Tobias Böhmelt

International Mediation Interaction

VS RESEARCH

Tobias Böhmelt

International Mediation Interaction Synergy, Conflict, Effectiveness

VS RESEARCH

1st Edition 2011

All rights reserved© VS Verlag für Sozialwissenschaften | Springer Fachmedien Wiesbaden GmbH 2011

Editorial Office: Dorothee Koch | Anita Wilke

VS Verlag für Sozialwissenschaften is a brand of Springer Fachmedien. Springer Fachmedien is part of Springer Science+Business Media. www.vs-verlag.de

No part of this publication may be reproduced, stored in a retrieval system or transmitted, in any form or by any means, electronic, mechanical, photo-copying, recording, or otherwise, without the prior written permission of thecopyright holder.

Registered and/or industrial names, trade names, trade descriptions etc. cited in this publicationare part of the law for trade-mark protection and may not be used free in any form or by anymeans even if this is not specifically marked.

Cover design: KünkelLopka Medienentwicklung, HeidelbergPrinted on acid-free paperPrinted in Germany

ISBN 978-3-531-18055-7

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie;detailed bibliographic data are available in the Internet at http://dnb.d-nb.de.

5

Acknowledgements The following book is based upon my PhD dissertation submitted to the Department of Government at the University of Essex, UK, in April 2010. Its completion took a significant amount of my time and I am indebted to more people than I can possibly list here. Nevertheless, first of all I want to thank those persons who had a direct influence on the contents of this manuscript. These firstly include my PhD supervisor, Kristian S. Gleditsch, as well as the members of my supervisory board, Hugh Ward and Vera E. Troeger. Without their expertise, helpful comments and support I could not have done this.

I also would like to thank those people who continuously provided me with advice and/or whom I have collaborated with on various papers. First and foremost, these are Ulrich H. Pilster and Spyros Kosmidis. In alphabetical order then follow Kyle Beardsley, Xun Cao, Nathan Danneman, Han Dorussen, Lawrence Ezrow, Natasha Ezrow, Vera Guill, Todd Landman, Nigel Lo, Nils Metternich, Daniel Millan, Andrea Ruggeri, Gerald Schneider, Willem Schudel, Thomas J. Scotto, Laura Seelkopf, and Chi-Mao Tang. In particular, however, I am indebted to Anja Neundorf who not only discussed with me the nature of research, science and academia on various occasions, but also supported and encouraged me in countless ways throughout my time in Essex.

As for the people who have influenced me as a person during the last three years, these are certainly too many to enumerate. A good way to sum this up, however, is to name – first among equals – my mother Christa followed by my siblings as well as my friends. Their continued help and support is also very well appreciated.

I am also grateful for the scholarships from the University of Essex and the Economic and Social Research Council (ESRC) for making all this possible in the first place as well as for a scholarship from Emory University that enabled me to work there as a visiting researcher during Fall Term 2009. Last but not least, I thank Anita Wilke from the editorial team at VS Research.

All errors and faults are my own.

Tobias Böhmelt

7

Abstract A specific set of factors, which are related to the belligerents, any third parties that (anticipate to) intervene or both of them, determine among other influences whether a conflict is likely to see mediation efforts in the first place and ultimately the prospects for successful mediation outcomes. However, although there is an extensive body of literature, qualitatively and quantitatively, on all these issues, the previous research rarely addressed the interaction between these actors in conflict and mediation attempts, i.e. how both belligerents and/or mediating parties are tied to each other and how they coordinate an intervention. Hence, inter alia, questions like “which kinds of relationships and ties between antagonists determine the likelihood of mediation onset” or “how does the interaction between multiple mediators affect successful conflict resolution” have not yet been entirely clarified – despite the fact that answering these questions not only helps to have a more precise understanding of mediation from an academic point of view, but is also likely to produce significant policy implications. In order to address these shortcomings, this book presents substantial chapters on four interrelated, albeit different issues of actors’ inter-actions in international mediation as a tool for settling conflicts peacefully.

Relying on a rationalist framework of conflict and actors as well as employing quantitative methods, the book demonstrates several important findings. First, mediation occurs according to the complex patterns of direct and indirect relationships between the disputants. While direct links, i.e. bilateral ties between the warring parties, are unlikely to facilitate mediation onset, indirect links that involve various intermediaries in a conflict seem to increase the chances that a dispute finally sees third-party mediation. However, depending on the type of these ties, the predicted effect of those interactions may not be that straightforward. More precisely, the book theoretically argues and empirically demonstrates that bilateral trade flows between antagonists significantly increase the chances of mediation occurrence, while trade links to parties outside this dyad actually decrease them. Second, most mediation attempts do not see one intervener, but a coalition of mediators instead. The following manuscript shows that this “multi-track diplomacy” or “multi-party mediation” and the interaction between those mediating parties is generally able to positively influence me-diation outcomes, but as soon as a coalition of mediators is disaggregated into its

8

distinctive micro foundations, it may well be that “too many cooks spoil the soup,” making mediation eventually less effective.

Accounting for these links and interactions between the conflicting and/or mediating parties, the findings demonstrate that – depending on the circum-stances – the chances of mediation onset may be higher than originally anticipated and also that mediation can be an effective instrument towards the peaceful resolution of conflicts. Also, while taking steps to mitigate selection problems and omitted variable bias, the results emphasize that traditional predictors of both mediation onset and effectiveness, such as dispute intensity or belligerents’ incentives, may play a far minor role than the previous literature may have suggested.

9

Table of Contents List of Figures................................................................................................................................. 11 List of Tables................................................................................................................................... 13 1. Introduction...................................................................................................................... 15 1.1. A Rationalist Approach to Conflict Management........................................................ 17 1.2. Overview of the Book Project...................................................................................... 19 2. Main Data Source: The International Conflict Management Data............................. 23 2.1. The International Conflict Management Data – An Overview.................................... 23 2.2. The International Crisis Behavior and SHERFACS Data – Alternative Data Sources......................................................................................................................... 25 2.3. The Utility of the ICM Data – An Introduction........................................................... 26 2.4. Conclusion.................................................................................................................... 30 3. International Mediation and Social Networks: The Importance of Indirect Ties................................................................................................................................... 31 3.1. Literature Review......................................................................................................... 32 3.2. A Theory of International Mediation and Social Networks......................................... 34 3.2.1. The Effect of Direct Links on Mediation Onset..................................................... 37 3.2.2. The Effect of Indirect Links on Mediation Onset.................................................. 39 3.3. Research Design........................................................................................................... 42 3.3.1. Data........................................................................................................................ 42 3.3.2. Operationalisation.................................................................................................. 43 3.3.3. Control Variables................................................................................................... 44 3.4. Empirical Findings....................................................................................................... 45 3.5. Conclusion.................................................................................................................... 50 4. The Impact of Trade on International Mediation......................................................... 53 4.1. The Relationship of Trade and Conflict – An Overview............................................. 55 4.2. International Mediation and Trade – Literature Review.............................................. 56 4.3. A Theory on International Mediation and Trade.......................................................... 58 4.3.1. An Opportunity Costs Model................................................. ................................ 61 4.3.2. The Impact of Bilateral Trade on Mediation......................... ................................. 62 4.3.3. The Impact of Trade Links to Third Parties on Mediation..................................... 64 4.3.4. The Impact of the Interaction of Bilateral Trade and Trade Links to Third Parties on Mediation.................................................................................

65

4.4. Research Design........................................................................................................... 67 4.4.1. Dependent Variable................................................................................................ 67 4.4.2. Explanatory Variables............................................................................................ 67

10

4.4.3. Control Variables................................................................................................... 68 4.4.4. Estimation Procedure............................................................................................. 70 4.5. Data Analysis............................................................................................................... 72 4.5.1. Empirical Findings................................................................................................. 72 4.5.2. Robustness Checks................................................................................................. 78 4.6. Discussion.................................................................................................................... 80 5. The Effectiveness of Tracks of Diplomacy Strategies in Third-Party Mediation...... 83 5.1. Literature Review: The Concept of Tracks of Diplomacy........................................... 85 5.2. The Effectiveness of Tracks of Diplomacy Strategies – Theoretical Model............... 87 5.2.1. Understanding Mediation Effectiveness............................................ .................... 87 5.2.2. Official Diplomacy – Track One............................................................................ 88 5.2.3. Unofficial Diplomacy – Track Two and Track-One-and-a-Half......................... 89 5.2.4. Multi-Track Diplomacy – Combined Efforts of Official and Unofficial Diplomacy.............................................................................................................. 91 5.3. Research Design........................................................................................................... 93 5.3.1. Methodology.......................................................................................................... 93 5.3.2. Data........................................................................................................................ 94 5.3.3. Operationalisation.................................................................................................. 94 5.3.4. Control Variables................................................................................................... 94 5.4. Results.......................................................................................................................... 96 5.5. Conclusion.................................................................................................................... 100 6. The Impact of Multi-Party Mediation in International Interventions: Synergy, Conflict, Effectiveness.................................................................................... 103 6.1. Literature Review......................................................................................................... 105 6.2. Theory: The Impact of Multi-Party Mediation – Size, Conflict, Cooperation and Regime Type......................................................................................................... 108 6.2.1. Size......................................................................................................................... 108 6.2.2. Conflicting and Cooperative Interactions............................................................... 110 6.2.3. Regime Type.......................................................................................................... 112 6.3. Research Design........................................................................................................... 113 6.3.1. Data........................................................................................................................ 113 6.3.2. Dependent Variable............................................................................................... 114 6.3.3. Explanatory Variables............................................................................................ 115 6.4. Empirical Findings....................................................................................................... 118 6.5. Robustness.................................................................................................................... 124 6.6. Discussion.................................................................................................................... 126 7. Conclusion......................................................................................................................... 129 7.1. Academic Contribution and Policy Implications......................................................... 131 7.2. Discussion of Research................................................................................................ 134 7.3. Avenues for Future Research....................................................................................... 135 Bibliography.................................................................................................................................... 137

11

List of Figures 3.1. Subset of the 1982 Interstate IGO Network......................................................................... 35 3.2. Constrained Subset of the 1982 Interstate IGO Network..................................................... 36 3.3. Predicted Probability of Mediation Onset by Network Measures....................................... 50 4.1. Subset of the 1958 Interstate Trade System......................................................................... 59 4.2. Constrained Subset of the 1958 Interstate Trade System..................................................... 60 4.3. Descriptive Statistics of Mediation Onset, Dyadic Trade and Trade Centrality.................. 75 4.4. Odds Ratio Estimates of Trade Centrality........................................................................... 77 4.5. Odds Ratio Estimates of Dyadic Trade................................................................................ 78 6.1. Expected Impact of Coalition Size on Mediation Effectiveness.......................................... 110 6.2. Inverted U-Shaped Relationship between Coalition Size and Mediation Effectiveness........................................................................................................................ 121

13

List of Tables 2.1. Frequencies of Mediation Onset.......................................................................................... 26 2.2. Frequencies of Mediation Effectiveness.............................................................................. 27 2.3. Cross Table of Conflict Duration and Dispute Fatalities..................................................... 28 2.4. Cross Table of Mediation Onset and Belligerents’ Power Discrepancies........................... 29 2.5. Cross Table of Mediation Effectiveness and Mediation Strategies..................................... 30 3.1. The Onset of International Mediation, 1946-1995.............................................................. 47 3.2. Predicted Probabilities of Mediation Onset......................................................................... 49 4.1. Descriptive Statistics of Variables....................................................................................... 70 4.2. Pairwise Correlations of Trade Variables............................................................................ 72 4.3. The Impact of Trade on International Mediation, 1948-1995............................................. 74 5.1. Descriptive Statistics of Variables....................................................................................... 95 5.2. The Effectiveness of Tracks of Diplomacy Strategies......................................................... 98 6.1. The Effectiveness of Third-Party Settlement Attempts, 1965-2000................................... 115 6.2. The Effectiveness of Multi-Party Mediation, 1965-2000................................................... 119 6.3. Predicted Probabilities of Mediation Effectiveness, 1965-2000......................................... 123

15

1. Introduction A consensus definition of mediation states that it is “a process of conflict management where disputants seek the assistance of, or accept an offer of help from, an individual, group, state or organisation to settle their conflict or resolve their differences without resorting to physical force or invoking the authority of the law” (Bercovitch, Anagnoson and Wille 1991: 8).1 In other words, belligerents that dispute over some issue(s) request the assistance of some third party that eventually agrees to intervene. More recently, the war between Israel and Lebanon (2006) or the conflict between Russia and Georgia (2008) have seen various mediation efforts; examples of successful mediations most prominently include the Good Friday Agreement (1999) for the Northern Ireland peace process or the Dayton Agreement (1995) settling the war in Bosnia. Nonetheless, international mediation is a costly way to solve conflicts peacefully (see e.g. Beardsley 2006; 2010). The warring parties have to agree on an intervener or a coalition of mediators, negotiate with this actor/these actors and ultimately lose some degree of control over the bargaining process to the mediator(s) (Hensel 2001). Against this background, it is apparent that both the combatants (demand side) as well as the mediator(s) (supply side) must have strong incentives both for mediation to occur in the first place and to effectively bargain in order to reach a satisfactory outcome (see Beardsley 2006; 2008; Beardsley and Greig 2009).

The previous research has examined this phenomenon to a large extent – both quantitatively and qualitatively – and was generally able to identify factors that are likely to influence if conflicts see mediation efforts at all and what determines if those are successful afterwards. With regard to the occurrence of mediation, existent work can be summarised under four interrelated clusters (Greig 2005). First, there is the likelihood of eventual mediation success, e.g. a dispute appears relatively easy to solve, which thereby could increase the chances that there is a third party willing to intervene. Second, the role of the previous conflict management history. This cluster examines whether any mediation attempts or other kinds of conflict resolution approaches have been employed during the dispute before and how this may affect the likelihood of

1 See also Touval and Zartman (1985) for a similar definition.

T. Böhmelt, International Mediation Interaction, DOI 10.1007/978-3-531-92812-8_1,© VS Verlag für Sozialwissenschaften | Springer Fachmedien Wiesbaden GmbH 2011

16

mediation incidence. Third, there is the question of how domestic and inter-national threats faced by the warring parties influence mediation onset. Finally, the characteristics of the conflict dyad, i.e. its previous relationships and ties towards each other, their regime types and/or capabilities, influence the chances of mediation as well.

With regard to mediation effectiveness, the existing literature frequently emphasizes three factors (Bercovitch and Langley 1993; Jackson 2000; Kleiboer 1996; Wall, Stark and Standifer 2001). The first one pertains to characteristics of the warring parties. The second cluster includes characteristics of the dispute, i.e. its intensity and duration or the issues at stake. The third factor describes the mediators as such or the type of mediation pursued.

Against this background, it may seem surprising that the previous research rarely addressed the interaction between the actors in conflict and mediation attempts, i.e. the warring parties and/or the mediator(s). I define interaction here as the interconnectedness and/or coordination efforts of both the belligerents and/or the mediating parties. However, neither the belligerents nor the intervener(s) are independent agents acting solely on their own and without any contact to each other before or during a conflict (e.g. Bercovitch and Schneider 2000; Frazier and Dixon 2006; Regan and Aydin 2006). For example, during the stage of negotiating the supply and demand of mediation, i.e. the warring parties and prospective mediators are trying to agree on an intervention mandate, and during the actual mediation bargaining process, the belligerents as well as the mediator(s) have to interact with each other and both the antagonists have to coordinate with the mediators and vice versa, thereby influencing on the development and effectiveness of other actors’ activities, which in turn is likely to shape the prospects not only of mediation incidence, but also of effective third-party conflict resolution.

In other words, inter alia, questions like “which kinds of relationships and ties between combatants determine the likelihood of mediation onset” or “how does the interaction between multiple mediators affect successful conflict resolution” have not yet been entirely clarified – despite the fact that answering those questions not only helps to have a more precise understanding of mediation from an academic point of view, but is also likely to produce significant policy implications (see also Beardsley and Greig 2009). First, recent research on mediation onset or the likelihood to reach effective outcomes uses a bargaining framework in order to understand the impact of various factors such as informational asymmetries, credible commitment problems and indivisible goods (see e.g. Beardsley 2006; 2008). Putting an emphasis on the interaction of belligerents and/or the mediator(s) sheds further light on these issues and helps us to better understand those bargaining and political processes.

17

Second and as indicated above, my research entails important policy impli-cations. As I will argue in the following, the likelihood of mediation onset and mediation effectiveness depends to a large extent on the interaction of the parties involved. Furthermore, if mediation generally is an effective instrument to restore peace, then both the actors involved in conflicts and those that seek to intervene for promoting peace will do well to know why mediation occurs in some disputes, but not in others. Further, my research then also sharpens their understanding of when they are likely to see a successful mediated settlement of a dispute. I will come back to these issues in each concluding section of the succeeding chapters.

Finally, an understanding of mediation occurrence is necessary for precise causal inferences of their effects, i.e. mediation effectiveness. Mediation usually is not randomly assigned, meaning that interveners strategically choose the conflicts they intend to mediate (Beardsley 2006; 2008). If this holds true, then any correlation between mediation incidence and effectiveness may only be a consequence of that mediation is chosen when certain outcomes – in particular effective ones – are more likely to occur anyway. Put differently, if mediation only happens in cases that are, for example, most difficult to resolve, then scholars will tend to underestimate the value that mediation actually has. In order to separate out the direct effects of mediation on outcomes, we need to know exactly which conflicts are likely to see mediation and why in the first place. 1.1. A Rationalist Approach to Conflict Management This book contributes to these areas by addressing the listed shortcomings. More specifically, in the following I will present four substantial chapters on inter-related, albeit different issues of actors’ interaction in international mediations as a tool for settling conflicts peacefully. Before I will outline the specific chapters in more detail, however, I briefly describe the underlying paradigm that I will employ in all succeeding sections more or less explicitly: a rationalist approach to conflict management (see Fearon 1995; Kydd 2010). In short, all parties act here according to rational calculations and preferences, i.e. by comparing costs and (prospective) benefits of certain actions, the parties will pursue that activity where the benefits outweigh the cost most. At the same time, however, disputes of any type impose costs on the parties involved. Since these costs would not occur if the warring parties would have pursued some kind of alternative peaceful patterns of behaviour in the beginning, conflict is therefore always an inefficient activity (Fearon 1995).

18

Consequently, in order to settle a conflict peacefully, an intervener must have a precise understanding why it happened or why this dispute continues, since the belligerents should already know in the first place that other, less costly avenues for behaviour were available. Fearon (1995) lists three causes of conflict. First, there are problems of issue indivisibility, i.e. actors fight over an issue that cannot be divided among the antagonists. This scenario appears rather unlikely, however, since most if not all issues in the international system are indeed divisible or actors could be compensated through side payments (Fearon 1998). Second, actors face uncertainty due to private information. In more detail, states, for example, have information about their own capabilities that the other side has not. In bargaining situations then, the warring parties might be tempted to deliberately transmit this information wrongly in order to strike a better deal, i.e. there are incentives to misinterpret private information. Finally, conflicts occur as a consequence of commitment problems. Although the belligerents might actually be able to reach a mutually preferable bargain, ultimately it may be unattainable because one or more of the concerned states would have an incentive to renege on the agreed terms.

Against this background, both demand and supply-side incentives exist that argue for mediation, since mediators are generally able to address all of these rational causes of conflict and, hence, are more likely to lead to an effective outcome compared to a situation without third-party engagement (Young 1967; Dixon 1996; Walter 1997; 2002; Wilkenfeld, Young, Asal and Quinn 2003; 2005; Kydd 2010). First, intervening actors can help show the disputing parties alternatives more beneficial than continued fighting by creating “turning points” and providing incentives (Crumm 1995; Filson and Werner 2002; Beardsley, Quinn, Biswas and Wilkenfeld 2006: 67; Regan and Aydin 2006). Second, third parties can decrease uncertainty by raising the flows of information between the antagonists and provide security guarantees (Walter 1997; 2002). Third, interveners can help overcome commitment and cooperation problems by imposing order, either by directly implementing peace agreements or increasing the costs of defection (see Doyle and Sambanis 2000: 781). Finally, and more directed at the perspective of the intervener(s), states might have incentives to mediate, since they could derive benefits out of their settlement attempt, prevent the spread of externalities that pose a threat towards their own security or due to humanitarian reasons. I will elaborate on all of these issues again and in more detail in the following chapters.

Frazier and Dixon (2006: 391) emphasize here correctly that “many scholars have found that all managers have had some degree of success in ameliorating conflict. The question that logically follows is which ones are more effective in doing so? More importantly, why are they more effective?” My book intends to

19

contribute to answering this important question as well as why mediation occurs in the realm of international mediation interaction in the following chapters.

1.2. Overview of the Book Project After a brief introduction to the main data source that will be used for most of my analyses, the next two chapters examine the conditions of mediation onset. Although it may seem that mediation effectiveness, the focus of Chapters 5 and 6, is “more important” than the actual incidence of mediation, I argue that understanding when mediation occurs in the first place is also fundamental to understanding when it is most likely to be successful. Put differently, a better knowledge of mediation onset may well translate into more effective mediation and avoid the consequences of failed efforts. Hence, the first substantial chapter extends the previous literature on actors’ interaction in conflict management by examining how direct and indirect links between fighting parties influence the prospects for mediation onset from a social network perspective. I argue, contrary to existing research, that direct links, i.e. bilateral ties between warring parties do not notably increase the likelihood of mediation. It is rather indirect links, i.e. ties that connect two states via one or more than one third party to each other, which are more likely to determine whether a conflict sees mediation. Although direct links can indicate mutual interests, shared preferences and decreased uncertainty between the warring parties, these ties either lose importance or become cut off when states enter a conflict. By contrast, indirect ties create a social network that involves outside parties into the dispute process, increase the exchange of information between belligerents and potential mediators, and the chances that third parties will have a vital interest in intervention. The empirical analysis using conflict and social network data from 1946 to 1995 provides strong support for my theory.

Second, based upon an opportunity costs model, the next chapter extends earlier research by examining how trade both between belligerents and between belligerents and third parties influences mediation onset. First, dense bilateral trade between combatants is characterised by high opportunity costs, since it would be more expensive to seek alternative markets and either belligerent is highly vulnerable to sanctions from the dyadic counterpart. Hence, fighting parties have an incentive to limit their hostile behaviour and to settle their conflict through mediation. Second, I will argue that belligerents’ trade with third parties decreases the likelihood of mediation. Combatants with multiple other alternatives for trade are not only able to mitigate the consequences of sanctions, but also to substitute for markets which are lost due to the outbreak of

20

a conflict. Finally, the nature of these different trade ties implies that opportunity costs of bilateral trade lower the opportunity costs of trade links to third parties and vice versa. Taking this interaction into account, this chapter demonstrates that the impact of bilateral trade varies conditional on trade ties to outside states.

Third, international mediation is not conducted solely by official actors such as states or international organisations. Non-official parties such as individuals and non-governmental organisations increasingly intervene as third-party mediators in conflicts. Both official and unofficial interventions are conceptualized as tracks of diplomacy (ToDs). Even though there is a vast quantitative literature on international mediation, most studies focusing on ToDs have been qualitative and produced few generalisable insights. This article extends the existing literature on third-party intervention by developing a theoretical model to explain the effectiveness of different and interacting ToDs interventions, which is then empirically tested in a large-N framework. The find-ings indicate that the leverage and resources of ToDs determine outcome effectiveness and demonstrate that the specific type of mediating actor and mediation interaction seem highly important. Track-One-Diplomacy, i.e. efforts by official actors, tends to be the most effective form of intervention as greater leverage and more resources invested can make track intervention more effective. Most importantly, however, I find evidence that combined mediation efforts of both official and unofficial tracks can be more effective than independent track actions. Since conflicts with mediation are unlikely to be a randomly selected set, a selection estimator is used to test the hypotheses on effectiveness in this chapter.

Finally, when studying the effectiveness of international mediation, the previous literature has largely ignored that multiple states intervene, and how they are tied to each other and coordinate their intervention. In addressing this shortcoming, I will examine how mediation outcomes are influenced by the size of an intervening group, conflicting relationships and cooperative interactions, as well as interveners’ regime types. I argue, first, that the size of an intervening coalition and mediation effectiveness are characterised by an inverted U-shaped relationship. Although more mediators are generally able to create synergy, a very large number of interveners increases the chances that coordination and collective action problems exist, hence lowering the margins of effective mediations. Second, interveners sharing a history of conflict and distrust will convey this relation into a mediation attempt, making it less effective. Consequently, states sharing friendly and cooperative ties with each other are more successful in managing conflicts. Finally, a coalition of mediators that is largely democratic should be more effective due to inclusivity, a shared culture of peaceful conflict resolution, and increased communication flows. Using data

21

from the Issues Correlates of War Project from 1965 to 2000, the empirical analysis largely provides support for the theory.

23

2. Main Data Source: The International Conflict Management Data

The focus of this manuscript is on belligerents’ and third parties’ interactions in international mediation efforts. In order to test my theoretical arguments quantitatively, data are necessary that provide comprehensive and detailed information on those actors as well as different aspects of conflict management and the dispute as such. The data on conflict and peaceful settlements that I use for most of the succeeding chapters are taken from the International Conflict Management (ICM) data (Bercovitch 1999).2 In the following, I will provide an overview of the ICM data set, outlining its strengths and weaknesses. I then briefly describe two alternative data sources that are commonly used for the study of international mediation and present reasons why I do not consider them for my analyses. Third, I frame the utility of the ICM data in more detail, providing statistics of my dependent variables – mediation onset and effectiveness – and simple inferential results. The last section briefly concludes. 2.1. The International Conflict Management Data – An Overview The ICM data set is a chronological account of international disputes and their peaceful management attempts from 1945 to 1995 (Bercovitch 1999: 2). Its primary aim is to “both further our understanding of mediation and facilitate the comprehensive investigation of different conflict management mechanisms” through rigorous empirical investigation along sound theoretical lines (Bercovitch 1999: 2).

Here, international conflict is defined as organized and continuous militarized conflict, or a demonstration of intentions to use military force involving at least one state (Bercovitch, Anagnoson and Wille 1991). This case selection criterion entails the advantage that the observations are not only constrained to wars, militarized interstate disputes or specific fatality thresholds as in other commonly used data sets, but include low-level disputes that may even involve non-state actors (Bercovitch 1999: 3). Further, this approach increases the 2 Chapter 6 is the only exception. I will explain in detail there why I employ alternative data.

T. Böhmelt, International Mediation Interaction, DOI 10.1007/978-3-531-92812-8_2,© VS Verlag für Sozialwissenschaften | Springer Fachmedien Wiesbaden GmbH 2011

24

number of observable implications of my theoretical claims, thereby enhancing the generalisability of empirical findings. A total of 295 disputes are identified in the period in question.3 These conflicts in turn experienced 3,207 different conflict management attempts, including negotiation, mediation, arbitration, referral to international institutions or multilateral conferences. Out of these, 1,859 were mediation of the form I focus on. The unit of analysis can vary according to the interest of the researcher, although the ICM’s default is a specific conflict management attempt or a conflict if there was no peaceful settlement at all. In the following chapters, I choose the non-directed conflict dyad with or without specific third-party resolution efforts as the most appropriate unit of analysis for the research questions addressed in the book.

The original underlying logic for identifying cases and obtaining information on specific variables builds upon the contingency model developed by Fisher and Keashley (1991; see also Fisher 2007). In short, this approach claims – in consistence with the rational framework of conflict management as outlined above – that the onset of conflict management efforts and their effectiveness depend on a number of contextual and process variables. More specifically, the ICM data relies on established dispute lists as well as primary and secondary sources such as Keesing’s or the CIA Factbook in order to obtain information on variables for the following three categories (Bercovitch 1999: 6ff).4 First, there are the characteristics of the parties involved, i.e. their identity, capabilities or relationships with each other. Second, the ICM provides detailed information on dispute characteristics such as issues, fatalities and outcomes. Finally, the data include conflict management agents and their activities, i.e. answers are provided for the questions if third parties intervened for ending a conflict peacefully (or in the case of bilateral negotiations: the belligerents themselves) and how they managed a settlement attempt. Note that these three categories perfectly match with those clusters identified by Kleiboer (1996), Wall, Stark and Standifer (2001) and Greig (2005) that are likely to play a major role for mediation effectiveness and onset: the characteristics of the disputants, the characteristics of the dispute and factors that describe the mediators as such or the type of mediation.

Against this background, the main advantages of these data are that the ICM covers conflict cases, which were selected on the basis of maximal comparison throughout different periods in the international system. Further, the data provide a diverse set of dispute actors as well as intervening parties that are not constrained to nation states only and vary in terms of age, regime type or de-

3 See Bercovitch and Jackson (1997) for a description of 292 of these disputes. 4 See Bercovitch (1999: 5) for a detailed and comprehensive list of all sources.

25

velopment. Finally, all cases in the data were cross checked over various sources and different coders, achieving inter-coder reliability of about 95% (Bercovitch 1999: 11). Hence, the confidence in the data’s validity is likely to be very high and the remaining noise should pose no major problems for the empirical analysis. A shortcoming of the ICM, however, is that much of its information on conflict management rests on ex post facto reflections or data collection of cases that were mentioned in public sources only (Bercovitch 1999: 4). Hence, management efforts that were carried out “behind closed doors” did not enter the data. It should be noted, nevertheless, that a more comprehensive or accurate measurement is not given for alterative data sets either. Finally, the final point of observation is about 14 years ago. Although an update of the data is therefore clearly necessary, the limited period of time should not be an obstacle to generalise my findings. 2.2. The International Crisis Behavior and SHERFACS Data –

Alternative Data Sources Although the vast majority of the following chapters rely on the ICM data, two other data sets exist that are prominently used for studying international mediation: the International Crisis Behavior (ICB) project (Brecher and Wilkenfeld 2009) and the SHERFACS data (Sherman 1995).

First, the ICB project defines a crisis along three necessary conditions. First, an international crisis requires that a state perceives another country as a threat to its basic values. Second, there needs to be a heightened perception of the threat of military hostilities. Finally, a crisis requires that the states face a finite time horizon in which an agreement has to be reached. The ICB data contain information on 452 international crises, 35 protracted conflicts and 994 crisis actors from the end of World War I through 2006 and have easily accessible case summaries to allow for more precise coding of who mediates in each crisis. In other words, although the coding procedures provide the same advantages as the ICM, the time span is significantly longer. However, the unit of analysis in the ICB is an interstate crisis, while the variables here do not vary over the course of such a dispute. Hence, due to this aggregation, it is likely that valuable information is lost, since many aspects of a conflict such as its intensity, the intervening parties or mediation techniques may vary over its course. This is the primary reason for arguing against the ICB and preferring the ICM data.

Second, unlike the ICB, the SHERFACS project provides intra-conflict information, as the unit of analysis is a distinct phase in each conflict. SHERFACS defines a conflict as “a situation where there exist mutually

26

exclusive sets of competing claims or challenges to sovereignty between at least two actors, one of which must be an internationally recognized political actor” (Sherman 1995: 11). In total we obtain 1,446 international conflict phases from 1937 to 1985. Therefore, although SHERFACS offers the advantage of a disaggregated study of conflicts, its time span is too limited, since the data collection process terminated more than 20 years ago.

To recap, both alternative data sets claim to have a similar purpose of analysis as the ICM project. Consequently the set of variables covered is essentially the same. However, while the ICB data aggregates its conflict information, the SHERFACS data go until 1985 only. The ICM data set on the other hand circumvents both problems, as it disaggregates disputes into specific conflict management attempts and has a longer period of study. 2.3. The Utility of the ICM Data – An Introduction Although the ICM data have been employed for a series of publications (e.g. Bercovitch and Langley 1993; Bercovitch and Houston 2000; Bercovitch and Schneider 2000; Böhmelt 2009), this section provides a few examples of how the ICM data might be used in detail. I will demonstrate a sample of descriptive statistics and simple inferential results, but I do not attempt to provide exhaustive empirical analyses here. This is done in the succeeding chapters. To start, Table 2.1. shows the descriptives for one of the two dependent variables examined in the following chapters: mediation onset.

Table 2.1. Frequencies of Mediation Onset.

Frequency Percent No Mediation 1350 42.10 Mediation 1857 57.90

Total 3207 100.00

As Table 2.1 demonstrates, about 58% of the ICM’s cases saw third-party

mediation, while about 42% did not experience such peaceful settlement attempts. Hence, there is certainly enough variation on that variable in order to employ it in empirical analyses. Note, however, that the zeros of Mediation Onset do not only include cases that did not experience any conflict management at all, but also non-mediation techniques such as bilateral negotiation, arbitration

27

or referrals to international organisations. I will come back to this issue in the following chapters.

Table 2.2. Frequencies of Mediation Effectiveness.

Frequency Percent No Management 70 2.18 Offered Only 192 5.99 Unsuccessful 1691 52.73 Cease-Fire 315 9.82 Partial Settlement 824 25.69 Full Settlement 115 3.59 Total 3207 100.00

Further, Table 2.2. shows the variation of the second dependent variable I

will study in this book: mediation effectiveness. 2.18% of all disputes did not involve any conflict management. The Shelling Incident between China and Taiwan in 1994 as such a dispute. 5.99% had mediation offers that were not accepted by the disputing parties. For example, after the invasion of the Hunish islands in November 1995, multiple actors offered Eritrea and Yemen to mediate. But these initial offers at the outbreak of the dispute were rejected. 52.73% of the conflict cases saw mediation action where conflict management was ultimately unsuccessful. Mediators of both official and unofficial nature have intervened on multiple occasions during the Caucasus conflict between Russia and Chechnya, but the dispute remained unsettled in the data set’s period of time. 9.82%, such as the Ecuador-Peru border conflict in 1995, ended in a cease-fire, while 25.65% of the ICM’s cases were partially settled, such as the Uganda-Kenya border dispute caused by the Ugandan civil war and subsequent refugee influx in 1987. Only 3.59% were effective in the sense of reaching a stable, peaceful settlement. The Saudi Arabia-Qatar border incident after the Gulf war in 1992 provides an example of this last category. Hence, Table 2.2. shows that as many as 60% of the conflicts could not even be resolved partially (Bercovitch and Regan 1999) – although, as illustrated in Table 2.1., 58% of the conflicts actually saw third-party mediation.

Third, in order to further uncover the utility of the ICM, I will look at some basic inferential statistics of three commonly cited arguments in the mediation literature. The ICM provides, for example, data on conflict duration and dispute fatalities. The literature argues that outside parties are less likely to intervene in long-standing conflicts that have “witnessed a long history of mutual attack and

28

atrocity” (Touval and Zartman 1985; Bercovitch and Langley 1993; Kleiboer 1996: 362f; Pearson 2001: 276). Further, Bercovitch, Anagnoson and Wille (1991: 13), Bercovitch and Gartner (2006b) and Quinn, Wilkenfeld, Smarick and Asal (2006: 454) emphasize that high-intensity disputes are more likely to see mediation. Within this context it is plausible to claim that the longer a dispute, the higher its intensity or the higher the number of fatalities (e.g. Greig and Regan 2008). Table 2.3. shows a cross tabulation of these two variables along with a test.

Table 2.3. Cross Table of Conflict Duration and Dispute Fatalities.

Fatalities (grouped)

Duration (grouped) 0-500 501-1000 1001-5000 5001-10000 10000+ Total

0-1 118 11 8 37 51 225 1-3 71 28 70 54 0 223 4-6 68 22 5 2 1 98

7-12 132 6 44 0 20 202 13-24 117 0 89 0 19 225 25-36 23 1 11 0 70 105

36+ 118 80 162 74 1686 2120

Total 647 148 389 167 1847 3189

Note: 24)=2.0e+3; p<0.00.

Table 2.3. demonstrates that there is a strong statistical relationship between the duration of a dispute and its intensity. Although the value does not tell us much about the direction of this relationship, the table entries indicate that the longer a dispute, the higher its intensity. As stated above, however, more sophisticated methods of analysis will be provided in the remainder of this manuscript.

Further, I claimed above that the non-directed conflict dyad is the most appropriate unit of analysis for my purposes. The following example demonstrates this, using variables on mediation onset and the power discrepancies of warring parties. Generally, the literature argues that belligerents approximating parity are more likely to accept mediation, i.e. when neither side has a clear military advantage (Ott 1972; Young 1967; Bennett and Stam 1996: 242; Mason and Fett 1996: 550). Power discrepancies decrease the incentives for the most powerful party to agree to a peaceful settlement as it stands a better chance of winning a conflict militarily. The ICM provides absolute differences

29

of the Cox-Jacobson scale of power relations (Bercovitch 1999: 44f).5 Table 2.4. illustrates this relationship with a cross table. First among others, there is a strong statistical relationship between the onset of international mediation and disputants’ differences in capabilities. The value is highly significant. Further, mediation seems to occur more often in low or medium levels of power discrepancies, hence providing initial evidence that there might be indeed a negative relationship between these two items – as claimed by the literature (e.g. Greig and Regan 2008).

Table 2.4. Cross Table of Mediation Onset and Belligerents’ Power

Discrepancies.

Mediation Onset Power

Discrepancies No Yes Total Low 404 588 992

Medium 404 729 1133 High 542 540 1082

Total 1350 1857 3207

Note: 2)=48.42; p<0.00.

The final example pertains to the disaggregated nature of the ICM data. Beardsley et al. (2006) argue that more coercive mediation strategies may ultimately be more effective in settling conflicts peacefully. Although disputing parties may not be convinced about peaceful solutions at the outset, coercive and enforcing efforts can make it harder for belligerents to dismiss these possibilities (see Beardsley et al. 2006). The authors make the distinction between facilitation, formulation and manipulation, where facilitation is the least coercive approach and manipulation the most enforcing. However, mediation strategies may vary over the course of a dispute if interveners change or simply want to use a more promising strategy depending on the circumstances. In consequence, if mediation data is aggregated, this information is likely to be lost. The ICM project does not have this disadvantage.

5 The Cox–Jacobson power index score is similar to the Composite Index of National Capability of the Correlates of War Project (Singer 1988). Its detailed calculation will be introduced in one of the succeeding chapters.

30

Table 2.5. Cross Table of Mediation Effectiveness and Mediation Strategies.

Mediation Effectiveness Mediation Strategies

Offered Only

Unsuccessful Mediation

Cease- fire

Partially Settled

Fully Settled Total

Offered Only 3 0 0 0 0 3 Facilitation 0 36 15 5 0 56

Formulation 0 18 6 14 0 38 Manipulation 0 54 17 16 1 88

Total 3 108 38 35 1 185

Note: 12)= 198.27; p<0.00.

Table 2.5. shows a cross table between mediation strategies and the outcomes of third-party mediation attempts. For simplicity, however, this table is constrained to the 185 mediation efforts during the Yugoslavian Civil War (since 1989). Table 2.5. emphasizes that mediation strategies highly differ during this particular dispute. While manipulative strategies have been employed for most of these interventions, formulation tactics have been used the least. Further, there is a statistically significant relationship between mediation strategies and mediation effectiveness. The table entries indicate that this relation is positively signed: more coercive approaches induce more successful outcomes. 2.4. Conclusion This section introduced the main data source for the following chapters. As shown, the ICM project provides an extensive and systematic data set on international mediation, its outcomes and the related conflicts. I have argued and demonstrated via a few examples that although alternative data sources do exist, the ICM is likely to prove most beneficial due to its disaggregated nature, the comparably large time span as well as the vast amount of variables included in the data. Building upon that data, the following chapters seek to analyse mediation in more detail now.

31

3. International Mediation and Social Networks: The Importance of Indirect Ties

In this chapter I extend the previous research on actors’ interaction in conflict management by examining how direct and indirect links between fighting parties influence the onset of international mediation.6 The literature on mediation frequently puts an emphasis on the demand-side incentives by arguing that stronger or more friendly relations between belligerents make the onset of mediation more likely (e.g. Deutsch 1973; Bercovitch and Houston 2000: 180f; Greig and Diehl 2006). The existing research on third-party conflict resolution generally examines this argument through direct links, i.e. bilateral ties. However, the theoretical reasoning for focusing on these kinds of ties is incomplete since it does not consider links to potential mediators. Also, the empirical support for the expectation that direct links facilitate mediation onset is not as strong as often assumed. Many conflict dyads either have no or few direct ties, yet see multiple mediation attempts. Israel and Egypt, for example, hardly had strong diplomatic contact, significant trade or other sorts of direct relations in the 1970s, but their various disputes over the Sinai Peninsula experienced many mediation efforts.

Previous studies have largely ignored that belligerents are typically interconnected through indirect ties. Indirect links, as established for example through the network of intergovernmental organisations (IGOs) (Dorussen and Ward 2008a), connect the disputing dyad to multiple other actors and may thereby create connections between belligerents even if direct links are either absent or torn apart. Unlike direct links, these ties incorporate both demand and supply-side influences, since the preferences of the combatants in the context of their relationship with outside actors are taken into account here.

6 I rely on the definition of mediation as stated in the introduction, i.e. mediation is “a process of conflict management where disputants seek the assistance of, or accept an offer of help from, an individual, group, state or organization to settle their conflict or resolve their differences without resorting to physical force or invoking the authority of the law” (Bercovitch, Anagnoson and Wille 1991: 8). However, in this chapter I focus only on state networks due to the limited availability of data and the fact that states remain “the most successful and enduring forms of social and political organization” (Bercovitch and Schneider 2000: 147).

T. Böhmelt, International Mediation Interaction, DOI 10.1007/978-3-531-92812-8_3,© VS Verlag für Sozialwissenschaften | Springer Fachmedien Wiesbaden GmbH 2011

32

By focusing on indirect links and the application of social network analysis, my approach overcomes the limitations in previous research. I argue that indirect, rather than direct, links between belligerents determine whether third parties are more likely to manage a conflict. Although direct links may indicate mutual interests, shared preferences and decreased uncertainty between the warring countries, these links either lose importance or become cut off as soon as parties enter a conflict. By contrast, indirect ties create a social network that directly involves outside parties into the dispute process. This will increase the exchange of information between belligerents and potential mediators as well as the chances that third parties have a vital interest in intervention. In other words, indirect links may substitute for direct ties. This in turn implies that existing direct relationships between the fighting parties may not necessarily be the most influential ones in determining the likelihood of mediation onset.

In the following sections of this chapter, I first review the relevant literature on actors’ incentives for mediation, emphasising the interconnectedness or interaction of disputants. I then develop a theoretical rationale combining international mediation, social networks and belligerents’ interconnectedness through direct and indirect ties. The subsequent sections detail the research design and the empirical test of the theory. The chapter concludes with a comprehensive discussion about the findings and implications of my research. 3.1. Literature Review Previous research has paid a great deal of attention to examining factors that may influence mediation effectiveness (Kleiboer 1996; Wall, Stark and Standifer 2001), but far less work has been done on when we are likely to see mediation attempts. For example, Terris and Maoz (2005: 565) note that “we know very little about the conditions under which mediation may be offered or accepted.” As outlined in the introduction, existent work can be summarised under four interrelated clusters (Greig 2005: 251): first, the likelihood of eventual mediation success; second, the role of the previous conflict management history; third, how domestic and international threats faced by the warring parties influence mediation onset; and fourth, the characteristics of the conflict dyad.

This study seeks to extend our knowledge in the latter area by considering belligerents’ interaction through both demand and supply-side incentives via the role of direct and indirect links. Bercovitch, Anagnoson and Wille (1991: 12) emphasize how “previous friendly relations between the [warring] parties facilitate mediation and increase its chances of success” (see also Deutsch 1973). Their study finds empirical support for this argument, but they focus on demand-

33

side incentives for mediation, i.e. direct ties, only. The authors also use a rather abstract measure based on whether the belligerents shared “no, friendly, antagonistic or hostile” relations prior to the dispute. Bercovitch and Schneider (2000) demonstrate in a follow-up paper that good, direct links between disputing parties as well as to actors who have mediated in the past make it more likely that a conflict will see mediation. This study makes a substantial contribution to our understanding of mediation, although it falls short of the stated goal of fully establishing “the circumstances under which the conflict parties agree to a mediation at all” (Bercovitch and Schneider 2000: 163). Bercovitch and Houston (2000) seek to address these shortcomings by considering all four clusters that may determine the onset of mediation. They emphasize the importance of the disputants’ relationship, arguing that strong, friendly, direct links between the fighting parties promote mediation through the focus on common interests and a regional order (Bercovitch and Houston 2000: 179). Extending this relationship through the analysis of links to (prospective) mediators, the authors show that these ties encompass greater legitimacy, leverage and authority. Their empirical findings indicate support for their claims. Given their rather abstract measures of the alignment of belligerents and mediators, however, the study does not go very far towards incorporating supply-side incentives, or more particularly indirect relationships and fails to capitalise on the advantages of social network analysis – which is able to examine ties involving more than one intermediary.

Greig (2005) focuses on the major power and former colony status of the warring parties, political institutions, contingency and recent policy changes as the “characteristics of the conflict dyad” that may influence the prospects for mediation. The author also argues that the history of the fighting parties’ relationship strongly influences whether combatants will have an incentive for mediation. However, Greig (2005) is hesitant to draw strong conclusions since the results are not statistically significant for most of his models.

More recently, Savun (2008) finds support for her argument that biased mediators increase the chances of successful mediation (see also Kydd 2006). Those outside actors have more incentives to intervene due to substantial information about conflict/war capabilities. Although she focuses on mediation effectiveness, Savun (2008) also claims that the same factors that determine mediation outcomes will influence whether we see mediation in the first place.

Greig and Diehl (2006) provide one of the most extensive studies on disputants’ interconnectedness and mediation. They argue that weak direct ties or hostile pre-conflict relations between belligerents undermine the willingness to talk and therefore the chances of mediation onset. Further, “positive contacts [will] ... soften the attitudes of each side and eventually allow them to see

34

peaceful, rather than hostile solutions to their differences” (Greig and Diehl 2006: 367). Greig and Diehl (2006: 371) measure the direct interconnections between belligerents by the number of shared IGO memberships. Their findings are inconclusive, however, as their direct links variable does not have a significant impact on mediation onset. The authors are unable to offer a “coherent explanation for this finding” (Greig and Diehl 2006: 377).

Although not addressing social network analysis, the study of Crescenzi, Kadera, Mitchell and Thyne (2008) is one of the few exceptions where the authors do not solely focus on direct ties between disputants, but emphasize systemic effects instead. Crescenzi et al. (2008: 11f) argue that information flows in the international system become more accurate and credible as the global community is becoming more democratic, which in turn makes mediation more likely. The authors find support for this claim and also that shared IGO memberships between the warring parties promote mediation onset. Their analysis does not fully capture indirect links as they use an average global democracy score in order to capture systemic characteristics. Also, the study is limited to observations in the Western hemisphere solely.

Finally, Dorussen and Ward (2008a) outline a theoretical argument from a social network perspective for why indirect links may promote mediation occurrence. Strong and dense links to outside parties are likely to “hold relevant information and to be better able to convey it to the parties” (Dorussen and Ward 2008a: 190). Their subsequent empirical test does not directly consider this argument, however, as the authors only examine conflict onset.

In sum, when studying the interaction of belligerents, the previous work has focused on direct links, i.e. demand-side influences and largely ignored the existence of indirect ties between disputants that may form both demand and supply-side incentives. When the latter have been considered, however, they have so far been studied only through rather abstract measures, without capitalising on the tools and insights of social network analysis. Moreover, existing findings have often failed to comply with the original theoretical framework. Towards this end, a coherent theory that is able to address these shortcomings through social network analysis is needed.

3.2. A Theory of International Mediation and Social Networks A social network is defined by a finite set of nodes, i.e. actors, and by the links, i.e. relationships, that tie two or more actors to each other (Wassermann and Faust 1997: 20; see also Maoz 2009). Shared IGO memberships, trade relations and other sorts of contacts can all be considered links between countries.

35

Figure 3.1. Subset of the 1982 Interstate IGO Network.

36

Figure 3.2. Constrained Subset of the 1982 Interstate IGO Network.

However, not all links in a network are created equal. Some ties are likely to promote mediation onset, while others may not necessarily facilitate it. On one hand, the belligerents may share direct links, i.e. dense and short connections between two states. For the purpose of illustration, Figure 3.1. shows a subset of the interstate IGO network in 1982. Most states are connected through direct links. The width of the edges is proportional to countries’ shared IGO membership: wider or more numerous ties are represented by denser and stronger links. Further, countries that have similar strong direct links are closely grouped together by adjusting to the nearest Euclidean distance. This is reflected in the clear cluster of South American countries as well as the outlier positions of North Korea or the Solomon Islands. As Figure 3.1. shows, although many countries have strong direct links, such as the U.S. and the UK, i.e. 54 joint IGO memberships, other states may have no or only weak direct ties. The UK and

37

Paraguay, for example, are connected through much weaker links (25 joint IGO memberships), whereas North Korea and the Solomon Islands have no direct link at all.

In addition to direct ties, warring parties may also share indirect links, i.e. ties to one or more than one intermediary that connects the fighting dyad. For example, although North Korea and the Solomon Islands have no shared IGO memberships with each other in Figure 3.1., the former is (weakly) connected to the UK, which in turn has a link to the Solomon Islands. Hence, North Korea and the Solomon Islands have an indirect link to one another via the UK as an intermediary. Furthermore and as indicated, social network analysis offers the possibility to analyse longer chains involving more than one intermediary. For instance, North Korea is linked to the Solomon Islands also via the link to Paraguay, which in turn is linked to Peru. The latter is connected to Brazil, which in turn is tied to the U.S. The U.S. completes the indirect linkage between North Korea and the Solomon Islands through its links to the latter. The constrained graph of this indirect interconnection between North Korea and the Solomon Islands via various third parties is shown in Figure 3.2.

Links in social networks, both direct and indirect ones facilitate transmitting information about interests and intentions and promote a common understanding. Links that are shorter and denser convey information with lower costs, more precisely and faster. This ultimately leads to decreased uncertainty and more trust as well as mutually accepted norms (Hafner-Burton and Montgomery 2006; Dorussen and Ward 2008a: 194f). According to this mechanism, it is generally plausible to assume that stronger and more efficient information flows between disputants lead to a higher likelihood of mediation, since conflict resolution depends on the “credible and truthful transmission of information” (Dorussen and Ward 2008a: 193; see also Greig and Diehl 2006: 363ff). At this point, however, we do not have any theoretical grounds to expect that indirect links should have the same effect as direct ties. 3.2.1. The Effect of Direct Links on Mediation Onset The circumstances of mediation are different from those of conflict onset, since mediation is usually employed during ongoing hostilities and conflict has already broken out over one or more issues. This in turn makes it less likely that direct links will have much influence on mediation onset, if the main function of direct links is to convey information and to decrease uncertainty.

First, most, if not all aspects, that led to a conflict become common knowledge after conflict occurrence (see Slantchev 2003). States’ behaviour over

38

the course of a dispute reveals their preferences, which decreases uncertainty significantly relative to the pre-conflict situation (see Wagner 2000). Hence, after the onset of a dispute, fighting parties know what the conflict is about, why it was initiated and there is less doubt on the other state’s objectives. Any direct connections between disputants are therefore unlikely to increase information or contribute anything substantial that will make mediation more likely once a conflict has broken out.

Second, conveying information through direct links is seen as helpful for enhancing cooperation between countries, since these ties develop a mutual understanding, common norms and good relations. These features are less likely to apply after the onset of a conflict, however, when disputing parties seek to maximise their own benefits over the opponent and considerations such as norms or a history of good relations are less likely to weigh heavily on the belligerents. For example, Britain had fairly good relations with Germany before World War I. The German Emperor Wilhelm II was a grandson of Queen Victoria, both countries were able to reach an agreement on the share of colonies in 1913 and also settled their differences on the Baghdad Railway in 1914. However, after the outbreak of war between Great Britain and Germany, neither side referred to their previous good relations, but concentrated on the war effort instead.

Third, mediation is a costly way to solve conflicts peacefully (Beardsley 2006; 2010). The warring parties have to agree on a third party, negotiate with this intervener and ultimately lose some degree of control over the bargaining process to the mediator (Hensel 2001). Thus, there is no a priori reason why warring parties should agree to mediation as a conflict resolution instrument when strong direct ties endure into conflict. In this case, other less costly conflict management techniques such as bilateral negotiations may be feasible and the antagonists should be likely to obtain a resolution of the conflict without intervention by a third party (see also Jackson 2000). Saudi Arabia and Yemen entered a border conflict in 1994, for example, and both countries shared reasonably strong direct ties with each other before. The belligerents did not seek mediation for settling their dispute, but engaged in bilateral negotiations instead.

Finally, my previous claims assume that direct links are retained during disputes. However, direct ties between parties tend to be torn apart when states engage in conflict. Once “deactivated,” any prior links are unlikely to promote mediation. For example, although the UK and Argentina shared 39 direct ties through IGOs and had various other sorts of contact through diplomatic ex-change in 1982, Britain cut all diplomatic relations to Argentina within 24 hours after the beginning of the Falklands War. Thus, even if one of the conflict countries might have wanted to approach the other party to discuss possible

39

mediation efforts and their specific terms, there would have been few direct avenues for doing so.

Summing up, ties in social networks convey information and thereby decrease uncertainty between belligerents. Direct links are certainly not an exception to this rule. However, these ties either lose their importance or become cut off once states enter a conflict. For these reasons, I formulate the first hypothesis:

Hypothesis 3.1. Direct links between belligerents will not influence the likelihood of mediation onset.

3.2.2. The Effect of Indirect Links on Mediation Onset Although I argued in the previous section that direct links will not influence mediation onset, some kind of communication between disputants is a necessary condition for coordinating mediation (Kleiboer 1996; see also Slikker and van den Nouweland 2001: 21ff). Ties connecting the warring parties via one or more intermediaries, or more specifically indirect links between belligerents, satisfy this requirement. But unlike direct ties, indirect links address both demand and supply-side incentives by involving the entire set of actors and nodes in a network, i.e. the international community (Mitchell, Kadera and Crescenzi 2005; Greig and Diehl 2006: 368; Ward 2006: 151). Information about interests, intentions and understandings are disseminated here via prospective mediators and not dependent on the disputing parties alone. In the following, I will demonstrate that indirect ties are a substitute for direct links within the context of mediation and indeed may replace them completely.

First, indirect links promote mediation onset for the simple reason that they expand the set of possible mediators in a network. Strong and dense links to a third party may increase the chances that the belligerents ask for mediation, but the prospective mediating country must also be willing to supply this good (Beardsley 2006; 2010). That is not always given, especially if conflict due to its characteristics may be of no vital interest for a third party. A dyad with multiple links to outside parties, on the other hand, has a higher propensity to attract mediation because it has a larger pool of potential mediators with sufficient interest in the disputing parties (see Princen 1992). Sri Lanka shares multiple links with the European Union (EU) countries, but the EU placed the Tamil Tigers on the list of terrorist organisations in 2006. Given the EU’s policy not to negotiate with terrorists, a European intervention became difficult. If the EU states were Sri Lanka’s only node to be indirectly linked to the Tamil Tigers, any

40

offers of mediation would have been unlikely. However, mediation through Norway was possible as this country had ties to both Sri Lanka and the rebels. In other words, the more indirect links connecting a dyad, the more possibilities of prospective mediators on the “market of mediation demand and supply” through all possible and existing chains.

More importantly, indirect links connect parties to other states outside the dyad and determine the density, i.e. the concentration or cohesiveness of a social network as measured by the number and strength of links (Wasserman and Faust 1997: 314f). An international system may have many and strong edges between its nodes or not. In the former case, the density of the network is high and there are multiple, well established channels of communication. Hence, a fighting party is better able to communicate indirectly with its dyadic counterpart and there are also more possibilities for transmitting information about preferences and intentions from a conflicting dyad to prospective mediators (Dorussen and Ward 2008a: 192f; see also Mitchell, Kadera and Crescenzi 2005: 8). Outside parties do therefore not only have knowledge about belligerents’ interests and intentions, but also a common understanding, since ties to both states in a dyad exist (Dorussen and Ward 2008a: 193). Due to that, third parties are more likely to have incentives to engage in mediation efforts. As stated, although Israel and Egypt did not have many direct relations with each other in the 1970s, both countries had close ties to the U.S. These connections allowed the U.S. to have a precise understanding of both Israeli and Egyptian points of view. Ultimately, the U.S. capitalised on this information, intervened and effectively negotiated the Camp David Accords as well as the Israel-Egypt Peace Treaty.

Third, indirect ties do not only facilitate transmitting information from the dyad to potential mediators. Countries with various ties to outside parties are extensively involved in relationships with the latter. This involvement makes states more embedded or central in the international system (Wasserman and Faust 1997: 173; Ward 2006: 151f). In Figure 3.1., the U.S. is clearly a central actor in this network, since there are strong links for almost every dyadic combination of the U.S. and another state. At the opposite extreme, the Solomon Islands or North Korea can hardly be seen as central. The level of centrality of a disputing dyad is essential for selecting the right country to intervene, since the centrality of belligerents determines the amount of information they receive about possible mediators. For mediation to occur, the disputants have to agree on an intervener (Beardsley 2006; 2010). They are more likely to reach an agreement on this, the more information they have about outside parties. Disputants that are more central will receive more information about prospective mediators in terms of their preferences, interests, truthfulness and credibility (Crescenzi et al. 2008: 13; Dorussen and Ward 2008a: 195). And this infor-

41

mation can be used when there is need for a mediator who is acceptable to both sides (Austen-Smith 1992; Wall, Stark and Standifer 2001; Boehmer, Gartzke and Nordstrom 2004: 14f; Kydd 2006).7 After the 2006 war in Lebanon, for instance, multiple prisoners of both sides remained in Israeli or Hezbollah custody. The warring parties were reluctant to accept countries such as the U.S. or the UK as a mediator. Especially the Hezbollah, although being the least central actor in this network, had enough information about these outside actors and believed them to solely mirror Israeli positions. At the end, however, the belligerents used their information from the fairly strong ties to Germany, which was perceived as an “honest broker” due to its general non-involvement in the region. Under these circumstances, Germany was accepted as a mediator and successfully brokered the last prisoner exchange deal between the belligerents in July 2008.

Finally, indirect links are also beneficial from the perspective of outside actors willing to intervene. Touval and Zartman (1985: 40) emphasize that “third parties are only accepted as mediators if they are likely to produce an agreement or help the parties out of a predicament and for this they usually need leverage.” In a similar manner, Ward (2006: 153) argues that states must have the ability to threaten or to bribe other countries, i.e. using leverage in order to ensure cooperation. Stronger indirect ties are likely to increase the incentives of actors in the international environment to assist in the conflict management process, as these ties make outside states more aware of potential negative externalities. As such interest rises, third parties will more strongly encourage mediation by establishing leverage over the combatants through positive and negative inducements such as financial, political or military support and coercion or force (Crumm 1995; Beardsley 2010; see also Princen 1992; Bercovitch and Houston 2000; Ward 2006: 153).

In sum, I argue that indirect links replace direct ties when we consider whether disputes see mediation or not. More and stronger indirect ties between a dyad increase the number of potential mediators, allow a more effective transmission and reception of information and also enhance the supply-side pressures in terms of leverage. This ultimately promotes the likelihood of mediation onset. I therefore claim:

Hypothesis 3.2. More and stronger indirect links between belligerents increase the likelihood of mediation onset.

7 This mechanism is related to the socialization effect arising from a more democratic international community, eventually leading to an increased perceived appropriateness of mediation as a way of settling conflicts peacefully (Mitchell 2002; Crescenzi et al. 2008).

42

3.3. Research Design 3.3.1. Data The data on conflict and mediation attempts are taken from the previously introduced ICM data set (Bercovitch 1999), where the sample for my analysis is the population of disputes in 1946 to 1995. The information on states’ positions in networks is taken from Dorussen and Ward (2008a). This is a new data base with information on direct and indirect ties between states via IGOs. I focus on IGO networks since I assume that IGOs generally lack the willingness and opportunity to play active, independent roles, making the specific purpose and institutionalisation of organisations less important. This does not imply, how-ever, that IGOs do not matter when assessing the effect of the network variables on mediation. IGOs serve as vehicles that establish links between its member states through the encounter of representatives in meetings and the working practices of these institutions. Hence, IGOs matter primarily through their network effects, but since the power of most IGOs is derived from their member states (Boehmer, Gartzke and Nordstrom 2004; Dorussen and Ward 2008a: 190), we can thus in practice ignore IGO influence as an intervening variable (see King, Keohane and Verba 1994: 173).

The data are essentially cross-sectional. However, some temporal dependence may persist as mediation offers are unlikely to be independent from previous mediation efforts throughout a conflict. More generally, there could be important uncontrolled factors specific to each conflict. To address this problem, I cluster the standard errors on disputes.8

8 However, this approach only corrects the standard errors of the estimated coefficients. Hence, I also consider two alternative model specifications. First, I estimated all models with a mediation–years variable and different sets of cubic splines (Beck, Katz and Tucker 1998). Note, however, that the data have missing values on the dependent item, which makes the former variables hard to interpret. The results do not change significantly when leaving out the mediation–years variable and the cubic splines for temporal correction and I thus report the findings without them. Second, a dyad's persistence in mediation acceptance/rejection may result from potential fixed unobserved heterogeneity. Chamberlain (1980) suggests using a conditional logit that focuses on changes of the dependent variable. This model is consistent and efficient if heterogeneity does exist, but the fixed effects conditional estimator drops all units that see no change on the dependent variable. The structure of the ICM data implies that this procedure would discard all conflicts that only have one observation, leading to a drop–out in N of about 300. As such, the cure is likely to be worse than the disease here, given the severe loss of information.

43

3.3.2. Operationalisation The dependent variable is a dichotomous measure whether a conflict sees mediation or not (1 = mediation occurred; 0 = otherwise). Since the dependent variable is binary, I use logistic regression to test the hypotheses.9

The independent variables of interest examine the direct and indirect ties between belligerents. Direct links measures the number of joint IGO memberships, e.g. 54 in the case of the U.S. and the UK in 1982. Consequently, this variable takes lower values for dyads with fewer shared IGOs. This direct-links measure is the same as in Greig and Diehl (2006) or Crescenzi et al. (2008).

In order to analyse the indirect links between disputing parties through various chains in a network, I employ Indirect Links and the Maximum Flow of Information from the Dorussen and Ward (2008a) data. With regard to the former, when belligerents ask for mediation it is plausible that they try to involve states that connect them to their dyadic counterpart via the shortest and most efficient ties, since these ties transmit information with lower costs, more precisely and faster (Dorussen and Ward 2008a: 197). In other words, belligerents look for a connection via only one intermediary that is as strong as possible. For example, Argentina asks the U.S. to mediate in the Falklands War. Indirect Links measures all short indirect connections of a dyad, while acknowledging denser links to third parties by substitution for weaker ties (Dorussen and Ward 2008a: 197). The variable is calculated at five-year intervals. Formally, the measure is defined as:

Indirect Links

jikjkvkivji

,)],(),(1ln[),( (3.1)

where i and j are the countries in a disputing dyad, k a third party that links the belligerents and v the value, i.e. the strength of a tie linking two actors.10

The Maximum Flow of Information measures the degree of centrality or embeddedness of a country by counting flows of information through all possible chains of intermediaries, i.e. the existent number of direct and indirect links a 9 There may be multiple reasons for observing a zero on the dependent variable. A value of zero on Mediation Onset includes not only disputes without any conflict management attempt, but also non–mediation techniques such as bilateral negotiation, arbitration, referral to international organizations, etc. In order to control for this unobserved heterogeneity, I re–run the analysis using a preliminary version of a zero–inflated logit estimator (Beger, DeMeritt, Hwang and Moore 2009). The core findings revealed only minor differences, though. 10 Dorussen and Ward (2008a) add 1 to the summed product of the values of ties between countries in order to avoid the log of zero. The variable is logged due to decreasing returns with respect to the increasing likelihood of mediation onset arising from more indirect links.

44

country has to other states in the entire international system (Ford and Fulkerson 1956). Since this variable includes both direct and indirect connections, I expect Maximum Flow of Information to positively influence the onset of mediation, but its effect should be smaller than for Indirect Links. The variable’s data are collected at five-year intervals and approximate the lowest degree of centrality in the dyad, i.e. the value of Maximum Flow of Information is determined by the least central country in a dyad (Dorussen and Ward 2008a: 196; see also Dorussen and Ward 2008b: 7). More precisely, Maximum Flow of Information is calculated by:

Maximum Flow of Information (i,j)=min(degree(i),degree(j)) (3.2)

In turn we have:11

degree(i)=ji

jiv100

),( (3.3)

3.3.3. Control Variables Although I focus on how different ties between belligerents influence the likelihood of mediation onset, other factors such as conflict characteristics have to be considered as well since they might increase the costs of both conflict and mediation. I therefore examine the effect of several controls which have been included in previous studies as some of the determinants of mediation onset (Kleiboer 1996; Wall, Stark and Standifer 2001; Greig 2005). As emphasized, mediation occurs according to both demand and supply (Beardsley 2006; 2010). The belligerents as well as the prospective intervening parties must have incentives for mediation, either in terms of accepting a proposal for mediation or offering to intervene (Kleiboer 1996: 367; Regan 2002: 59; Terris and Maoz 2005: 566f; Beardsley 2006: 81; Greig and Diehl 2006: 357; Regan and Aydin 2006: 740). I include two other variables tapping actors’ incentives. The first variable measures dichotomously whether a third party offered to mediate. The second item measures on a 3-point scale the disputants’ demand for mediation (0=no disputing party asked for mediation; 1=one belligerent asked for mediation; 2=both disputants asked for mediation).

11 In order to calculate the maximum flow of information with this equation, the network must

satisfy the condition r (n–2) 21

, where n is the number of countries and r is the ratio of the largest edge value in a network to the smallest edge value (Dorussen and Ward 2008a: 196).

45

Second, Bercovitch, Anagnoson and Wille (1991: 13), Bercovitch and Gartner (2006b) and Quinn et al. (2006: 454) emphasize that high-intensity disputes are more likely to see mediation. However, high-intensity conflicts are also more complex, which leads to greater uncertainty. This makes it more difficult for the belligerents to agree on mediation (see Kressel and Pruitt 1989; Jackson 2000). Finally, due to the interdependence of actors and norm promotion in dense networks (see Mitchell 2002), conflicts may be lighter here. In this case, third parties will increasingly intervene, as these disputes pose lower risks and are more likely to be mediated successfully. Based on this discussion, I include the number of fatalities and its squared term to control for a nonlinear relationship between conflict intensity and mediation onset (see also Mason and Fett 1996: 551).12

Third, outside parties are less likely to intervene in long-standing conflicts that have “witnessed a long history of mutual attack and atrocity” (Touval and Zartman 1985; Bercovitch and Langley 1993; Kleiboer 1996: 362f; Pearson 2001: 276). Hence, conflicts are less likely to see mediation the longer they last. I consider a variable that measures the number of months elapsed until mediation occurred.

Finally, democracies have a common structural background and institutional procedures, which increase the chances that fighting parties will agree on mediation (e.g. Dixon 1994; Raymond 1994; Bercovitch and Houston 2000; Mitchell 2002). The data are taken from Polity IV (Marshall and Jaggers 2002), ranging from -10 (full autocracy) to +10 (full democracy). I use a weakest link specification, where the least democratic country in a dyad determines the likelihood of mediation onset. 3.4. Empirical Findings I now empirically test how direct and indirect ties between belligerents influence the likelihood of mediation onset. Table 3.1. provides estimates for six models, displaying the marginal effects while holding everything else at the means (or modes for binary variables). Models 1-3 show the results for the respective network measures only. I consider these items separately in the regression estimates because of the high pairwise cross-correlations of the network variables.13 Models 4-6 incorporate the controls in addition.14

12 More specifically, I expect the coefficient of intensity to be positive and intensity squared to be negative reflecting the hypothesised non–linear relationship. 13 The pairwise cross–correlations range in the interval [0.62; 0.79].

46

The results in Table 3.1. provide support for the theory that it is rather indirect links between a dyad that determine mediation. The Wald statistics are significant for five out of six models, allowing me to reject the null-hypothesis that all marginal effects are jointly indistinguishable from zero. Model 1 is the only exception. This result is not surprising, however, and provides some initial support for my theory. When considered as the only explanatory item, the marginal effect of Direct Links is basically zero, while not being significant. Comparing the values of the Akaike Information Criterion (AIC) further supports my arguments. The Direct Links model performs far worse than the model estimates with Indirect Links or the Maximum Flow of Information. For Models 1-3, Model 2 has the best fit with an AIC value of 2,471.91. When comparing Models 4-6, it is apparent that the controls contribute to the ability to account for mediation, leading to an overall decrease of the AIC. In addition, we still find that the models with Indirect Links and the Maximum Flow of Information perform better than the estimates with Direct Links.

With regard to my hypotheses, Table 3.1. indicates that all network variables have the predicted sign and statistical significance. Direct Links has a weak positive value that approximates zero. As expected in the first hypothesis, the item is not significant at conventional levels in any model. It seems indeed that direct links are either cut off or lose their importance after states enter a conflict. Hence, direct links between fighting parties tell us little about the likelihood of mediation onset. This finding is consistent with Greig (2005), but not with the results of Crescenzi et al. (2008) who find that shared IGO memberships actually increase the likelihood of mediation. As implied by the chapter’s second hypothesis, more and stronger indirect ties make mediation onset more likely. Indirect Links and the Maximum Flow of Information have positive estimated marginal effects that are significant at the 5% level. A small positive change to Indirect Links increases the chances that a dispute will be mediated by about 9% on average, while the marginal impact of a change in Maximum Flow of Information increases the likelihood of mediation by about 1%. In other words, indirect links replace direct ones when trying to predict the likelihood of mediation. Note, however, that Indirect Links has a stronger impact than Maximum Flow of Information. This is not surprising given the operationali- 14 The control variables Incentive Disputants, Incentive Third–Party and Intensity could also be seen as intervening variables, since these items are usually not causally prior to the formation of the IGO network. However, as controlling for intervening covariates generally weakens the results, the inclusion of these determinants of mediation should not pose a problem for my substantive findings. Nevertheless, in unreported analyses I run the models without those three covariates and my findings remain robust. Further, the empirical models in the next chapters all include a control item for the military capabilities of the belligerents. That variable has been dropped in this analysis due to insignificance, but the results remain robust when included.

47

sation of the former variable. When demanding mediation, states look for the shortest and most efficient ties to third parties. On the other hand, Maximum Flow of Information as a centrality measure also reflects direct ties to some extent.

Table 3.1. The Onset of International Mediation, 1946-1995.

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Direct Links 0.00 0.00 (0.00) (0.00) Indirect Links 0.10 0.08 (0.04)** (0.04)** Maximum Flow of Information 0.01 0.01 (0.00)** (0.00)** Incentive Disputants -0.34 -0.35 -0.35 (0.05)*** (0.05)*** (0.06)*** Incentive Third-Party 0.63 0.66 0.61 (0.11)*** (0.12)*** (0.11)*** Intensity 0.77 0.51 0.67 (0.28)*** (0.25)** (0.29)** Intensity (sq) -0.16 -0.09 -0.13 (0.06)*** (0.05)* (0.06)** Duration -0.00 -0.00 -0.00 (0.00) (0.00)** (0.00)** Democracy 0.02 0.02 0.02 (0.01)* (0.01)* (0.01)* Constant -0.04 -1.13 -0.23 -0.39 -1.00 -0.45 (0.15) (0.49)** (0.14)* (0.26) (0.43)** (0.24)* N 1960 1831 1923 1725 1598 1698 Log Pseudo Likelihood -1347.68 -1233.96 -1278.98 -639.13 -606.11 -606.15 Wald (df) 0.49(1) 5.54(1) 4.77(1) 164.47(7) 153.59(7) 150.21(7) Prob> 0.49 0.02 0.03 0.00 0.00 0.00

AIC 2699.37 2471.91 2561.95 1294.27 1228.22 1228.31

Note: table entries are marginal effects while holding all other variables at their means (or modes for binary variables); clustered robust standard errors in parentheses; * significant at 10%; ** significant at 5%; *** significant at 1% (two-tailed).

48

The results of the control covariates are mainly consistent with the existing literature. Duration has a negative sign that is significant in almost all models. The marginal effects of this item suggest a rather small impact, though. Also, the more democratic the least democratic state in a dyad, the higher the chances for mediation onset. Furthermore, third parties are more likely to intervene when they have incentives to do so. The marginal effect of Incentive Third Party approaches 64% on average. Moreover, I find a nonlinear relationship between dispute intensity and mediation onset, where interventions are the least likely in conflicts that are either more easy or difficult to solve. Finally, the negative sign of Incentive Disputants is more difficult to interpret. My results suggest that incentives for belligerents may be insufficient for mediation in the absence of the right incentives for mediators. The chances to see mediation in a conflict decrease by about 34% if the value of this variable is one unit lower.

To further assess the implied magnitude of my findings, Table 3.2. reports the predicted probabilities for the lowest and highest value of each item while holding all other variables at their means (or modes for dichotomous variables).15 The predicted probability for mediation incidence increases by about 16%-points when we move from the lowest to the highest value of Direct Links. However, this is only a small increase and this substantive effect should be interpreted cautiously due to the insignificance of Direct Links in Table 3.1. By contrast, when we move to the highest value of Indirect Links and Maximum Flow of Information, we see an increase in the predicted probability of mediation onset by about 60%-points and 50%-points, respectively. This is an impressive increase for both variables given their relatively small baseline probabilities of about 2% (Indirect Links) and 28% (Maximum Flow of Information). Thus, more indirect ties make a substantial difference to our expectations about whether mediation is more likely or not.

15 The predicted probabilities of the control variables mirror the findings in Table 3.1. Due space limitations, I will not discuss these further. Also note that the distribution of Indirect Links is highly skewed. About 90% of the distribution lies above 10.13. Since the range of possible variation may not be entirely comparable, I re–run the analyses without all observations that have a value smaller than 10.13 on Indirect Links. The findings do not change substantially.

49

Table 3.2. Predicted Probabilities of Mediation Onset.

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Min Max Min Max Min Max Min Max Min Max Min Max

Direct Links

0.46 0.64

0.44 0.59

Indirect Links

0.02 0.64

0.03 0.58 Maximum Flow of Information

0.28 0.79

0.28 0.74

Incentive Disputants

0.75 0.12

0.72 0.12

0.76 0.12

Incentive Third-Party

0.50 0.95

0.45 0.93

0.51 0.94

Intensity

0.03 1.00

0.08 0.98

0.04 1.00

Intensity (sq)

0.96 0.00

0.84 0.01

0.94 0.00

Duration

0.58 0.07

0.55 0.04

0.61 0.04

Democracy

0.41 0.75

0.35 0.74

0.40 0.77

Note: predicted probabilities are shown for the minimum and maximum values of each specific variable; calculations were done while holding all other variables at their means (or modes for binary variables).

Figure 3.3. depicts the predicted probabilities for the network measures in

their full model specifications, i.e. Models 4-6. Again, consistent with H3.1. and H3.2., it matters significantly if belligerents interact via direct or indirect linkages. All three graphs show an increasing likelihood of mediation with ties. However, the slope of Direct Links starts at a very high baseline probability and is only marginally positive, demonstrating that direct links do not exert much help in predicting mediation occurrence. Maximum Flow of Information has a comparably larger, positive slope and replaces Direct Links in terms of accurate predictions for mediation onset around its mean value. When comparing the maxima of Direct Links and Maximum Flow of Information, the probability to see mediation is almost 15%-points higher for the latter variable. As indicated by the marginal effects, the highest increase is found for short and efficient links via third parties. Starting at a rather low probability for mediation onset (3%), the probability for a dispute to see mediation is ceteris paribus more than 58% at the maximum value of Indirect Links. This means an increase of more than 1,930%. To illustrate these figures with an example, consider the Lake Tiberias Dispute between Syria and Israel, which started in 1951 and has seen five mediation

50

onsets since the start of the conflict. The direct links between the two belligerents are unlikely to be good predictors here, since Israel and Syria only shared 14 joint IGO memberships in 1951. At the same time, the disputants were indirectly linked, with a fairly high value of 9.6 on the Indirect Links scale.

Note: dashed lines are 95% confidence interval; predicted probabilities are shown for standardized values of the network variables (i.e. min=0 to max=10). All other variables held at their means (or modes for binary variables).

Figure 3.3. Predicted Probability of Mediation Onset by Network Measures. 3.5. Conclusion This chapter has sought to advance our understanding of when mediation is likely to occur. In examining the interconnectedness or interaction between belligerents from a network perspective, I have developed a theoretical rationale

51

arguing that indirect ties between parties can provide a better understanding of the prospects for mediation in disputes.

My findings suggest that mediation and conflict management researchers must consider the position of states in networks and the role of indirect ties in disputes more carefully. Whereas previous research has anticipated the importance of ties, it has overemphasized the role of direct ties at the expense of indirect ones. The empirical results strongly support my claim that indirect rather than direct links between fighting states are important influences on whether a conflict sees mediation or not. More and stronger indirect ties between a dyad increase the number of potential mediators, allow a more effective transmission and reception of information and also enhance the supply-side pressures in terms of leverage. This increases the likelihood that a conflict ultimately will see mediation. Put differently, when predicting mediation, direct links are likely to become severed once conflict breaks out and are replaced by indirect ties.

The results also have strong implications for policy makers. For example, private institutions or individuals, i.e. “Track Two Diplomacy” (Diamond and McDonald 1996), increasingly intervene in conflicts by setting up workshops, reconciliation programs and so forth in order to improve the direct relationship of belligerents. The ultimate goal of these interveners is to increase the prospects for successful mediation. Given their focus on direct links, my research raises doubts about whether such efforts are likely to generate more third-party conflict management attempts. Also, belligerents who seek a peaceful settlement need to examine their ties to prospective mediators cautiously in order to increase the chances that a request for intervention will not be rejected. Third parties that are closely tied to both disputants will be more likely to accept a demand for mediation. Consequently, belligerents will save valuable time and resources that are necessary for effective dispute settlement if they avoid asking outside parties that – despite high leverage, strong resources or much power – have weak ties to the antagonists and therefore will not be willing to intervene in the first place.

Although my empirical analysis demonstrates a great deal of empirical support for the theory, many important questions remain. Further research along the following lines may help to shed further light on when conflicts see mediation. First, although I clarified the impact of direct and indirect linkages on whether a conflict will see mediation, I did not examine their effects on the effectiveness of mediation efforts. Second, this first step combining social network analysis and mediation focused on when mediation is more likely. Although we would need a different research design and data, network analysis may also help provide insights into what third parties are best positioned to mediate or most likely to succeed, for example through incorporating social network measures into the data of Bercovitch and Schneider (2000). Third, it

52

would be worthwhile to go beyond the state-centred focus here and study the effects of networks incorporating non-state actors. Furthermore, although I have only examined the interconnectedness between belligerents, actor constellations in conflicts may be much more complex and involve additional direct and indirect linkages of both mediator(s) and belligerents. Finally, I limited myself to networks through IGO ties here, but it is plausible that other kinds of ties such as cultural, trade or alliance based links may have different effects on mediation. I will address the last issue in the following chapter consequently.

53

4. The Impact of Trade on International Mediation Trade features prominently in the study of actors affecting the risk of interstate conflict. Several studies have argued that the higher the amount of trade between two countries, the more dependent they are on each other, which then leads to a significantly lower propensity of conflict (see e.g. Barbieri 1996; 1998; Dorussen 2006; Dorussen and Ward 2008a; 2010; Polachek 1980; Polachek, Robst and Chang 1999; Russett and Oneal 2001).16 In light of this, it is rather surprising that trade has generally been neglected when studying the process of international mediation, since the disputing parties and/or an outside actor(s) must have interests at stake when considering resort to mediation (see Beardsley 2006; 2010; Crescenzi et al. 2008). Trade, either between the belligerents themselves or between third parties and the disputants, by definition entails costs and benefits.17 This in turn is likely to influence the utility function of belliger-ents and prospective mediators, and, hence, not only making it plausible that trade shapes the prospects for mediation, but also emphasizing that the latter is a largely neglected facet in the broader relationship of trade and conflict (Polachek 1980; see also Polachek, Robst and Chang 1999). For example, early border tensions between Ecuador and Peru evolved into full-scale offensives in 1995, at a time when none of the countries were well embedded into the world’s trade system. Neither Ecuador nor Peru had prospects for winning the conflict militarily and the ongoing dispute was costly to both states in the sense that the disrupted trade between the two could not easily find substitutes through alternative markets. In other words, both states were increasingly losing re-sources through continued conflict, and were as such more willing to see the dispute mediated.

In one of the first steps towards examining this relationship, this chapter studies trade ties both between belligerents and between belligerents and states outside this dyad in order to explain when conflicts see mediation. Starting with

16 Previous research emphasises, however, that the trade–conflict relationship may be reciprocal, i.e. that conflict also influences trade patterns. I will address this point more closely in the next section as well as when discussing the estimator. 17 Throughout this chapter, the term “third party” does not necessarily refer to a mediator, but rather an actor outside the conflict that the belligerents trade with. Hence, “third party” and “outside state/party” are used interchangeably.

T. Böhmelt, International Mediation Interaction, DOI 10.1007/978-3-531-92812-8_4,© VS Verlag für Sozialwissenschaften | Springer Fachmedien Wiesbaden GmbH 2011

54

an opportunity costs model (Polachek 1980; Polachek, Robst and Chang 1999), the theory centres on substitution and the impact of sanctions. I argue that dense bilateral trade links between a fighting dyad are characterised by high opportunity costs, since it would be more expensive to seek alternative markets and creating trade links to other parties than simply maintaining the benefits of the already existing ties. In addition, dense bilateral trade with the dyadic counterpart would increase a country’s vulnerability to sanctions from its opponent in the case of conflict. Hence, fighting parties have an incentive to limit their hostile behaviour and ultimately to seek mediation. On the other hand – and contrary to previous research – I argue that belligerents’ trade ties to outside parties decrease the likelihood of mediation on both the supply and demand side. Combatants with multiple other alternatives for trade are not only able to mitigate the consequences of sanctions either from their opponent or some third parties, but also have the possibility to substitute for markets, which may be at risk or even lost, since economic agents in either belligerent country suffer a current loss of trade or expect a decline in the future profitability due to the outbreak of a conflict (Li and Sacko 2002: 13f).18 Similarly, outside parties that supply similar markets may have an incentive to see the disputants continue fighting in order to reap the benefits from fewer competitors. Finally, the nature of these different trade ties implies that the opportunity costs of bilateral trade lower the opportunity costs of trade links to states outside a belligerent dyad and vice versa. Taking this interaction into account, I demonstrate that the impact of bilateral trade varies conditional on trade ties to outside parties.

This chapter has crucial theoretical and empirical implications for examining when we are likely to see mediation. While little work has been done on the determinants of mediation in general, the trade-mediation nexus is particularly theoretically underdeveloped. As I will show in the literature review on me-diation and trade, many theoretical arguments are complementary and consistent with each other, but often too limited and see little empirical support. Furthermore, better understanding of how trade links influence the prospects for mediation onset can help policymakers to use such tools of conflict resolution successfully through sanctions. I will return to these topics in the conclusion.

18 I will elaborate this issue in more detail in the next section. Pollins (1989a; 1989b) and Mansfield (1994) argue more generally that more conflictual relations between states will tend to reduce the prospects for trade. Barbieri and Levy (1999) as well as Levy and Barbieri (2004) argue against this and present evidence suggesting that conflict does not necessarily reduce the amount of trade be–tween belligerents substantially. Despite compelling theoretical arguments, however, Anderton and Carter (2001) hold that these studies only look at outliers, and in addition they largely lack statistical significance and/or are hardly generalisable (see e.g. Barbieri and Levy 1999: 470ff).

55

The remainder of the chapter proceeds as follows. Before focusing on the actual trade-mediation relationship, I briefly review the relevant research on the broader trade-conflict nexus and clarify how my study is related to this. I then detail the previous literature on trade and international mediation. Based on an opportunity costs model, I afterwards develop a theory that helps understanding how different sorts of trade links influence mediation onset. The subsequent sections outline the research design, data analysis and robustness checks. I conclude with a discussion of my findings and the broader implications. 4.1. The Relationship of Trade and Conflict – An Overview A realist conception of trade argues – in a simplified way – that trade increases the wealth of trading countries, which in turn enables these states to transform gains from commercial exchanges into military capabilities (e.g. Gowa 1989; Gowa and Mansfield 1993). States would therefore seek to manipulate trade patterns to promote commerce with friends and restrict trade with enemies. Such a strict “state-centred” perspective on trade has become less prominent, since “trade flows are affected by the decisions of social actors at every level” (Pollins 1989a: 738), and many other actors other than states, e.g. individuals, interest groups or companies are influential economic agents in trade (Pollins 1989a; 1989b; see also Keshk, Pollins and Reuveney 2004: 1164; Li and Sacko 2002).

The perspective on state-centred conceptions of political biases in trade patterns can be extended to non-state actors if economic agents at other levels also follow similar influences in making decisions on importing or exporting goods. More precisely, Pollins (1989a: 738f; 1989b: 470) argues that individuals, interest groups, companies, country elites and so forth take into account both political and economic concerns, i.e. the place of origin of these products and the political relationship between nations will matter in addition to the price and quality of goods and services. In other words, these distinct actors operate at different levels, but their trade utility functions can be assumed to be similar and their interactions with each other ultimately converge to the same outcome. Therefore, despite the fact that my theory will examine different units of analysis, the level of data aggregation used for the succeeding analysis lowers the importance to precisely distinguish between a perspective that focuses on economic agents like individuals or interest groups and the state-centred perspective (see Keshk, Pollins and Reuveny 2004: 1164). For example then, actors’ activities in the domestic political arena that may influence the behaviour of a national government towards a foreign state, and, hence, the trade volume

56

between those countries (Pollins 1989a: 740) will be treated empirically as if this change has been caused by the government directly.

Within this context it is worth noting that the causal relationship between trade and conflict is unlikely to be as straightforward as presented in the introduction. Previous studies demonstrate in a compelling way that the causal mechanism of trade and conflict may be reciprocal, in the sense that trade may influence the onset of conflict, while conflicting or cooperative relationships between actors also determine the level of prior trade between them (Reuveny and Kang 1996). For example, while Polachek (1980), Polachek, Robst and Chang (1999), Russett and Oneal (2001) or Dorussen and Ward (2010) show that larger trade flows between states decrease the likelihood of conflict, Pollins (1989a; 1989b) as well as Li and Sacko (2002) measure conflicting ties between actors in order to predict trade levels. Still, despite the wide acknowledgement that the trade-conflict relationship is indeed reciprocal, most scholars employed single-equation models for testing this, although this is likely to lead to biased estimates. Keshk, Pollins and Reuveny (2004) and Kim and Rousseau (2005) are noteworthy recent exceptions that systematically address this simultaneity bias by examining the trade-conflict relationship via a simultaneous equation estimator. Their results suggest that there is indeed a significant negative impact of dyadic conflict on trade, but that trade does not seem to have a discernable impact on the risk of conflict. 4.2. International Mediation and Trade – Literature Review As stated above, despite the relatively extensive work on what factors may influence mediation effectiveness (Kleiboer 1996; Wall, Stark, and Standifer 2001), we know little about when we are likely to see mediation attempts in the first place (see Bercovitch and Schneider 2000; Greig 2005). One of the rarely addressed determinants of mediation onset is trade. To my knowledge, only five previous studies have examined this relationship directly.

First, Kathman (2007) shows that regional economic ties to outside parties are likely to increase the chances of mediation in civil wars. Third parties have a vital interest in securing their economic benefits with the incumbent government through peaceful settlement and are therefore more willing to intervene as a mediator.

Second, Greig and Diehl (2006: 367) argue that positive contacts will soften the attitudes of belligerents and eventually allow each side to see “peaceful, rather than hostile solutions to their differences.” Among other variables, the authors use the total amount of trade as a percentage of GDP between fighting

57

parties for testing this claim. The results demonstrate that this item has a negative, but insignificant impact on mediation onset. Greig and Diehl (2006: 377) lack a coherent theoretical explanation for this, however, and conclude that trade is unlikely to influence conflict management at all.

Third, Crescenzi et al. (2008) are interested in the credibility of a prospective mediator. The authors claim that if outside parties share trade relationships with a disputant, then the former will be biased towards this belligerent. The combatants interpret this as a “truth-telling constraint,” which makes a third party more credible as a mediator. Crescenzi et al. (2008: 22) operationalise this bias via the trade ties of potential mediators to either the challenger or the target state in a conflict. The empirical findings are mixed, however. Although trade between a potential mediator and the challenger does not seem to affect the likelihood of mediation, the amount of trade between the target state and an outside party positively influences the prospects for mediation significantly.

In a similar manner, Regan and Aydin (2006) show that bilateral trade ties between a disputant and a third party increase the probability of outside interventions in intrastate wars. Conflict may affect the supply and demand of goods, which can make it less likely that outside parties benefit from trade with an antagonist. However, when the scholars disaggregate economic ties into inducements and sanctions, they find that neither is statistically significant. Regan and Aydin (2006: 751) suggest that economic ties are therefore less influential for civil war interventions.

Finally, Greig and Regan (2008) examine in a follow-up paper the impact of trade between one combatant and prospective mediators on the likelihood of mediation offers and whether these are accepted in civil wars. The authors argue that conflict may disrupt trade ties between fighting countries and third parties, but since the latter would like to preserve the benefits from trade, they have incentives to foster a peaceful settlement through mediation (Greig and Regan 2008: 763). Their results are largely inconclusive, lacking significance at conventional levels. One result appears to hold true throughout all model specifications, however: contrary to Greig and Regan’s (2008: 773) expectations, third parties that are more reliant on a disputing country are actually less likely to offer to mediate. The lack of support for their theoretical argument suggests that other mechanisms may be at work between trade and mediation. The authors present a set of potential explanations for this (Greig and Regan 2008: 774). First, prospective mediators may be reluctant to intervene out of concerns over further complicating matters. Second, other conflict management tools may be more appropriate when there are strong trade links to one of the belligerents. The authors propose economic sanctions or military interventions. Finally, dense trade ties of a third party to one of the disputants may increase the incentives for

58

the former to substitute existing trade patterns for new commercial partners. Additionally, states in civil wars may see mediation offers as a possibility of establishing new trade flows with a prospective mediator. Although all three explanations seem plausible, they are not tested as such and all these alternatives apart from the last one ignore the possibility that conflicts may not necessarily disrupt the trade with outside parties (see e.g. Li and Sacko 2002). On the contrary, combatants may be even more dependent on trade with any third party if they intend to win a conflict militarily, since trade channels with the opponent are likely to become disrupted or decrease significantly with disputes due to the following reasons (see Anderton and Carter 2001; Li and Sacko 2002: 13; Mansfield 1994; Pollins 1989a; 1989b). First, either government could perceive the trade exchange between the firms in their country as harmful to its military cause and may therefore cut them or impose sanctions. Second, if not fully cut off yet affected by sanctions, relying on these trade channels can become too costly for individuals or companies and force them to look for alternative markets. Third, these actors fear a loss in gains due to the higher level of risk that is caused by the fighting as such, the level of violence and so forth.

To recap, trade as a determinant of mediation has only recently received attention. The arguments why trade may increase the chances of mediation are largely consistent with each other, but focus on intrastate wars, and have generally not received much support in empirical tests. Two reasons appear as possible explanations for these shortcomings. First, the theoretical arguments have to date not considered that belligerents may want to increase their capability for continuing the dispute if there are ways to do so. Antagonists therefore do not have incentives for a peaceful settlement if trade links with the adversary can be substituted for trade to parties outside the dyad. Consequently, third parties may also have an interest in seeing belligerents to continue to fight. Second, and with regards to the empirical evidence, the literature did not capitalise on looking at trade flows along multiple and various chains of intermediaries and how these differ from and may modify the effects of bilateral trade links. The following sections try to address both of these issues. 4.3. A Theory on International Mediation and Trade Individuals, firms, interest groups and countries generally have a rational interest in trade. It enables them to enjoy things that would be unavailable under autarchy and makes a dyad at any level mutually dependent on each other, since trade is welfare enhancing for both (e.g. Barbieri 1996: 31; 1998; Dorussen and Ward 2010; Gowa 1989; Gowa and Mansfield 1993). Figure 4.1. shows the trade

59

relationships between six countries in 1958. A trade link equals the natural log of the averaged sum of imports and exports between a dyad in 1948 US dollar value (see Dorussen and Ward 2010). The width of the edges is proportional to the countries’ amount of trade: more numerous trade connections between two countries are represented by denser links.

In the most common form, trade ties are bilateral between pairs of states. Figure 4.1. demonstrates that the vast majority of states are connected through those trade links. For example, the U.S. and France trade comparably a lot with each other (edge value of 731.1). On the other hand, some countries are only barely linked through bilateral trade or do not share any direct commercial exchange at all. Morocco and the USSR, for instance, are linked through a rather weak bilateral trade relationship (edge value of 2.3), while the USSR is the only bilateral connection for North Korea, since the latter does not have any direct commerce with other states.

Figure 4.1. Subset of the 1958 Interstate Trade System.

60

In the case of conflict between two countries, their bilateral trade ties usually become torn apart (Gowa 1989; Gowa and Mansfield 1993; Mansfield and Bronson 1997; Li and Sacko 2002) and states have to rely on trade with third parties, i.e. markets other than their dyadic counterpart in order to obtain goods. Imagine that the Cold War would have cut off all trade relationships between the U.S. and the USSR. In this case and assuming further that Figure 4.1. provides all possible trading partners in 1958, the set of trade ties for the U.S. would look different. Figure 4.2. illustrates these third-party trade links for the U.S.-USSR dyad from a U.S. ego-perspective. As can be seen, the U.S. would have trade links with only three other countries (France, Morocco and Tunisia) and would have to compensate the lost trade with the USSR (edge value of 18.3). At this point, however, and similar to my argumentation in Chapter 3, there are no theoretical grounds for expecting that trade relationships with third parties should have the same effect as trade with the dyadic counterpart.

Figure 4.2. Constrained Subset of the 1958 Interstate Trade System.

61

4.3.1. An Opportunity Costs Model As elaborated above, in the case of conflict between two countries, their bilateral trade flows usually either become disrupted or decrease significantly, since it becomes too costly for firms to rely on them due to sanctions or the risk inherent in conflict (Anderton and Carter 2001; Li and Sacko 2002; Mansfield 1994; Pollins 1989a; 1989b; see also Mansfield and Bronson 1997). For example, bilateral trade between the UK and Argentina was rapidly halted after the outbreak of the Falklands-Malvinas conflict in 1982 (Pollins 1989a: 739). Similarly, despite large trade flows between the U.S. and Cuba in the 1950s, the bilateral trade between these states essentially collapsed after Castro came into power in 1959 (Kim and Rousseau 2005: 526). Under these circumstances of decreasing bilateral trade, states have to rely on trade with outside parties, i.e. markets other than their dyadic counterpart. Against this background, it still remains unclear, however, how individuals, firms or more generally countries value trade.

Economic agents value trade links differently according to the opportunity costs allocated to these ties (Dorussen and Ward 2008b; Polachek 1980; Polachek, Robst and Chang 1999). Trade links with high opportunity costs are valued more than links characterised by fewer opportunity costs. Trade has high opportunity costs if it is densely established, since this makes it more costly for economic agents to cut off these ties, to find alternative markets and to circumvent sanctions from the trading partner. In other words, the value of opportunity costs depends on the elasticity of substitution between suppliers and the impact of possible sanctions (see Dorussen and Ward 2008b, 5; Li and Sacko 2002; Rowe 2001).

Imagine a dyad consisting of actors i and j. When neglecting any domestically produced consumption, i’s welfare function Wi(C) is derived from the consumption of imported goods and the benefits due to exports (see Polachek, Robst and Chang 1999):

Wi(C) =n

ii

n

ii mx

11

(4.1)

where xi denotes the benefits due to exports from actor i and mi signifies imports to i. The imports and exports, i.e. i’s overall trade flow can be disaggregated into trade with the dyadic counterpart j and one or more than one third party, where xij denotes the benefits due to exports from i to j and mij stands for the imports to i from economic agent j.

62

Wi(C) = ijij

jn

ii

jn

ii mxmx

11

(4.2)

If a country’s needs have to be covered partially by imports and exports,

either from/to the dyadic counterpart or from/to third parties, it is reasonable to assume that the more trade flows are covered through bilateral trade, the less trade is necessary with actors outside the dyad and vice versa. Put differently, fostering bilateral trade decreases the opportunity costs of a country’s trade links to third parties, while raising the amount of trade with outside parties lowers the necessity to trade with the dyadic counterpart. Specifically, we can introduce factor ai that models this trade-off:

Wi(C) = iijij

jn

iii

jn

ii amxamx )()1)((

11 (4.3)

Actor i values bilateral trade less if the imports and exports with the dyadic counterpart become fewer in number, i.e. ai is defined by 0 ai 1 and that ai decreases either with an increase in jn

ii

jn

ii mx

11)( or with a lower value of

)( ijij mx . Thus, factor ai ensures that as soon as bilateral trade becomes worth less, the opportunity costs of trade ties with outside actors increase and thereby their value for i.

4.3.2. The Impact of Bilateral Trade on Mediation Extending the logic from the previous section, well established trade links to the dyadic counterpart make it more expensive for both states in a dyad or other economic agents within these countries to cut off their trade relations in the first place. However, if a conflict breaks out between a dyad,19 which usually ends 19 Although this chapter addresses processes after the outbreak of a dispute, one may question why states engage in conflicts in the first place despite their trade flows. One possible answer could be due to belligerent’s pre–dispute uncertainty about their actual trade dependence or their focus on capabilities. The crucial point is again that “states fight and firms trade” (Li and Sacko 2002: 13; see also Keshk, Pollins and Reuveney 2004: 1164), meaning that we may be dealing with actors at different levels where different factors weigh more heavily for some economic agents, even if the actors’ utility functions are qualitatively similar. However, my theory is not intended to address this issue and I am reluctant to speculate on this here.

63

most of the commercial exchanges between the belligerents or makes them in-creasingly inefficient (see Anderton and Carter 2001; Li and Sacko 2002; Mansfield 1994; Pollins 1989a; 1989b; see also Mansfield and Bronson 1997), dense bilateral trade will also make it more costly for either warring party or private actors in these belligerent countries to obtain goods from some alternative markets and to create trade links to other actors that can substitute for the benefits of the already existing ties. This is likely to hold true especially if trade with countries outside the dyad does not or only weakly exist.

Second, each belligerent wants to increase its chances of winning a dispute. If combatants or other economic agents within these states shared trade ties with each other before a conflict, either belligerent has incentives to impose sanctions on the opponent in order to disrupt the target’s economy for impairing its ability to pursue its objectives (Li and Sacko 2002: 14; Rowe 2001: 4f). This implies that the higher the amount of bilateral commercial exchanges between two disputants, the more detrimental the impact of sanctions on the target country’s economy, as the latter would be highly vulnerable to sanctions (Rowe 2001).

Consequently, I argue that the costs of an interruption of trade with the dyadic counterpart should encourage states to limit their hostile behaviour. Due to the assumption that trade relationships with an opponent either become lost with disputes or highly inefficient due to sanctions and other risks, states with dense bilateral trade are less likely to obtain the necessary resources for continuing as well as winning the conflict. Furthermore, domestic firms, interest groups and trading individuals, motivated by their own economic interests, will now pressure the government for an end to a conflict. In addition, sanctions that are imposed on dense bilateral trade links would undermine the overall economic situation for either belligerent significantly. This increases the likelihood that disputants sharing strong bilateral trade links with each other will be more willing to seek a peaceful solution. Hence, in complementing the argument on the “liberal peace” (e.g. Dorussen and Ward 2010; Russett and Oneal 2001), warring parties that have an interest in resolving a dispute peacefully might treat mediation as an attractive possibility under the circumstances outlined above, since all economic agents at each level may expect greater benefits from a mediated outcome than a negotiated one (see Simmons 2002: 834). For example, mediators are primarily seen as legal authorities that provide incentives or security guarantees. Furthermore, mediators can help overcome commitment problems by decreasing the level of uncertainty between the warring parties, imposing order and raising the costs of fighting. These are attractive character-istics that bilateral negotiations cannot offer – despite the fact that mediation implies losing some degree of control over the bargaining process to the mediator (Hensel 2001). But even if the antagonists do not value these aspects of

64

a mediator, which are particularly likely to appear attractive under the aspect of trade that involves by definition huge costs and benefits, there are still reasons to choose mediation, as domestic actors may find it more attractive to make concessions to a third party than to a political adversary (see Simmons 2002: 834). For these reasons, I formulate the following hypothesis:

Hypothesis 4.1. The higher the bilateral trade volume between

belligerents, the higher the likelihood of mediation onset.

4.3.3. The Impact of Trade Links to Third Parties on Mediation Opportunity costs also apply to the commerce with third parties, albeit differently. As in the case of bilateral trade, trade links with one or more than one outside actor are characterised by high opportunity costs if they are strongly established. The impact of sanctions on those trade ties would be highly disruptive and there would be large costs for finding alternative markets if this kind of trade is torn apart or becomes inefficient.

First, however, a significant portion of trade links to third parties will most likely not be affected by a dispute between a dyad. On the contrary, if a conflict breaks out between two countries, both the belligerents want to perpetuate their trade ties with outside actors and at least some third parties will try to secure existing trade to either belligerent to increase their benefits under the new circumstances (see Dorussen 2006; Greig and Regan 2008; Li and Sacko 2002: 14; Mansfield and Bronson 1997). Belligerent i, for example, if equipped with multiple and dense trade links to third parties, would have more possibilities to obtain goods than actors less embedded in the trade system. From a slightly different perspective, since belligerent i may be able to substitute for markets that are lost or at risk to become lost due to the outbreak of conflict (Dorussen 2006), those third parties that are able and willing to provide these alternative markets increasingly have incentives to raise their trade shares with i, as the former will obtain benefits from this. Therefore, at least belligerent i not only loses the incentive to keep or re-establish bilateral trade, but focuses on the war effort instead, since this country has the possibility to obtain its necessary goods from somewhere else. Ultimately, this will decrease the willingness of disputants to settle the conflict by peaceful means due to the given prospectus of winning it militarily and there is no mutual incentive for mediation on both the supply and demand side (see Gasiorowski and Polachek 1982; Polachek and McDonald 1992). After the onset of a dispute, by contrast, weak trade ties to countries outside the dyad are not likely to provide belligerents with more resources,

65

which I argue makes the belligerents more likely to seek mediation. The example discussed above shows how Ecuador and Peru faced such a situation in 1995.

Second, although some third parties will provide opportunities for trade substitution with countries in disputes, others have few interests in doing so, since they might want to decrease their commercial exchanges instead or impose sanctions on the fighting parties. This could be due to an alliance with either belligerent, a potential bias towards one or both disputing parties, etc. However, disputants with multiple third-party trade links are not only more likely to substitute even those lost markets for trade possibilities with other partners, but may also be able to circumvent and manipulate the impact of sanctions that are either imposed by the dyadic counterpart or some outside states (Rowe 2001: 20ff). Iran, for example, as one of the major oil producers in the world has largely been able to mitigate the consequences of international and U.S. trade sanctions. Since the demand for oil from other countries that have not joined the sanctioning efforts remains stable, Iran is in a good position to withstand these sanctions. Put differently, many trade links to outside parties will decrease the chances of mediation from a sanctions perspective as well, since countries that have the possibility of using dense third-party trade are less likely to demand a peaceful intervention and the supply-side of mediation is likely to be lower as well. My next hypothesis argues accordingly:

Hypothesis 4.2. The higher the trade volume with third parties, the

lower the likelihood of mediation onset. 4.3.4. The Impact of the Interaction of Bilateral Trade and Trade Links to

Third Parties on Mediation

As indicated above, it is plausible that countries substitute bilateral trade for trade with third parties and vice versa. Conditional on the existence of bilateral trade between disputing parties, trade ties to outside states may lower the opport-unity costs of bilateral commerce, since there is the possibility that goods can be purchased elsewhere and it is more likely that sanctions can be circumvented. Israel and Lebanon did not have any bilateral trade relationship in the 1970s, for example, but both countries were able to obtain some of their necessary goods from the U.S. or Arab countries, respectively.

Hence, the effect of trade relations to third parties on mediation partly depends on the trade between the two belligerents in a dyad and vice versa. Imagine that a dyad of states i and j is connected through some bilateral trade relations. This implies that i’s dependency on trade with j is lower if the former

66

has many trade ties to actors outside the dyad. Depending on the nature of substitution between bilateral trade and trade links to third parties, this may either increase or decrease the likelihood of mediation onset. On one hand, it could be that i and j share many bilateral trade ties. My argument on the independent effect of bilateral trade predicts that this increases the chances of mediation. We should observe this outcome as well if there are only a few possibilities for this dyad to mitigate sanctions from the opponent or to obtain their goods from outside markets, i.e. if the trade with outside parties is weakly established. During the “Military Bases Conflict” between France and Tunisia in 1958, for example, both countries shared many and dense trade links with each other, but Tunisia in particular was not well embedded in the worldwide trade system, since Tunisia was granted independence only two years before the conflict. Hence, Tunisia was dependent on the trade with France and thereby more likely to seek a fast and peaceful settlement through mediation. Ultimately, Anglo-American mediators intervened to settle the dispute (Bercovitch and Jackson 1997).

On the other hand, it may be that i and j have strong bilateral trade links with each other, but at the same time either belligerent has a high trade volume with actors outside the dyad. Hence, due to the existence of the latter, there is at least the possibility to substitute the dyadic counterpart for another trading partner and both disputants are more likely to circumvent any sanctions from the foe – despite the fact that bilateral trade ties are characterised by high opportunity costs. In other words, conditional on the trade with third parties it may well be that bilateral trade affects mediation onset negatively. I therefore claim in this chapter’s last hypothesis:

Hypothesis 4.3. Bilateral trade conditions the effect of trade ties to outside states on mediation and vice versa. The more bilateral trade between belligerents, the smaller the impact of trade links to third parties. The higher the trade volume with outside parties, the less positive the influence of bilateral trade on mediation.

67

4.4. Research Design 4.4.1. Dependent Variable The dependent variable measures dichotomously if a conflict sees mediation (1) or not (0). Again, the data on conflict and mediation attempts are taken from the ICM data project (Bercovitch 1999). 4.4.2. Explanatory Variables Based on Gleditsch’s trade data (2002) and following Dorussen and Ward (2010), bilateral trade is operationalised by a measure of dyadic total trade flows, i.e. a bilateral trade link equals the sum of imports and exports between a pair of states in 1948 U.S. dollar value (see Dorussen and Ward 2010). Averaging the values, the variable is then modified by calculating the natural log (after adding 1 to the base), while zero stands for no bilateral trade relationship at all:

Dyadic Trade= jiv ,1ln (4.4) where i and j are the states of a belligerent dyad, and v the value of a trade link, i.e. the trade volume between i and j. Since Dyadic Trade is considered to increase the likelihood of mediation onset, I expect a positive impact for this item in the statistical models.

Second, trade with parties outside a belligerent dyad is operationalised by Trade Centrality, a network measure that considers a country’s whole range and any strength of all trade ties to the entire international system (Dorussen and Ward 2010). Trade Centrality therefore measures the degree of centrality or embeddedness of a state: the more central the position of a country in the trade system, the more possibilities for trade via direct and indirect linkages do exist (Ford and Fulkerson 1956). More precisely, Trade Centrality is calculated by:

Trade Centrality (i,j)=min(degree(i),degree(j)) (4.5)

with:

degree(i)=n

jikiv ),( (4.6)

68

where k signifies any country in the international trade system and v the value of a trade link, i.e. the trade volume to any other state. I employ a re-scaled variable that is divided by 100. Since Trade Centrality is a monadic measure, I approximate the lowest degree of that item, i.e. the value of Trade Centrality is determined by the least embedded country in a dyad (see Dorussen and Ward 2008b; 2010).

For estimating the impact of the substitution between bilateral trade and trade links to third parties, I multiply Dyadic Trade with Trade Centrality and include the new variable simultaneously in the models. In addition, I run estimations with the ratio of bilateral trade and trade links to third parties (see also Barbieri 1996: 36):

Trade Ratio=

CentralityTradeTradeDyadic (4.7)

4.4.3. Control Variables Since I focus on how trade influences the likelihood of mediation, I follow the rationale in the previous chapter and control for other factors that may influence the dependent variable and trade relations in order to avoid potential omitted variable bias. Further, including determinants of mediation that other studies have identified may help control for possible selection effects as well, since it is possible that only certain types of disputes actually see mediation. For the controls, I closely follow Chapter 3 by considering the four interrelated clusters of mediation determinants (see Greig 2005: 251), i.e. the likelihood of eventual mediation success, the role of the previous conflict management history, domestic and international threats faced by the warring parties and characteristics of the conflict dyad (see also Bercovitch and Langley 1993; Greig 2005; Kleiboer 1996; Wall, Stark and Standifer 2001).

First, mediation occurs according to both demand and supply (Beardsley 2006; 2010; Crescenzi et al. 2008). As stated, the belligerents as well as the prospective intervening parties must have incentives either in terms of accepting a proposal for mediation or offering to mediate (Greig and Diehl 2006: 357; Kleiboer 1996: 367; Regan 2002: 59; Regan and Aydin 2006: 740). Therefore, one variable measures dichotomously whether a third party offered to mediate. I also include an item that measures the disputants’ demand for mediation (0=no disputing party asked for mediation; 1=at least one disputant asked for mediation). These two variables not only operate at the conflict characteristics level, but also are likely to influence the chances of eventual mediation success.

69

Second, belligerents that approximate military parity, i.e. neither side has a clear military advantage are more likely to accept mediation (Bennett and Stam 1996, 242; Mason and Fett 1996: 550). Power discrepancies decrease the incentives for the most powerful country to agree to a peaceful settlement as it stands a better chance of winning the conflict militarily. Using data from the Correlates of War Project (Singer 1988), I include a variable that is the natural log of the ratio of the stronger country’s capability to that of the weaker country (Singer, Bremer and Stuckey 1972). Thus, higher values on this item should be associated with a lower propensity for mediation.

Third, I also control for the intensity of a dispute. Since sanctions or the disruption of bilateral trade are usually not employed in low-intensity conflicts, I control for selection effects here (Regan and Stam 2000: 241). In addition, Bercovitch, Anagnoson and Wille (1991: 13) and Quinn et al. (2006: 454) emphasize that high-intensity disputes are more likely to see mediation. However, the relationship between dispute intensity and mediation onset may be non-linear (Mason and Fett 1996: 551), if more intense conflicts are also more complex, which leads to greater uncertainty and may decrease the prospects for eventually obtaining stable outcomes (Beardsley et al. 2006; Kressel and Pruitt 1989). Based on this discussion, I include a fatalities variable and its squared term to allow for an inverted U-shaped relationship between conflict intensity and mediation.

Fourth, mediation becomes less likely in conflicts the longer they last (Bercovitch and Langley 1993; Kleiboer 1996: 362f; Touval and Zartman 1985). I consider a variable that measures the number of months elapsed until mediation occurred. This variable pertains to the cluster of ultimate mediation success and captures aspects of the previous conflict management history as well as threats faced by the warring parties. Finally, including a duration item also corrects for possible temporal dependencies and selection effects (see Regan and Stam 2000).

Finally, democracies have a common structural background and institutional procedures, which increase the chances that fighting parties will agree to mediation (e.g. Bercovitch and Houston 2000; Dixon 1994; Mitchell 2002; Raymond 1994). Also, the trade patterns of countries may vary according to regime type. Autocracies, for example, could have mercantilist tendencies and weight imports negatively. I therefore use a variable from Polity IV (Marshall and Jaggers 2002), ranging from -10 (full autocracy) to +10 (full democracy). In order to include this individual state characteristic in my dyadic research design, I rely on a weakest link specification, i.e. the least democratic belligerent in a dyad determines the likelihood of mediation onset, while I expect this item to be positively related to mediation.

70

Table 4.1. provides the descriptive statistics of the variables.

Table 4.1. Descriptive Statistics of Variables.

N Mean Std. Dev. Min Max Mediation Onset 3207 0.58 0.49 0 1

Dyadic Trade 1940 1.80 1.91 0 7.45 Trade Centrality 1940 0.25 0.16 0.02 0.75

Dyadic Trade*Trade Centrality

1940 0.54 0.78 0 5.24

Trade Ratio 1940 8.46 9.57 0 38.67 Third-Party Incentive 3207 0.25 0.43 0 1 Disputants Incentive 2847 0.48 0.50 0 1

Capability Ratio 2015 1.91 1.37 0.002 9.50 Intensity 3166 2.00 1.10 1 4

Intensity (sq) 3166 5.19 5.09 1 16 Duration 3207 92.75 116.10 0 900

Democracy 1992 -4.78 5.05 -10 9

4.4.4. Estimation Procedure The binary dependent variable suggests a logit estimator. However, mediation attempts may depend on previous mediation efforts in a conflict and disputants may have underlying and unobserved propensities to accept or reject mediation offers due to some dyad-specific heterogeneity. Either issue would lead to serial correlation, which, if not addressed, could overstate the effect of trade on mediation. One approach for dealing with this is the inclusion of fixed effects in the model (Chay and Hyslop 1998). We obtain:

i

i

z

z

med

e

ePob ~

~

1)1(Pr (4.8)

where ~

iz incorporates ß’xi, i.e. a coefficient and a vector of independent

variables as well as a dyad-specific heterogeneity term i, so that

~

iz = ß’xi + i . Based upon this, Chamberlain (1980) suggests a maximum-likelihood estimator,

71

where a likelihood function is maximised that is conditional on the sum of each dyad’s outcome:

i titititiiii ppPpPpPobL ],...,,[Pr 2211

(4.9)

Here, t represents time units that actually see mediation, i.e. the coefficient

estimates are based upon dyads that change their mediation behaviour over time. For example, we observe mediation for a specific dyad in time t1, whereas in t2 there is no mediation in that dyad’s conflict. Dyads that do not change their mediation pattern over time do not contribute to this likelihood function and are thereby dropped from the analysis. The Chamberlain (1980) conditional logit estimator does not have distributional assumptions about dyadic effects and is consistent as well as efficient if heterogeneity does exist – unlike the standard logit model. Hence, after running Hausman specification tests to see whether there is unobserved heterogeneity, I employ conditional logit models for testing my hypotheses.

Two other issues are important to acknowledge here with respect to the previous discussion on trade and conflict, and my research question. First, since mediation is an essential instrument for ending conflicts peacefully, a study seeking to shed light on the impact of trade on mediation is only but one facet of the larger trade-conflict nexus. Hence, we also have to carefully judge if the potential caveats outlined above also apply to both my theoretical discussion and the empirical analysis. Related to the first point, albeit differently, the second issue pertains to the direction of causation. Since I intend to explain an outcome related to conflict, one might also ask whether simultaneity could lead to biased findings with regards to how trade influences prospects for mediation. However, my focus here is to account for when conflicts see mediation, not the effective-ness of mediation. Although understanding the conditions that make mediation more likely conceivably can also help understand what makes mediation more effective (Bercovitch and Gartner 2006), the fact that mediation per se occurs does not say much about its eventual degree of success or failure. But it is only the outcome of a mediation attempt that – one way or the other – redistributes the leverage, resources or influence between belligerents, ultimately creating more peaceful or hostile relationships between the warring parties (see Beardsley 2008). This brings us back to the causal influence of conflicting/cooperative relationships between two states on trade (see Pollins 1989a; 1989b; Li and Sacko 2002). In sum, although I believe the simultaneity problem to be important for understanding the broader trade-mediation relationship, I argue

72

that it only applies to studies combining mediation effectiveness and trade, and that research that solely focuses on mediation onset can proceed by a single equation estimation.

4.5. Data Analysis 4.5.1. Empirical Findings I start with the pairwise correlations between the trade measures. Table 4.2. demonstrates for Dyadic Trade and Trade Centrality that the correlation tends to be fairly low (r=0.27). The correlations increase to r=0.85, however, when the multiplicative term and Trade Ratio are considered. The simultaneous inclusion of the former variable is necessary for modelling existing interaction effects, but I will include Trade Ratio separately in the estimations.

Table 4.2. Pairwise Correlations of Trade Variables.

Dyadic Trade

Trade Centrality

Dyadic Trade*Trade Centrality

Trade Ratio

Dyadic Trade 1.00 Trade Centrality 0.27 1.00

Dyadic Trade*Trade Centrality 0.85 0.59 1.00 Trade Ratio 0.74 -0.22 0.31 1.00

Note: All figures significant below 1%.

I provide estimates for six models examining how trade influences the likeli-

hood of mediation. In Table 4.3., Model 1 shows the results for Dyadic Trade and Trade Centrality only. Model 2 introduces the interaction term, while Model 3 replaces the previous variables by Trade Ratio. For Models 4-6, I examine the impact of the same set of variables as in Models 1-3, respectively, but here I include the control variables as well. Hausman tests indicate that the conditional logit estimator is both consistent and efficient. The null hypothesis stating the consistency of the conditional logit against the regular logit is rejected below 1%.20 The results display the odds ratios instead of coefficients for the covariates on the response. Although the signs and standard errors of coefficients can be

20 Model 1: 2(2)=61.31; Model 2: 2(3)=57.79; Model 3: 2(1)=51.07; Model 4: 2(9)=134.27; Model 5: 2(10)=140.00; Model 6: 2(8)=155.97.

73

interpreted directly, they cannot be considered as slopes or elasticities for non-linear models. Odds ratios are essentially the effects on the probability of a binary event over the probability of the event not occurring following a change in the independent variable. This comparison is the switch from zero to 1 for dichotomous explanatory variables or a one unit increase for continuous ones. Odds ratios range from zero to an unbounded maximum, where values smaller than 1 indicate a negative impact on the probability that a conflict sees mediation following an increase in an explanatory variable.

Before I will come to my core explanatory variables, I briefly discuss the results of the control items in Models 4-6. The controls largely have the predicted relative size and are consistent with the literature. Duration has a negative impact that is significant in all models, but the size of the odds suggests a rather small impact. The capabilities variable is also statistically significant in all models and has the expected negative effect. On average, a one unit increase in Capability Ratio leads to a decrease in the chances of mediation by about 4,344%. Furthermore, Democracy and Third-Party Incentive are consistent with my expectations. Table 4.3. reveals that an increase by one unit of Democracy increases the likelihood of mediation by the factor 1.12 on average. This effect is significant in all models. Also, third parties are more likely to intervene when they have incentives to do so. The chances of seeing mediation in a conflict are larger by about 4.71 times on average where a third party offered to mediate. In addition, I find a non-linear relationship between dispute intensity and mediation onset, but the overall effect is insignificant on average. Finally, the negative impact of Disputants Incentive is consistent with my earlier findings. Again, it seems that my results suggest that incentives for belligerents may be insufficient for mediation in the absence of the right incentives for potential mediators. Disputes where at least one of the belligerents asked for mediation have a chance of about 0.04 to see mediation as compared to conflicts where this was not given.

With regard to my hypotheses, I will start by focusing on Models 1 and 4, i.e. the estimates that do not consider the substitution of bilateral trade for trade links to third parties and vice versa. I obtain support for this chapter’s first hypothesis. As argued, more bilateral trade between belligerents significantly increases the chances of mediation onset. A one unit increase in Dyadic Trade raises a dispute’s odds of seeing mediation by a factor of 50% and 57% in Model 1 and 4, respectively. It therefore seems that belligerents are indeed more dependent on each other if they have well and densely established trade links. This makes them more inclined to seek a peaceful settlement of their conflict and mediation ultimately becomes more likely. I also obtain support for my second hypothesis here. Trade Centrality has odds ratios that are below 1, indicating a negative influence on mediation onset. More specifically, Trade Centrality decreases

74

Table 4.3. The Impact of Trade on International Mediation, 1948-1995.

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Dyadic Trade 1.50 1.56 1.57 3.31 (0.10)*** (0.17)*** (0.18)*** (0.73)*** Trade Centrality 0.01 0.01 0.03 0.94 (0.01)*** (0.01)*** (0.04)** (1.63) Dyadic Trade * Trade Centrality 0.85 0.08 (0.29) (0.04)*** Trade Ratio 1.09 1.15 (0.01)*** (0.03)*** Third-Party Incentive 4.84 4.64 4.65 (2.26)*** (2.21)*** (2.18)*** Disputants Incentive 0.04 0.04 0.04 (0.01)*** (0.01)*** (0.01)*** Capability Ratio 0.03 0.02 0.02 (0.02)*** (0.01)*** (0.01)*** Intensity 0.15 0.22 0.27 (0.18) (0.27) (0.32) Intensity (sq) 1.60 1.42 1.38 (0.45)* (0.41) (0.37) Duration 0.99 0.99 0.99 (0.00)*** (0.00)*** (0.00)*** Democracy 1.13 1.12 1.11 (0.05)*** (0.05)*** (0.05)** N 1639 1639 1639 1383 1383 1383 Log Likelihood -694.7 -694.6 -700.23 -253.47 -243.94 -244.06 LR 83.07 83.29 72.02 657.34 670.64 676.16 Prob> 0.00 0.00 0.00 0.00 0.00 0.00 AIC 1393.41 1395.19 1402.46 524.94 507.87 504.11

Note: table entries are odds ratios; standard errors in parentheses; * significant at 10%; ** significant at 5%; *** significant at 1% (two-tailed).

the likelihood of mediation onset by about 9,900% (Model 1) and 3,233% (Model 4). In other words, if a dyad has multiple other options for trade in the international system, then both belligerents not only lose the incentive to keep or

75

re-establish bilateral trade, but focus on the war effort instead. The possibility exists that goods for continuing the warfare can be purchased through alternative markets and sanctions either from the dyadic counterpart or some third parties can be circumvented. The dependency on trade with the dyadic counterpart is not given and ultimately strong embeddedness in the world’s trade system lowers the chances of mediation onset. Adding or suppressing controls from the models does not alter this result.

Figure 4.3. Descriptive Statistics of Mediation Onset, Dyadic Trade and Trade Centrality.

Concerning my third hypothesis, I begin by looking at simple descriptive

statistics of the variables of interest. Figure 4.3. shows a smoothed plot of the occurrences of mediation according to the values of Dyadic Trade and Trade

76

Centrality. Despite its limited analytical value, this figure clearly demonstrates that mediation significantly varies depending on the interaction of the bilateral trade volume and the trade with outside states. More precisely, although we see a lot of mediation attempts for high values of bilateral trade, the number of conflicts that are likely to see mediation drops notably if Trade Centrality is held at high values, too. This gives some initial support for the third hypothesis.

Furthermore, as expected by my theory, we find a positive impact of Trade Ratio, i.e. the rate of substitution of Dyadic Trade for Trade Centrality, on the onset of mediation. A one unit change in that variable increases the odds of seeing mediation by 9% and 15% in Model 3 and 6, respectively. In other words, the more bilateral trade ties between belligerents as compared to the amount of trade with third parties, the higher the likelihood that their dispute will be mediated.

To further uncover this relationship, I now focus on the constitutive terms, i.e. Dyadic Trade, Trade Centrality and the related interaction variable. Note, however, that we cannot directly interpret the size, signs and the z-statistics (see Ai and Norton 2003; Braumoeller 2004; Brambor, Clark and Golder 2006). To unveil the substantively meaningful results, I calculate the odds ratios for Trade Centrality according to Dyadic Trade and the odds ratios of Dyadic Trade along the values of Trade Centrality (see Braumoeller 2004: 815ff). Figures 4.4. and 4.5. depict my findings.21

Recall that the third hypothesis in this chapter argued that the effect of bilateral trade ties on the onset of mediation is conditional on trade links to third parties and vice versa. I find mixed support for this hypothesis. As Figure 4.4. shows, Trade Centrality has a negative effect on mediation onset that does not seem to be conditional on bilateral trade ties. The odds of Trade Centrality rapidly approach zero with higher values of the bilateral trade measure, but this only mirrors the findings from Models 1 and 4 in Table 4.3. Hence, the more central the least embedded country in a fighting dyad, the less likely is mediation onset, but this does apparently not depend on the existence of bilateral trade ties. Leaving aside very low values of Dyadic Trade, the relationship between Trade Centrality and the onset of mediation is negative and highly significant along all values of Dyadic Trade. We obtain an odds ratio of 0.26 (not significant) for the minimum of Dyadic Trade and an odds ratio of 3.97e-9 for the maximum value of that variable.

21 The figures depict the effects for full model specifications, i.e. Model 5 only. The results do not differ substantially when leaving out the controls, however.

77

Note: 95% confidence interval marked with dashed line. Dotted-dashed line signifies point of 1.

Figure 4.4. Odds Ratio Estimates of Trade Centrality.

My claim on the impact of bilateral trade conditional on trade links to third parties finds strong support, however. As Figure 4.5. demonstrates, the slope of Dyadic Trade is negative, emphasising that if the amount of trade to parties outside the fighting dyad increases, the belligerents have the possibility to substitute bilateral trade for trade to third parties, which in turn increasingly affects mediation onset negatively. The negative slope of the bilateral trade variable is significant until the mean value of Trade Centrality and then again for the last 10% percentile of the latter. Along the continuum of Trade Centrality, the odds of Dyadic Trade range between 2.91 (minimum of Trade Centrality) and 0.47 (maximum of Trade Centrality). Further, the predicted odds of mediation onset decrease by about 111.66% if Dyadic Trade is increased by one unit and Trade Centrality is held at its maximum. In other words, although we found a general positive impact of bilateral trade on mediation onset on average in Models 1 and 4, Figure 4.5. highlights that we have to consider the trade to third parties as well. Conditional on the latter kind, the impact of bilateral trade on mediation may be negative: the more trade ties to parties outside the fighting

78

dyad, the lower the opportunity costs of bilateral trade ties, since belligerents increasingly acknowledge the possibility of substituting trade markets. This decreases the chances of mediation onset even if there were many bilateral trade ties between disputants.

Note: 95% confidence interval marked with dashed line. Dotted-dashed line signifies point of 1.

Figure 4.5. Odds Ratio Estimates of Dyadic Trade. 4.5.2. Robustness Checks In order to ensure the robustness of my findings, I changed a variety of model specifications and re-run the estimates again. First, Clarke (2005) shows that the inclusion of control variables may actually increase the bias instead of decreasing it. In addition, some of the controls, especially the two dichotomous variables on actors’ incentives might undercut the significance and size of the trade variables that are supposed to model why belligerents demand mediation. Hence, Models 1-3 do not include the control covariates and demonstrate that the results do not depend on whether the control variables are included or not.

79

Second, as outlined above, the influence of trade on mediation belongs to the broader trade-conflict nexus. Since the direction of causality for the latter is likely to be reciprocal, doubts may persist about the direction of the trade-mediation relationship despite my argumentation above. Still, in order to control for this possibility, I run all analyses with the trade measures being lagged one year, but these estimations did not alter the substance of the findings.

Third, due to the structure of the ICM data, Chamberlain’s (1980) conditional logit estimator discards all conflicts that only have one observation, leading to a drop-out in N of about 300. This loss of information might induce selection bias. I therefore run the analysis again, using standard logit models. The findings reveal only minor differences, which are likely to be caused by the inconsistency of the estimator in this context.

Forth, the previous approach basically corrects the standard errors, but not the size of the coefficients. Hence, I also estimate the logit models with a mediation-years variable and different sets of cubic splines (Beck, Katz and Tucker 1998). Since the data have missing values on the dependent item, the mediation-years variable and the cubic splines for temporal correction become hard to interpret. However, the core results do not change significantly when incorporating the latter variables.

Fifth, there may be multiple reasons for not observing mediation in conflicts. A value of zero on Mediation Onset does not only include disputes without any conflict management attempt, but also non-mediation techniques such as bilateral negotiation, arbitration, referral to international organizations, etc. This could be an indicator for another selection problem. In order to control for this unobserved heterogeneity, I re-run the analysis using a preliminary version of a zero-inflated logit estimator (Beger et al. 2009). Although there are a few changes in terms of the control variables, the core items essentially remain robust.

Finally, my analysis employs Gleditsch’s (2002) trade data exclusively. These data, however, not only contain observations that are directly reported by official sources (i.e. IMF Direction of Trade Data, World Export Database, etc.), but also estimated trade figures via reverse flows and a relatively large number of cases where missing values were replaced via interpolating or extrapolating techniques (see also Barbieri, Keshk and Pollins 2009; Boehmer, Jungblut and Stoll 2009; Keshk, Pollins and Reuveny 2004: 1168f). Gleditsch’s (2002) imputation techniques may introduce noteworthy problems. The studies of Boehmer, Jungblut and Stoll (2009) and Barbieri, Keshk and Pollins (2009) point out that the assumptions made by Gleditsch could also induce biased estimates, since values are unlikely to be missing at random and missing data should not be treated as zero trade between countries. To consider this, I first estimated the

80

models again, relying only on those observations that were either directly observed or estimated. Second, I excluded all observations that were not directly observed, i.e. also dropping those cases that were obtained via reverse flows. The core results stay essentially the same. Unreported figures also unveil that this holds true for the postulated interaction effects between Dyadic Trade and Trade Centrality. The latter variable is an exception to this rule, however, since Trade Centrality only approaches conventional levels of significance in one of the models employing directly observed or estimated observations. One reason for this is likely to be the reduction in the variance of this item due to the drop of about 8% of the cases. A second reason could be problems in the Gleditsch (2002) data, although random measurement errors in general should increase the standard errors and reduce the likelihood of finding statistically significant results. While this raises the likelihood of missing true relationships, it is unlikely to increase the odds of finding spurious relationships. In other words, the relative inefficiency of statistical models based on noisy data does not invalidate statistically significant findings and against the background that all other variables, the interaction terms and Trade Ratio perform as expected, this sensitivity analysis enables us to place further confidence in the overall pattern of my results. 4.6. Discussion In this chapter, I offer a theory for examining the inconsistent findings on the impact of trade on mediation onset. My theory is based upon an opportunity costs logic that focuses on the possibility of substitution and sanctions. Trade links incorporate opportunity costs. Stronger and denser ties entail more opportunity costs, which in turn makes it more difficult and costly for a belligerent to circumvent sanctions or to find alternative markets for trade. By specifying the effects of bilateral trade and trade ties to third parties, my theory goes beyond existing theoretical arguments in two ways. First, I consider various trade ties instead of only bilateral, direct commerce. Second, unlike previous theories I explicitly argue that trade ties to outside parties are not necessarily cut off when two states enter a dispute.

The theory built upon that logic receives strong and robust empirical support for that bilateral trade between fighting states positively influences the chances of mediation in their conflict. If combatants shared dense and strong trade relationships with each other before the dispute, then they have incentives to enjoy the benefits of these already existing ties. This ultimately lowers violent behaviour and leads to conflict resolution through mediation. On the other hand,

81

additional trade channels to outside parties are likely to undermine the prospects for a peaceful intervention. Belligerents with multiple other alternatives for trade are not only able to circumvent probable sanctions, but also have the possibility to substitute for markets, which may be lost or are at risk to be lost due to the outbreak of a conflict. Similarly, third parties that provide those markets have an incentive to see the disputants continue fighting in order to make benefits. In addition, I argued for and found a different impact of bilateral trade conditional on trade ties to outside parties. However, the results do not confirm that the conditional effect applies the other way round. In short, trade links to outside parties can lower the opportunity costs of well established bilateral trade, since goods might be purchased elsewhere. Hence, although bilateral trade largely has a positive impact on mediation, we are more accurate in predicting the relationship between these variables when considering the trade linkages to third parties as well.

This chapter’s research has several notable implications. First, with regards to the policy advice, decision makers have to identify cautiously the different sorts of trade ties that belligerents have before mediation. Examining the trade patterns of fighting parties according to their perceptions of opportunity costs well before intervening may increase the chances of success in intervention efforts. Countries that share dense trade ties with each other will be more inclined to accept mediation offers than disputants that have multiple opportunities for obtaining the necessary goods of warfare through imports and exports. If conflict actors are of the latter type, then other forms of conflict management or intervention could be more appropriate. Furthermore, the interaction of different trade ties has to be acknowledged in advance. Sanctions might work for antagonists that are closely tied to each other via trade, but have few other possibilities to substitute for markets otherwise. As soon as prospective mediators face belligerents that have multiple opportunities for obtaining their goods through trade with outside parties, those disputants will most likely not be encouraged to accept mediation enforced by sanctions if they do not have an interest in a peaceful settlement on their own.

Second, although I clarified the impact of bilateral trade and trade linkages to third parties on mediation onset, I did not examine their effects on the effectiveness of mediation. Equally important, mediation is but one facet of multiple techniques for settling conflicts peacefully. Although I argued above that trade links determine especially mediation to appear as an appealing option, other instruments of conflict resolution, such as bilateral negotiations, arbitration, or adjudication may provide alternative attractive characteristics for the disputants for settling their conflict. In other words, it is plausible that the pacifying effect of trade ties does not only hold for mediation per se, but other

82

conflict resolution instruments as well. Since there is a large body of literature on both topics, they certainly are prospective avenues for future research.

Finally, my treatment of trade ties captures the average effect on mediation onset, but also opens up the possibility to distinguish between the kinds of goods that are exchanged between states. Some goods may be of a higher strategic value than others and might thereby be characterised by higher opportunity costs. Strategic or dual-use goods provide examples. My theory presents and the empirical analysis provides support for a theoretical argument that holds true in the given context. But the theory could easily be extended if one differentiates between different kinds of goods in order to make a more precise conclusion.

83

5. The Effectiveness of Tracks of Diplomacy Strategies in Third-Party Mediation

After the focus on mediation onset and belligerents’ interactions in the previous two chapters, I now turn to the effectiveness of international mediation and the coordination of interveners. Conflict resolution in inter- or intrastate conflicts has been on the agenda of international relations for a long time, but we know little about the effectiveness of third-party mediation (see Rubin 1981; Bercovitch and Langley 1993; Diamond and McDonald 1996; Dixon 1996; Bercovitch and Houston 2000; Gleditsch and Beardsley 2004; Nan and Strimling 2004; Beardsley 2006; Beardsley et al. 2006; Bercovitch and Gartner 2006a; Regan and Aydin 2006). Bercovitch and Gartner (2006a: 325) emphasize “which mediators use which strategies or resources in which conflicts and with what effects” as a key issue that future studies must examine.

In this chapter I will use the tracks of diplomacy (ToDs) framework to identify accordingly the most effective method for resolving conflicts (Ziegler 1984; De Magalhaes 1988; Diamond and McDonald 1996). The tracks of diplomacy concept has the advantage of providing a coherent and parsimonious framework to consider variation in mediator characteristics, their resources, strategies and interaction. ToDs can be defined as diplomatic initiatives by outside state or non-state parties to transform a dispute through communicating information, proposing new solutions and influencing directly the crisis through carrots and sticks that can help generate movement toward potentially overlapping bargaining positions (see Crumm 1995; Dixon 1996: 653; Bercovitch and Jackson 1997: 25f; Jackson 2000: 324; Frazier and Dixon 2006: 387; Regan and Aydin 2006: 745). It is common to distinguish between official and unofficial ToDs (Montville 1991; Diamond and McDonald 1996; Mapendere 2000; 2005; Nan 2002; 2003; Nan and Strimling 2004). Official diplomacy such as Track One (T1) involves state and official actors. Unofficial diplomatic intervention is formed by Track Two Diplomacy (T2) and Track One-and-a-Half Diplomacy (T1.5). T2 encompasses unofficial, informal interaction between members of adversarial groups or nations. Similarly, T1.5 comprises public or private interaction between official representatives of conflicting actors mediated by a third party not representing a political institution.

T. Böhmelt, International Mediation Interaction, DOI 10.1007/978-3-531-92812-8_5,© VS Verlag für Sozialwissenschaften | Springer Fachmedien Wiesbaden GmbH 2011

84

I intend to contribute to the existing research on third-party intervention by developing and empirically testing a theoretical model to explain which track – both on its own and interacting with other tracks – is likely to be most effective. As far as I am aware, this is one of the first large-N study on ToDs and mediat-ion effectiveness. I argue that the leverage and resources of ToDs determine effective outcomes. The more leverage is available and resources are invested by a specific track, the more effective a track intervention (see Diamond and McDonald 1996; Notter and McDonald 1996: 36; Chigas 2003). ToDs that have more resources and leverage than other interveners are more likely to apply more enforcing strategies. This also yields more effective outcomes (see Wilkenfeld et al. 2003; Beardsley et al. 2006; Quinn et al. 2006). Finally, if official T1 diplomacy is facilitated by unofficial approaches, i.e. if T1 and T2/T1.5 interact with each other, then mediation effectiveness is likely to be higher as it helps actors pooling their individual resources, decreases uncertainty and ensures the support at the grassroots level.

My empirical test of the theory extends previous research evaluating mediation effectiveness as a two-stage process (Gartner and Bercovitch 2006; Schneider, Bercovitch and Selck 2006). Assessing the effectiveness of particular forms of third-party interventions that we observe in disputes must also consider the prior selection stage of intervention, or the question of how conflict characteristics may influence the willingness of third-parties to intervene in conflicts. Both stages are intertwined and the conflicts that actually see track mediation are unlikely to be a random sample. On one hand, we may surmise that third parties are more likely to intervene in disputes that are easy to mediate. However, it may also be the case that third parties primarily intervene in the more difficult disputes, as these cases pose higher risks of instability to the international order. If the first scenario applies, we may exaggerate the effectiveness of mediation, while the latter may lead us to underestimate it. More generally, previous research ignoring the role of selection may have yielded misleading results on the effects of third-party intervention (Gartner and Bercovitch 2006; Schneider, Bercovitch and Selck 2006).

The remainder of this chapter is structured in the following way. I first review the relevant literature on ToDs. I then develop a theoretical model that combines ToDs, their leverage and resources with mediation effectiveness. Afterwards, I detail an appropriate research design to test the theory, using an estimator that takes selection into account. I conclude with a discussion of my findings and their implications.

85

5.1. Literature Review: The Concept of Tracks of Diplomacy As briefly outlined in the introduction, research on international mediation has examined the impact of multiple factors that may influence the outcome of a third-party intervention. These features can be summarised under three con-ceptual clusters (Bercovitch and Langley 1993: 673; Kleiboer 1996; Jackson 2000: 323). The first cluster pertains to characteristics of the disputants. For example, Terris and Maoz (2005) suggest that disputing democracies are more likely to accept mediation and agree on effective outcomes than autocracies. Schwarzer (1998) shows that effective mediation is more likely when the capabilities of the belligerents are balanced. The second cluster includes characteristics of the dispute. Bercovitch and Langley (1993) show that the duration of a conflict negatively effects mediation and that longer conflicts are less likely both to see third-party intervention as well as an effective outcome. Quinn et al. (2006), Bercovitch, Anagnoson and Wille (1991) and Bercovitch and Gartner (2006b) argue that high-intensity disputes with a high number of fatalities are more likely to experience third-party mediation, as an ongoing conflict with many casualties poses a more severe threat to international security. The third cluster encompasses factors that describe the mediators or type of mediation. Kydd (2006), for example, demonstrates that neutral interveners are more likely to reach effective outcomes, as these are generally more trusted than actors biased towards one of the disputing parties. Bercovitch and Schneider (2000) furthermore show that UN Security Council members are more likely to receive mediation offers.

Despite this recognition that characteristics of mediators and the type of mediation may be important, existing work has generally ignored the concept of ToDs as a framework to capture the structure of mediator characteristics, resources, strategies and interaction. T1, the intervention of decision makers, involves state and other official actors (Diamond and McDonald 1996: 28ff; Chigas 2003; Nan 2003). T1 is essentially an interstate process where communication goes from one official party directly to the decision-making apparatus of another actor (see Said and Lerche 1995: 69; Nan 2003). Examples of T1 include the UN intervention during the crisis with Iraq in 1991 or the in-volvement of Britain’s Lord Carrington during the Rhodesia-Zimbabwe in-dependence negotiations. T1 works on behalf of the authority and power of an official entity. The skills, resources and interests from these “principals” directly influence the performance of T1 (Diamond and McDonald 1996: 30; Mapendere 2000: 45).

T2 is an unofficial, informal interaction between members of adversarial groups or nations with the goals of developing strategies, influencing public

86

opinions and organising resources in ways that might help resolve the conflict (Montville 1991; Diamond and McDonald 1996: 37ff). T2 actors are recruited from unofficial representatives, NGOs, regional and local leaders or grassroots groups. The foreign campaigns against Apartheid in South Africa or the en-gagement of Harvard professor Roger Fisher in El Salvador are examples of T2. T2 can be more subtle, personal and free from the constraints of T1, as it involves NGO activity and back channel measures (see Chataway 1998: 270; Boehmer, Gartzke and Nordstrom 2004: 13; Kydd 2006: 451). Montville (1991) and Chigas (2003) argue that value-based conflicts about identity, survival and fears of the other only can be effectively addressed by T2 diplomacy that seeks to change the underlying relationships so as to promote a mutual understanding and acknowledging each others’ concerns.

T1.5 emerged as a response to the problem when official actors have no incentives to engage in a conflict and/or T2 efforts show no effect at the grassroots level (Mapendere 2000; 2005; Nan 2002; Gleditsch and Beardsley 2004: 400). As an unofficial track, T1.5 is public or private interaction between official representatives of disputants that is mediated by a third-party not representing a political institution (Mapendere 2000: 16; Nan 2002). As such, the main difference between T2 and T1.5 as unofficial mediation interveners lies in their level of operation. While T2 actors engage solely at the grassroots or middle leadership stage, T1.5 operates with the top leadership or official diplomacy level (Mapendere 2005: 76). T1.5 actors span former prominent politicians such as previous U.S. president Jimmy Carter, academic scholars or religious institutions. Examples of this track include the Oslo Peace Process, the work of Sant'Egidio in negotiating between the warring parties in Mozambique or the Guinea Worm Ceasefire mediated by the Carter Center in Sudan. This track, although being categorised as unofficial intervention, can combine the strengths of both T1 and T2 to a certain degree, but potentially also their weaknesses. On one hand, T1.5 efforts can influence actors’ relative power as it partly relies on resources and power capabilities provided by official parties, even if it is not driven by governmental agendas (Mapendere 2005: 72ff). On the other hand, T1.5 may be able to establish a T2-like private and non-public environment.

Although we have a great deal of research on ToDs as well as on particular cases where different tracks have been applied, few studies have empirically examined how specific tracks may vary in effectiveness (Kaye 2001). For example, Regan and Aydin (2006) argue that track mediation, especially when it interacts with other intervening actors, should be likely to significantly reduce the duration of a conflict. However, they focus only on official or T1 interventions in interstate conflicts. Dixon (1996) found that the type of track

87

strategy matters for the prospects for peaceful settlement and that more enforcing strategies seem to induce more effective outcomes. However, he did not differentiate among specific types of third-party actors. Finally, Lieberfeld (2002) examines the contribution of T2 to effective conflict resolution in South Africa. He demonstrates that T2 efforts influenced the peace negotiations during the Apartheid era, but ignores the impact of other tracks. Further, the findings from his qualitative case study of three T2 interventions offer only limited generalisable findings.

In sum, previous work on ToDs suffers from certain limitations and constraints and the existing literature on third-party interventions has not taken advantage of the possibility of using ToDs to measure mediator characteristics in a comprehensive and systematic way. Hence, a large-N study examining a wider range, characteristics and the interaction of ToDs as well as their impact on mediation effectiveness can contribute both to the general literature on inter-vention and broaden existing work on ToDs. 5.2. The Effectiveness of Tracks of Diplomacy Strategies – Theoretical

Model 5.2.1. Understanding Mediation Effectiveness My aim here is to relate variation along the ToDs to provide an answer to the question which mediating tracks are more successful than others in third-party conflict resolution. However, existing research does not provide a standard template for evaluating effectiveness (Bercovitch 1985; Bercovitch, Anagnoson and Wille 1991: 8; Kleiboer 1996: 361f, 376; Lieberfeld 2002: 355). For example, Wilkenfeld et al. (2003; 2005) measure the effectiveness of third-party mediation by the extent of actor satisfaction with the outcome. Other scholars (e.g. Ott 1972; Maoz and Terris 2006; Schneider, Bercovitch and Selck 2006) simply classify outcomes as “successful” and “unsuccessful” dichotomously in a rather imprecise fashion. I employ conceptual research on mediation effectiveness by distinguishing along the following range of outcomes (see Bercovitch and Langley 1993: 674). First, there are cases without a peaceful settlement and the warring parties continue fighting. This includes disputes where no conflict management took place, where mediation has been offered but not accepted and failed attempts of track mediation. Second, we have cases where formal agreements have been negotiated and signed by the actors. An agreement by itself does not, however, necessarily imply that the conflict will not re-emerge in the near future. Third, there are conflicts that have partially

88

been settled and see some improvement or adjustment of the warring parties. But, even if actors are better off after the involvement of ToDs elicits cooperation, there may be uncertainty over whether the disputing parties will comply with the obligations (see Dixon 1996; Walter 1997: 336; Terris and Maoz 2005; Kydd 2006; Regan and Aydin 2006: 739f). Finally, we have cases where the conflict has been fully settled (see Jackson 2000: 330f; Kaye 2001: 50). 5.2.2. Official Diplomacy – Track One T1 is an official diplomatic effort, where states or official international organisations such as the UN intervene and mediate in conflicts. This track directly relies on state power with high-ranking government officials, diplomats or heads of state aiming to influencing political power structures (Mapendere 2005: 67; Sanders 1991). It is generally better funded than unofficial efforts and supported by intelligence, security as well as logistical resources rarely available to non-state actors (Stein and Lewis 1996; Bercovitch and Houston 2000; Schneider, Bercovitch and Selck 2006: 11). For example, Lord Carrington could rely on Britain's colonial control over Rhodesia and the resources of the UK government during the Lancaster talks about Rhodesia-Zimbabwe independence in 1979. Withholding recognition of independence until a settlement was reached allowed Britain to demonstrate its power and exert a strong control over the negotiations. The greater power and resources for peaceful conflict settlement makes T1 likely to be the most effective track (Rubin 1981; Diamond and McDonald 1996: 40, 150; Mapendere 2005: 68; Nan 2003; Frazier and Dixon 2006: 402; Schneider, Bercovitch and Selck 2006: 11f). Only T1 can rely on clear positive or negative incentives such as financial, political or military support as well as coercion or force (Crumm 1995). These incentives are provided through directive strategies or manipulations, i.e. using leverage to directly influence the crisis bargaining process. This changes the perceived costs of conflict and/or the benefits of a peaceful settlement (Touval and Zartman 1985). Such efforts attempt to shift actors’ reservation points to increase the probability of identifying mutually superior alternatives to continuing fighting. Coercive or enforcing strategies yield generally more effective outcomes as they are more likely to transform a game of conflict into one of cooperation (Chigas 2003; Beardsley et al. 2006). Such strategies are better able to show belligerents mutually superior alternatives to fighting. Even though the disputing parties may not be convinced about these alternatives at the outset, coercive and enforcing efforts can make it harder for belligerents to dismiss these possibilities (see

89

Notter and McDonald 1996: 36; Bercovitch and Lee 2003; Mapendere 2005: 67; Nan 2003; Beardsley et al. 2006; Bercovitch and Gartner 2006b: 338f). During the Indo-Pakistani war of 1965, for example, the UN was able to enforce a ceasefire through the manipulative intervening strategies of the U.S., which placed a military embargo on both states. Pakistan in particular subsequently became more willing to negotiate the dispute peacefully as it was highly dependent on U.S. arms supplies. Also, the initial 1994 mediation efforts of the Contact Group in Bosnia, i.e. the U.S., Russia, France, Britain and Germany, failed mainly due to Bosnian Serb unwillingness to agree to the terms and the imposed embargo on the Drina river. The NATO operation “Deliberate Force,” which bombarded the Bosnian Serb military, was intended to bring the Bosnian Serbs back to the negotiation table through negative incentives or sanctions.

To summarise, due to T1’s vast amount of resources, leverage and enforce-ment power, I argue:

Hypothesis 5.1. Greater leverage, resources and prospects for

enforcement will make T1 the most effective track. 5.2.3. Unofficial Diplomacy – Track Two and Track-One-and-a-Half Although states undeniably have more resources, Schneckener (2002), for example, holds that less leverage can promote more effective outcomes in some instances. In particular, T1 can encourage competitiveness as well as positional bargaining (Nan 2003). Furthermore, T1 is highly dependent on the position and interests of the “principal actor” (Mapendere 2005: 67f) and such official diplomatic efforts may be less flexible due to their ties to states (Nan 2003). The Israelis and Palestinians have, for example, both increasingly become sceptical about the use of T1 in the U.S. based “Road Map for Peace” initiative (Agha et al. 2003). Israelis hold that they will lose too much by accepting the Road Map, while Palestinians believe that the proposal only suits the interests of the U.S. as an ally of Israel. A less coercive approach than T1 may help overcome such fears of competitiveness, bargaining and biased interests (see Pearson 2001; Lieberfeld 2002; Mapendere 2005; Bercovitch and Gartner 2006b: 333).

Unofficial approaches such as T2 and T1.5 may be able to do this. These tracks do not primarily rely on coercion or exercising power, but rather seek to foster support for agreements at the local level (Diamond and McDonald 1996: 38). T2 efforts can be effective through convincing actors, appealing to their common understanding and establishing a peaceful settlement in private discussions (Lederach 1997; Kaye 2001; Mapendere 2005: 68). This is done on

90

one hand by developing and maintaining a wide network of contacts, able to spread respect and trust among the disputing parties in the intermediary and the intermediation process (Mapendere 2000: 49) as well as providing a neutral, low-key, safe and non-judgmental environment such as in workshops or reconciliation programmes to facilitate interaction (Chigas 2003). These circum-stances can make participants freer to share fears and explore ideas for resolution, free from the constraints of government positions in a non-binding and flexible way. In this sense, T2 is a form of what Kydd and Snidal (1997: 123f) characterise as “cheap talk” strategies that can convey information about the range of mutually acceptable agreements and sustain coordination. Afterwards, parties discuss ways to end a conflict, but “can walk away from them without penalty.” More generally, T2 interventions can bring the antagonists closer to peaceful settlement without committing the parties too early to a binding agreement. This track mainly relies on communication-facilitation. As an integrative strategy,22 it “serves as a channel of communication among disputing parties” (Keashly and Fisher 1996), but does not contribute materially to the negotiation process itself. Unofficial T2 mediators have applied communication-facilitation strategies in the conflict between Russia and Chechnya, where multiple workshops and talks between grassroots leaders on both sides have been organised since 1992.

T1.5 as another unofficial approach is a track that partly relies on the top-down approaches and coercive techniques of T1 (Nan and Strimling 2004; Mapendere 2005: 66). Mapendere (2005: 70) argues that the ability to switch between T1 and T2 conflict resolution techniques depending on the specific situation faced can make T1.5 efforts effective in conflict resolution. T1.5 may, for example, rely on lower-level mediation forms to gain the trust of the parties when a direct, high-level intervention is not possible (Mapendere 2005: 73). T1.5 may also influence power structures without being seen as completely “driven by governmental political agendas” (Mapendere 2005: 72). The Carter Center’s 1995 Guinea-Worm Ceasefire negotiations in Sudan demonstrate this diplomatic agility. “The USA had branded Sudan a supporter of world terrorism and accusations of religious persecution were a major issue in the civil war” (Carter 1995: 184). The hostile relationship between the U.S. and Sudan made it im-possible for the former to serve as a mediator in Sudan’s civil war, but efforts combining the distinctive elements of T1 and T2 helped fill this gap. T1.5 efforts normally employ procedural strategies, where ToDs mediators “conceive and propose new solutions to disputants” (Beardsley et al. 2006). Beardsley et al.

22 An integrative strategy allows the disputing parties to identify alternatives that are already present within an overlapping bargaining space (Beardsley et al. 2006: 63).

91

(2006: 63) conclude that this style can be more effective than pure communicating-facilitation by creating focal points for the combatants. The T1.5 attempts in the Georgia-South Ossetia Abkhazia secessionist conflict provide examples of procedural strategies.

However, T2 and T1.5 efforts also have characteristics that could make dispute resolution success less likely. First, adversaries may argue that non-state third parties are overly dependent on their home countries and their official agendas. For example, when former U.S. president Carter mediated between the U.S. and North Korea in 1994, North Korea had just withdrawn from the International Atomic Energy Agency, thereby raising fears that the country was developing nuclear weapons capabilities. Carter went to North Korea as an unofficial party, representing the Carter Center – a non-governmental institution. His mediation efforts ended the crisis in the short-term (Mapendere 2005: 74), but a long-term settlement proved illusive as the North Koreans perceived Carter as an official representative of the U.S. Second, T1 elements may undermine the low-key environment intended to ensure pressure free discussion among belligerents associated with T2 interventions. Finally, unofficial efforts often take a long time to produce results (Chigas 1997; 2003; Lederach 1997; Chataway 1998: 278; Nan 2003; Mapendere 2005: 68). Even more important, although some unofficial efforts such as T1.5 can in principle capitalise on the capabilities and resources provided by official actors, they may still lack sufficient resources and power to enforce agreements (Mapendere 2005: 71). This suggests the following hypothesis:

Hypothesis 5.2. T2 and T1.5, i.e. unofficial diplomacy will be less

effective than T1, since these tracks have less leverage and resources.

5.2.4. Multi-Track Diplomacy – Combined Efforts of Official and Unofficial

Tracks Many people view T2 and T1.5 less as tracks on their own and more as useful supplements to T1 efforts (Davidson and Montville 1981; Diamond and McDonald 1996; Chataway 1999: 275; Kaye 2001: 52; Lieberfeld 2002: 358; Nan 2003). As outlined above, given the lack of political power and resources for sustained leverage and the long time before results can be obtained, I expect T2 and T1.5 to be generally less effective than T1 when these tracks are used on their own. In order to be effective, unofficial tracks must have some opportunity to influence T1 actors through introducing their ideas and results into official

92

peace negotiations (e.g. Notter and McDonald 1996: 33; Chataway 1998: 271ff; 1999; Mapendere 2005: 76). A treaty with enforcement provisions will often be necessary to solve a conflict and this can generally be done by official actors only who have access to political power and capabilities (Mapendere 2005: 68). For example, the initial Israeli-Palestinian peace process began with informal discussion between Yair Hirschfeld, an Israeli academic and the PLO official Abu Alaa, mediated by Norwegian sociologist Terje Rød Larsen. This resulted in a “declaration of principles” between the officials of the two parties that established the framework for later mediation efforts. Thus, different track efforts may each influence the likelihood and effectiveness of other activities in an interactive manner, leading to “multi-track diplomacy” (Diamond and McDonald 1996: 30, 156ff; Crocker, Hampson and Aall 1999; Mapendere 2005: 74ff). The Contact Group for Bosnia used the combined leverage of resources, intelligence and logistics of their member states to enhance the effectiveness of its mediation effort. Such interaction between the track parties is shaped and influenced through common ties or social relationships that help actors pool their individual resources to increase their effectiveness (Maoz, Kuperman, Terris and Talmud 2006: 665; Dorussen and Ward 2008a). Interacting ToDs are more likely to establish a stable peace agreement and increase effectiveness (see Greig and Diehl 2006: 363f; Kydd 2006: 450f). First, combined ToDs can create obligations and expectations that help enforce compliance (see Ward 2006). Due to the intervention of unofficial tracks, compliance is likely to be enforced at the grassroots levels as well. Second, since non-cooperating actors can be excluded from other forms of reciprocity, interaction increases the likelihood that track parties will pursue shared goals through complementary means. Third, the connection of issues and interest establishes trust, a crucial element for over-coming collective action problems. Finally, track interaction enables comm-unication among actors, which facilitates coordination and decreases uncertainty (see Kaye 2001: 64; Frazier and Dixon 2006: 391ff).

In other words, the effect of combined efforts of official and unofficial diplomacy is hypothesised to be as follows:

Hypothesis 5.3. T1 efforts facilitated by unofficial diplomacy will yield more effective outcomes than independent efforts.

93

5.3. Research Design 5.3.1. Methodology Since those conflicts where we actually see mediation efforts are unlikely to be a random sample of conflicts with opportunities for mediation, I follow recent research conceptualising tracks of diplomacy effectiveness as a function of a two-stage decision-making process (Gartner and Bercovitch 2006; Schneider, Bercovitch and Selck 2006). The first stage is the selection of mediation, while the second stage pertains to the effectiveness of specific tracks. Since features that are associated with mediation effectiveness may also affect the prospects for whether mediation takes place, the two stages must be considered jointly. Understanding selection requires that we examine the features that may influence actors’ decision to intervene or not. Existing literature has suggested a number of structural variables that may affect the decision to mediate, which will be discussed in greater detail below as control variables when evaluating effectiveness. Consistent with Schneider, Bercovitch and Selck (2006), I assume that belligerents choose between specific tracks based on their anticipated effectiveness after mediation has started and not before deciding to solicit mediation or not.

Although I argue that track related characteristics only exert an influence at the outcome stage, many important control variables for assessing the effectiveness of ToDs are also likely to be important at the selection stage. Since the stages are unlikely to be independent of each other and features that affect the likelihood of mediation may also have an impact on the prospects for success, treating the two stages as independent of one another can lead us to either under- or overestimate the effectiveness of third-party interventions (Schneider, Ber-covitch and Selck 2006: 5f). Because of this self-selective character of ToDs, applying a standard OLS regression estimator to effectiveness in mediated conflicts can give rise to misleading and biased findings (Heckman 1979). To deal with the problem of selection, I employ a Heckman selection model, where the estimated mean function in the outcome stage is conditioned on the first stage selection process and thereby provides a consistent estimate for the truncated distribution of the second stage sample (Heckman 1979). This procedure also follows closely my theoretical argument, since the track characteristics are assumed to affect only the outcome and not all structural control variables influence both stages. The heckit estimator is appropriate as long as one can find at least one variable that influences only the first but not the second stage and is more robust than competing estimators.

94

5.3.2. Data Once again, the data are taken from the ICM data set (Bercovitch 1999). While 1,859 were ToDs mediation of the form I focus on in this chapter, 194 dispute management efforts experienced combined official and unofficial ToDs mediation efforts. After accounting for missing values in the data, the selection stage of the empirical analysis considers 3,166 observations, while the second stage considers the sample of 1,321 conflict cases where ToDs mediation occurred. 5.3.3. Operationalisation The conflict outcome variable is measured according to my definition of mediation effectiveness. I measure ToDs mediation effectiveness by a six-point scale ranging from 0 (no mediation) to 5 (full settlement of conflict).

Second, I identify the distinctive tracks through the functional identity of the mediator, i.e. the nature of the group the mediator represents (Bercovitch 1999: 20). From this I derive two dummy variables for both official and unofficial ToDs to show their direct effect. Finally, I include a binary variable for the interaction of official and unofficial track mediation, which takes on a value of 1 if a conflict observation experienced the intervention of combined official-unofficial tracks, i.e. multi-track diplomacy. 5.3.4. Control Variables Track activity takes place in an environment where third parties both choose whether to intervene and the form of intervention (Ott 1972: 597; Bercovitch 1997; Greig 2005; Beardsley et al. 2006: 73). I estimate the likelihood of specific tracks being chosen based on features shown in previous research to affect the prospect and effectiveness of third-party mediation (Bercovitch and Langley 1993; Kleiboer 1996; Mapendere 2000: 41; Pearson 2001; Greig 2005; Terris and Maoz 2005; Schneider, Bercovitch and Selck 2006: 7ff).

First, as elaborated in the previous two chapters, mediation occurs according to both demand and supply (Beardsley 2006; 2010). I thus include the same two variables tapping the incentives of the actors for mediation in the first stage.

Second, power discrepancies were identified to decrease the incentives for the most powerful party to agree to a peaceful settlement as it stands a better chance of winning the conflict militarily in Chapter 4. Hence, I consider the

95

capabilities of the fighting parties here as well, but use the absolute differences of the Cox-Jacobson scale of power relations (Bercovitch 1999: 44f)23. Third, based upon my previous discussion and empirical findings I also consider a possible inverted U-shaped relationship between conflict intensity and track mediation onset as well as effectiveness, by including the square of the intensity variable: both low and high intensity disputes are less likely to see interventions and effective outcomes (Mason and Fett 1996: 551). Forth, consistent with the literature (Bercovitch and Langley 1993; Kleiboer 1996: 362f; Pearson 2001: 276; Chigas 2003) and my previous analyses, the succeeding empirical models also include a duration variable at the selection stage.

Table 5.1. Descriptive Statistics of Variables.

N Mean Std. Dev. Min Max Effectiveness 3207 2.62 1.1 0 5 Official Diplomacy (Track 1) 3207 0.47 0.5 0 1 Unofficial Diplomacy (T2 – T1.5) 3207 0.05 0.21 0 1 Official – Unofficial Interaction 3207 0.06 0.24 0 1 Track Dummy 3207 0.58 0.49 0 1 Third-Party Incentive 3207 0.25 0.43 0 1 Disputants Incentive 3207 0.63 0.81 0 2 Power Difference 3207 2.03 0.8 1 3 Intensity 3166 1.1 1.1 1 4 Intensity(sq) 3166 5.19 5.09 1 16 Duration 3207 92.75 116.1 0 900 Ethnicity 3207 0.03 0.18 0 1 Associate 3207 0.75 0.43 0 1

In addition, many argue that ethnopolitical and identity-based disputes are

more difficult to settle (Diamond and McDonald 1996: 133ff; Rothman and Olson 2001: 289; Schneckener 2002) and agreements are likely to be less stable in a fractionalised environment with competing ethnic groups and identities

23 As indicated above, the Cox–Jacobson power index score is similar to the Composite Index of National Capability (CINC) of the Correlates of War Project (Singer 1988). It is calculated by adding its scores on the measures of GNP, GNP per capita, military spending, population and territory. All currency–based measures are in US dollars at current prices. For further specification see Cox and Jacobson (1973: 437ff). The reason for using this scale instead of the CINC score is that non–state actors’ power levels are captured by the Cox–Jacobson index as well.

96

(Chigas 2003; Doyle and Sambanis 2000: 783; Pearson 2001; Rothman and Olson 2001: 291f; Schneckener 2002; Walter 1997: 343). Hence, I consider a dichotomous variable for ethnopolitical issues.

Finally, a larger number of parties involved in a dispute may increase un-certainty (Walter 2002; Schneider, Bercovitch and Selck 2006). This makes it more difficult for the belligerents to agree on mediation and for ToDs to mediate effectively. Following the operationalisation of Schneider, Bercovitch and Selck (2006), I use a dummy variable that has the value of 1 if at least one other party actively supports one of the belligerents.

Table 5.1 provides the descriptive statistics of the variables. 5.4. Results I estimate three models to demonstrate that my findings are robust and take the censoring of ToDs effectiveness into account.24 The first model includes the track variables and the control variables on the first stage only. The second model is limited to the control variables for both stages. Finally, the third model includes all explanatory variables for both stages.

Table 5.2. clearly demonstrates that conflicts that experience intervention cannot be considered a random sample of conflicts. The significant estimate of the parameter indicates that the Heckman selection model fits the data better than independent estimates of the selection and outcome equations. The is an estimate of the correlation of the error terms in the two stages. Its estimate can be highly sensitive to model specifications and should be expected to be negative if unobserved features that increase the likelihood of selection (i.e. track mediation) actually decrease the probability of more effective outcomes. The fact that is positive in all the models here implies that unobserved features that make interventions more likely also induce more effective outcomes. This may perhaps be likened to a positive impact of unobserved forms of resolve among the inter-veners as the literature suggests multiple other forms of unobserved influences that could have these effects arising from the characteristics of the disputants and 24 I report robust standard errors to correct for the bias due to non–constant variances. Next to running three models, to test the robustness of my findings I also employed other econometric instruments that are not reported in this chapter. First, as my unit of observation are sub–cases of particular conflicts, the errors may not be independent or identically distributed from observations of the same dispute due to common factors not explicitly considered. Thus, I estimated the models with robust standard errors through clustering on conflicts. Second, I re–sampled the observations with replacement using the non–parametric bootstrapped method. Third, in order to control for multicollinearity, I calculated the variables’ cross–correlation. The robustness tests did neither differ significantly nor indicate that the main results reported here have to be interpreted with caution.

97

the conflict itself (see Bercovitch and Langley 1993: 673; Kleiboer 1996; Gartzke 1999; Jackson 2000: 323). The significant Wald statistics for all three models allow me to reject the null-hypothesis that all coefficients are jointly insignificant and indistinguishable from zero. This suggests that my models im-prove on Schneider, Bercovitch and Selck (2006), who only obtain significant Wald statistics for two out of three of their suggested specifications. Both the ToDs and the control variables are largely highly significant and have the hypothesised signs. However, the interpretation of heckit model results is often not straightforward from coefficient estimates (Heckman 1979; Sigelman and Zeng 1999). The coefficient for a variable x that appears only in the outcome equation can be interpreted as the marginal effect of a one unit change in x on the dependent variable y. However, if a variable appears in both the selection and outcome equations, then its coefficient in the outcome equation must be interpreted conditional on its inclusion in the selection equation. Sigelman and Zeng (1999: 179) suggest that a more informative way to evaluate the contribution of an independent variable in this case is to consider its marginal effect on the conditional expectation of the dependent variable. Where appropriate, I report these estimates.25

Starting with an interpretation of the first stage variables, the term for ethnicity is insignificant and generally has a positive sign. This is contrary to my expectations, but supports other studies suggesting that ethnopolitical fractionalizations do not notably influence the course of disputes significantly (e.g. Fearon and Laitin 2003). Duration has a negative coefficient, which is significant in all models, but its relatively small size suggests that the effect is minimal. The estimated coefficients for the structural control variables of third-party incentives, power discrepancies and intensity in the first stage are all consistent with my expectations. Third parties are more likely to intervene when they have incentives to do so. Large differences in capabilities make track mediation less likely. Furthermore, I find a non-linear inverse relationship between a dispute’s intensity and mediation onset, where track interventions are the least likely in conflicts that are either relatively more easy or difficult to solve. However, contrary to common arguments in the literature, I find a positive sign for association and a negative sign for incentives of disputants. The result for the association variable means that the likelihood of intervention is higher if at least one of the belligerents is supported by another party other than the track

25 Sigelman and Zeng (1999) suggest to calculate the “true beta coefficient” by the coefficient in the outcome equation minus a product of 1) the corresponding coefficient in the selection equation, 2) the correlation between the errors in the two equations, 3) the error from the outcome equation and 4) a function of the inverse mills ratio (Heckman 1979). The calculations can be obtained from the author upon request.

98

Table 5.2. The Effectiveness of Tracks of Diplomacy Strategies.

Model 1 Model 2 Model 3

Effective-

ness Track

Selection Effective-

ness Track

Selection Effective-

ness Track

Selection Official Diplomacy (Track 1) 2.27 1.70 (0.04)*** (0.16)*** Unofficial Diplomacy (T2 – T1.5) 2.11 1.61 (0.09)*** (0.18)*** Official – Unofficial Interaction 2.52 1.89 (0.08)*** (0.19)*** Third-Party Incentive 1.79 1.17 1.70 (0.13)*** (0.27)*** (0.14)*** Disputants Incentive -0.51 -0.41 -0.51 (0.03)*** (0.10)*** (0.03)*** Power Difference -0.10 -0.05 -0.07 -0.08 -0.12 (0.03)*** (0.00)*** (0.03)** (0.00)*** (0.03)*** Intensity 0.68 1.02 1.26 0.29 0.81 (0.14)*** (0.01)*** (0.12)*** (0.00)*** (0.14)*** Intensity(sq) -0.13 -0.22 -0.27 -0.07 -0.16 (0.03)*** (0.00)*** (0.02)*** (0.00)*** (0.03)*** Duration -0.0007 -0.0003 -0.0006 (0.00)*** (0.00)* (0.00)*** Ethnicity 0.03 0.07 0.16 -0.001 0.04 (0.18) (0.00)*** (0.16) (0.00)*** (0.18) Associate 0.50 0.29 0.61 0.13 0.56 (0.06)*** (0.00)*** (0.06)*** (0.00)*** (0.06)*** Constant -0.46 -1.12 -0.58 (0.15)*** (0.12)*** (0.15)*** N (Censored N) 3166 (1321) 3166 (1321) 3166 (1321) Log Pseudo Likelihood -4118.26 -4154.77 -4094.72

0.52 0.89 0.61 Wald ( ) 86.34*** 12.04*** 84.71*** Wald 4426.98 1163.38 4273.18

Note: robust standard errors in parentheses; Sigelman and Zeng (1999) corrected coefficient estimates reported where appropriate; * significant at 10%; ** significant at 5%; *** significant at 1% (two-tailed).

99

mediator. Although I will refrain from speculating on the origin of this finding in much detail here, since it runs counter to initial expectations, one might argue from a network perspective that possible indirect links between belligerents and prospective track mediators actually could increase willingness to contribute to a peaceful settlement by intervention (see Böhmelt 2009; Dorussen and Ward 2008a). Finally and consistent with the previous two chapters, the disputants’ incentive variable has again a negative sign. I will come back to this result in the conclusion.

The findings for the structural control variables in the outcome equation mirror the results for the selection stage. Power disparities among belligerents decrease the effectiveness of track mediation. The predicted inverted U-shaped relationship between conflict intensity and ToDs effectiveness is highly significant. As at the first stage, the association variable is positive and significant at the 1% level, while the ethnic variable becomes significant in the full model after employing Sigelman and Zeng’s (1999) correction.

With regard to the relevant hypotheses, Table 5.2. shows that all track related variables have positive and significant coefficients, which provides strong support for the utility of the theoretical framework of ToDs. The findings indicate that track intervention has a positive impact on the effectiveness of intervention. Moreover, the relative size of the coefficients demonstrates that T1 is more effective than unofficial efforts such as T2 and T1.5. Even though official and unofficial tracks still have a positive effect on conflict resolution, it seems that resources, leverage, coercion and enforcement power in general do matter and make T1 most effective. Table 5.2. reveals that T1 mediation increases the chances of an effective outcome by about 170% in Model 3. This gives strong support for the chapter’s first hypothesis. In other words, individuals, NGOs or other non-official actors may have an influence, but official parties like states and IGOs ultimately matter more. Nonetheless, the im-pact on the peaceful settlement of conflicts is still positive for unofficial mediation, but less by about 16%-points (Model 1) to 9%-points (Model 3). The chapter’s second hypothesis seems to be confirmed, therefore. Finally, inter-acting tracks are likely to produce synergistic outcomes. This variable is significant at the 1% level. Obtaining evidence for my third hypothesis, this clearly demonstrates that T1 mediation is more effective when it is facilitated by unofficial tracks and vice versa. The positive impact on ToDs effectiveness is about 20%-points higher than pure official interventions. The marginal effect of combined efforts approaches 252%, when the controls are excluded in the outcome equation (Model 1) and about 189% in the full model.

As such, we can conclude not only that track intervention has a significant impact on peaceful conflict resolution, but also that the specific kind of track

100

itself actually is of higher importance. These findings do not support the claim of Ott (1972) that the particular kind of mediator does not matter notably for effectiveness. 5.5. Conclusion This chapter has sought to expand our understanding of the effectiveness of different tracks of diplomacy in inter- and intrastate conflicts. Whereas previous research has tended to treat conflict mediation efforts in an undifferentiated or truncated manner, I have developed a theoretical model that considers different ToDs, their interaction and effectiveness. Although existing research has told us that several structural features are likely to be important, this is only one of the earliest works (Gartner and Bercovitch 2006; Schneider, Bercovitch and Selck 2006) to combine structural variables and third-party characteristics in a joint model while controlling for possible selection effects and the first seeking to assess possible differences in effectiveness between different ToDs both in the case of unilateral interventions and coordinated efforts. My findings suggest that more enforcing ToDs with high leverage and resources have a significant positive impact on the effectiveness of third-party mediation and show strong evidence for differences in the relative effectiveness of independent interventions as well as cases where official and unofficial tracks jointly mediate. Therefore, my research implies that the policy community has to think carefully about with whom they intervene in order to reach most effective outcomes: T1 as the more powerful and enforcing mediator is more likely to come to stable and effective outcomes. If this track is facilitated by unofficial mediators, its effectiveness is even higher, due to pooled resources, decreased uncertainty and the ensured support at the grassroots level (see Diamond and McDonald 1996; Gleditsch and Beardsley 2004: 382; Frazier and Dixon 2006: 404; Regan and Aydin 2006).

Against this background, this chapter opens up new avenues for further research on ToDs in international mediation. Although my results show that different tracks of diplomacy in general have positive impacts on the effectiveness of interventions, there may be specific dispute characteristics that make one track less effective than another. Issues like “why did the state actors’ attempts to reach a peaceful settlement in South Africa during Apartheid fail” (see Lieberfeld 2002) or “why were the efforts of the Community of Sant'Egidio so crucial for the effective conflict resolution in Mozambique” remain underexplored. It seems questionable to merely assume that ToDs are able to mediate all disputes, in all contexts (Bercovitch 1985). Future research has to unveil the interaction effects between conflict characteristics and ToDs. Finally,

101

although the results here indicate that the interaction of official and unofficial tracks does matter significantly, I have only used a rather crude operationali-sation that does not take into account how actual interactions between different tracks may be much more complex. As such, a more comprehensive theoretical model of varieties of track interaction may help shed further light on these specific issues. The final chapter addresses this point.

103

6. The Impact of Multi-Party Mediation in International Interventions: Synergy, Conflict, Effectiveness

As demonstrated in the previous chapter, many third-party attempts for settling disputes do not see one mediator, but a coalition of mediators instead that intervene simultaneously. The mediation effort of the Contact Group, i.e., the U.S., Russia, France, Britain and Germany, to settle the conflict in Bosnia in 1994 provides one example out of many. When studying the effectiveness of these mediations, however, the previous literature largely ignored how multiple mediators interact with each other and coordinate their intervention. So far, scholars rather focused on the relations both between the warring parties and to actors who either have mediated in the past or potentially will in the future (see e.g. Bercovitch, Anagnoson and Wille 1991; Bercovitch and Schneider 2000; Gent and Shannon 2007). Very little research tries to answer the question how “multi-party mediations” (Crocker, Hampson and Aall 1999: 9; see also Crocker, Hampson and Aall 2001a), i.e. those mediation attempts that involve multiple interveners influence conflict settlement outcomes in a systematic way.

Nonetheless, disaggregating intervening coalitions and “opening the black box” of those mediations seems an effort worth making, since combined mediation attempts are likely to be different from those interventions that only see one mediator. Mediators are driven by their own incentives, constrained by their own domestic factors and are tied to each other through various kinds of relationships. They share information and resources, and pursue other means of synchronising and/or integrating activities. The Contact Group, for example, used the combined leverage of resources, intelligence and logistics to enhance the effectiveness of its intervention. On the other hand, however, a coalition of interveners may increase the complexity of the mediation process, leading to unanticipated coordination and collective action problems, which ultimately can induce a serious deterioration of the overall conflict situation. The crisis in Burundi in the 1990s provides an example for this (Crocker, Hampson and Aall 2001b): the sheer number of intervening third parties, e.g. the UN, the EU or the Arusha Group of states, eventually undermined official claims that the international community itself wanted peace due to the not necessarily overlapping interests, motivations and agendas of each mediator. The UN Secretary-General Ban Ki-moon (2009) emphasized accordingly: “Multiple

T. Böhmelt, International Mediation Interaction, DOI 10.1007/978-3-531-92812-8_6,© VS Verlag für Sozialwissenschaften | Springer Fachmedien Wiesbaden GmbH 2011

104

actors competing for a mediation role create an opportunity for forum shopping as intermediaries are played off against one another. Such a fragmented international response reinforces fragmentation in the conflict and complicates resolution.”

In other words, there are numerous examples of conflict resolution, where a coalition of mediators either was effective or unsuccessful in settling a dispute. Although interventions with more than one mediator can make these attempts ultimately more synergetic, they also might create conflicts among the third parties that actually hamper effective dispute management. The crucial question that follows is: why is it so? As a matter of fact, we do not know. Comprehensive studies seeking to address this question both theoretically and in a rigorous empirical way are surprisingly rare. At the same time, the existing literature on multi-party mediation produced many contradictory claims about its effective-eness and overly derives its conclusions from single case studies. Beardsley and Greig (2009: 247) consequently emphasize that there is not yet “a clear under-standing of the degree to which the presence of multiple mediators improves or detracts from the conflict management process.”

As one of the first steps towards filling this gap, this chapter develops a new theoretical framework that analytically approaches the existing claims by drawing on the concepts of collective action and recent work on collaboration as well as democracy. More specifically, I first examine the size of an intervening group and argue that size and mediation effectiveness are characterised by an inverted U-shaped relationship. More mediators are generally able to create synergy due to combined efforts, making them more effective than a single third party (Crocker, Hampson and Aall 1999; 2001a; Greig and Diehl 2006: 363ff; Kydd 2006: 450f). On the other hand, a very large number of interveners implies higher organisational costs and greater heterogeneity of interests (Buchanan and Tullock 1962; Snidal 1994; see also Olson 1965). This increases the chances that coordination and collective action problems exist, which ultimately lower the margins of effective mediations. Second, I analyse multi-party mediation through conflicting relationships and cooperative interactions between the interveners. Mediators sharing a history of conflict and distrust will convey this relation into the mediation process, making these interveners less likely to be effective. Consequently, states having friendly and cooperative interactions with each other are likely to be more effective in managing conflicts peacefully due to higher communication flows, more trust and lower transaction costs (Axelrod and Keohane 1985; Crescenzi and Enterline 2001;Crescenzi, Enterline and Long 2008; Keohane 1984). Finally, I examine the impact of interveners’ regime types on mediation success. A coalition of mediators that is largely democratic should be more effective, since these mediators are characterised by a shared culture of

105

peaceful conflict resolution, inclusivity and increased communication flows (Dixon 1994; Raymond 1994; Bercovitch and Houston 2001; Mitchell 2002).

This chapter is likely to have important theoretical and empirical implications that go beyond the usual scope of work on international mediation. Little work has been done on the systematic study of multi-party mediation from an academic point of view and better understanding how combined mediation efforts influence the outcomes of conflict resolution attempts can also help policymakers to use third-party mediation more successfully. Obviously, knowing in advance that cooperation with some states might create synergy, while mediating together with other countries could worsen the situation significantly, is likely to make a crucial contribution to how decisionmakers perceive multi-party mediation in the future. Furthermore, since the chapter’s theoretical framework introduces new concepts to the study of mediation, it is also likely that the theoretical argumentation and the empirical findings might influence the rich literature on collective action and its related problems or research on actors’ collaboration and interactions. I will return to these topics in detail in the analysis and discussion sections of this chapter.

In the remainder of the chapter, I first review the relevant literature on multi-party mediation in conflict resolution attempts and highlight the many contra-dictory claims about its effectiveness. I then develop a theoretical framework in order to approach these claims analytically, drawing on the concepts of collective action and group size, collaboration via conflict and cooperative interactions as well as regime type. Afterwards, I detail the research design and, employing data from the Issues Correlates of War Project (ICOW) (Hensel and Mitchell 2007; see also Hensel 2001; Hensel, Mitchell, Sowers and Thyne 2008) in 1965-2000, empirically test the theory. The empirical analysis and the robustness checks largely provide support for the theory, but also show that both conflicting relations between the interveners and more democratic mediation coalitions do not seem to be crucially important for settling conflicts peacefully. The article finishes with a comprehensive discussion about its findings, the implications for policy makers and the avenues for future research. 6.1. Literature Review The following research disaggregates mediating coalitions by examining how third parties interact with and are tied to each other and analyses what kind of impact this might have on the outcome of mediation attempts. As stated above, when studying this process, I use the term multi-party mediation (Crocker, Hampson and Aall 1999; 2001a; see also Nan and Strimling 2006), which can be

106

defined as “information sharing, collaborative analysis and strategizing, resource sharing, formal partnerships and other means of synchronizing and/or integrating activities to make own individual efforts more effective together as interconnected pieces of a larger peace process” (Nan 2003). Building upon this definition, the previous research, though being sparse and mostly qualitative in its nature, has produced an interesting body of insights.

The most extensive studies on multi-party mediation are conducted by Crocker, Hampson and Aall (1999; 2001a; 2001b). The scholars approach multi-party mediation in a threefold sense. First, there are simultaneous interventions of multiple mediators that do not necessarily coordinate their efforts. Second, conflicts may see sequential mediated interventions that involve more than one third party. Finally, there are composite actors such as coalitions of states (Crocker, Hampson and Aall 1999: 210; Crocker, Hampson and Aall 2001a). In this study, I primarily focus on the third “composite” category where third parties actually form a coalition and seek to represent a single interest.26

Crocker, Hampson and Aall are aware of the potential of interventions conducted by more than one mediator. First, their structural theoretical argument suggests that multi-party mediation implies smaller shares of the fiscal burden and the political risk for each member of this coalition. These actors can further pool resources and skills and, thus, have in principle more leverage than a single intervener. Ultimately, this should make a multi-party mediation more effective than single-party interventions. For example, Argentina, Brazil, Chile and the U.S. formed a mediating coalition in 1995 to resolve the boundary dispute between Peru and Ecuador. Crocker, Hampson and Aall (2001b: 59) highlight in particular that peace was eventually established due to the combined leverage of these four powerful countries.

Second, multi-party mediation can create effective outcomes as “the lead mediator gains the benefit of the partners’ insights, relationships, credibility, resources, diplomatic ‘reach,’ and political ‘balance.’ Broadening, when successful, isolates the spoilers and ‘rejectionists’ common to most conflict situations” (Crocker, Hampson and Aall 1999: 230). The process that led to the Oslo Peace Accords, for example, was a successful multi-party mediation effort in which the Norwegian mediators could draw upon the insights of their American partners (Crocker, Hampson and Aall 2001b: 63). More specifically, the formal and public U.S.-led mediations were on the verge of failing as both

26 The other two categories are of minor theoretical importance and I leave their examination to future research. Still, the data I use for the succeeding empirical analysis comprises cases of these two categories as well and although I do not specifically address them in my research, controlling for these influences does not seem to have any substantial effect. Furthermore, in the following I only focus on states’ multi–party mediation due to the limited availability of data.

107

Palestinians and Israelis were unwilling to take any risks and – in the light of media attention – feared exposure. Accordingly, Norway – with the full backing of the U.S. – opened up confidential back channels that directly circumvented these constraints.

Finally, a prerequisite for these two conditions is, however, that the members of a mediating coalition largely share similar goals and are willing to work together (Crocker, Hampson and Aall 1999: 22). They further have to maintain coherence, coordinate and sequence their initiatives as well as have staying power over the process (Crocker, Hampson and Aall 1999: 33, 38f). Garb and Nan (2006) emphasize that these factors positively influenced the establishment of conflict resolution networks in the Georgian-Abkhaz peace intervention. However, their article derives its conclusions from a qualitative analysis that is based on a single case, which makes it difficult to know to what extent the findings also apply to other cases of multi-party mediation. Similarly, Strimling (2006: 94; see also Kelman 2006) argues that some level of coordination between mediators is a necessary requirement to achieve effective outcomes. She categorises mediators’ interaction along the clusters of communication, coordination, collaboration and integration. For example, higher forms of communication and coordination increase understanding and efficacy of resource allocation. Strimling’s analysis lacks empirical evidence, though, and she consequently highlights that “there is need for […] much more rigorous empirical research on the impacts of cooperation” (Strimling 2006: 92).

On the other hand, and as the title of their book “Herding Cats” implies, Crocker, Hampson and Aall (1999; see also 2001a) are rather sceptical of multi-party mediation attempts. The authors highlight that caveats of multi-party mediation may be higher complexity and costs, more difficulties in maximising leverage and potentially conflicting agendas (Crocker, Hampson and Aall 1999: 39f). Put differently, multi-party mediation can lead to serious coordination problems and a lack of transparency over the allocation of responsibilities: “Problems of handoff between one peacemaker and the next are all too frequently encountered when different mediators try to engage parties in negotiations over a prolonged period of time. […] During this period, misunderstandings and conflicting interpretations about implementation are common” (Crocker, Hampson and Aall 2001a: 507). The implementation of the Bicesse Accords in Angola between 1991 and 1993, for example, saw multiple mediators from the U.S., Portugal, Russia or the Organization of African Unity who were neither able to effectively coordinate their intervention nor agree on a consistent and coherent agenda for peace.

Summing up, the existent research on multi-party was able to identify crucial aspects of this complex phenomenon, but often merely describes multi-

108

party mediation processes while more systematic work is still rare. Many claims about the effectiveness of multi-party mediation even contradict each other and there is an overly strong conclusion from single case studies. I will address this shortcoming in the next sections by developing a theoretical framework of multi-party mediation and its impact on effectiveness. More specifically, I offer a new way to interpret the many opposing arguments on mediating coalitions by focussing on public choice approaches as well as recent work on collaboration and democracy. 6.2. Theory: The Impact of Multi-Party Mediation – Size, Conflict,

Cooperation and Regime Type 6.2.1. Size As stated above, one stream in the mediation literature generally argues that a coalition of mediators is more likely than a single intervener to establish a stable peace agreement (see Crocker, Hampson and Aall 1999: 22; 2001a; Greig and Diehl 2006: 33f; Kydd 2006: 450f). First, Touval and Zartman (1985: 40) emphasize that “third parties are only accepted as mediators if they are likely to produce an agreement or help the parties out of a predicament and for this they usually need leverage.” Due the combined leverage of a mediating coalition, multi-party mediation can then create obligations and expectations that help enforce compliance. A single mediator does often not have such an amount of leverage that may be necessary for effectively solving conflicts peacefully. Second, states as part of a mediating coalition have connected issues and interests. This establishes trust, which is a crucial element for overcoming collective action problems and being ultimately more effective in settling disputes. Finally, interacting third parties usually have a higher rate of communication, which facilitates coordination and decreases uncertainty (Frazier and Dixon 2006: 391ff; Kaye 2001: 64).

However, I contend that it is likely that there are decreasing returns with respect to higher levels of mediation effectiveness arising from higher leverage. In other words, multiple mediators are only to a certain extent more effective than a single third-party. Mediating parties vary in terms of their leverage and the larger a coalition of mediating states, the higher the likelihood of unanticipated problems of coordination, which ultimately may cause that a sufficiently large coalition of mediators does not contribute significantly more to effective outcomes than a smaller set of interveners (Crocker, Hampson and Aall 1999: 40).

109

Related to, albeit differently from the last claim is the public choice arguing that larger groups providing a public good – in this case global or regional peace – are more difficult and more costly to organise (Buchanan and Tullock 1962: 46ff; Chamberlain 1974). Organisational costs increase with group size in two different ways. First, the larger a group of states trying to settle a conflict peacefully, the more likely it is that the intervening coalition is characterised by greater heterogeneity of interests (Snidal 1994). This in turn makes it more difficult, both for the mediators and the belligerents, to reach any agreement at all or something beyond a minimal consensus. Second, states can participate in mediation efforts without contributing much or anything at all, but obtaining the reputation as a peaceful broker afterwards. Beardsley (2009; 2010) argues in a similar way when suggesting a collective action problem in which the presence of multiple third parties with the potential to offer mediation reduces the likelihood that any mediation will take place at all. Under these circumstances, enforcement through monitoring and/or sanctioning is essential in order to prevent states from free-riding on other states’ mediation efforts (Fearon 1998: 270; see also Crocker, Hampson and Aall 2001b). Monitoring procedures ensure that contributions become perceptible for the intervening coalition. In addition, only the prospect to detect non-contributions to an effort of international mediation allows states to sanction defectors accordingly (Axelrod and Keohane 1985: 235; North 1990: 57; Olson 1965: 45f). However, with an increasing number of mediators, the more costly it is to organise monitoring devices that provide perceptibility of individual contributions and enforcing actors’ com-pliance becomes less likely to succeed (Axelrod and Keohane 1985: 234ff; Frohlich and Oppenheimer 1970; North 1990: 57; Olson 1965: 36).

To recap, my argument on the impact of coalition size on mediation effectiveness centres on a mediating coalition’s leverage and organisational costs. While I contend that the impact diminishes for leverage, I claim that it strengthens for organisational costs. In other words, the relationship between coalition size and mediation effectiveness should be characterised by an inverted U-shaped relationship (see also Buchanan and Tullock 1962). Figure 6.1 graphically depicts this argument and I therefore postulate:

Hypothesis 6.1. The size of an intervening group and mediation effectiveness are characterised by an inverted U-shaped relationship: both small and very large groups are less likely than medium-sized intervening coalitions to mediate disputes effectively.

110

Mediation Effectiveness

0 Coalition Size Figure 6.1. Expected Impact of Coalition Size on Mediation Effectiveness.

6.2.2. Conflicting and Cooperative Interactions The overview of the literature showed that previous work on multi-party mediation effectiveness emphasized the maintenance of coherence and coordination between the coalition members. However, this previous research did not unpack this mechanism. In order to address this in my theoretical frame-work, I focus on the interactive ties between mediators in a coalition. In more detail, states in the international system are tied to each other in multiple forms. Countries may have good and friendly relations with each other for decades, but they could also share a relationship of common distrust, hostility and war instead. In turn, this implies that states’ interactions in the past may increase our understanding of countries’ future behaviour and its impact on mediation effectiveness (Crescenzi and Enterline 2001).

Before examining the likely impact of mediators’ conflicting or cooperative interactions on mediation effectiveness, however, we first need to define the actual concept in question. Crescenzi and Enterline (2001) elaborate the underlying causal mechanisms of interstate interactions along four dimensions (Goertz and Diehl 1993: 195f). First, according to “accumulation,” if states interact frequently, then the relationship is well defined, but further interaction also has a decreasing marginal effect. Second, the duration between specific

111

interaction events matters. The longer the time elapsed until a new interaction between states occurs, the less related they are and the more likely it is that these events will be independent from each other. Third, the intensity or degree of interstate interactions determines the impact of these relations. For example, a simple trade dispute is likely to have a significantly smaller impact on future interactions than a previous full-scale war. Finally, the intensity of states’ inter-action may change over time. The last interactive event between states does not disappear immediately, but gradually diminishes with duration. As a result, states’ interactions last for a certain period, albeit with a different intensity.

Defining mediators’ interactions along those four dimensions demonstrates that interaction does not occur in a historical vacuum, varies over time and may well have different effects on states’ behaviour depending on its intensity. Against this background, I can formulate the argument on states’ conflicting or cooperative interactions and their impact on mediation effectiveness quite simply now. If states negotiate with each other or, as in the focus of this chapter, a coalition of countries tries to mediate conflicts, their decisions and the likelihood of certain outcomes are “conditioned to some extent by the prior relationship” (Crescenzi and Enterline 2001: 410). Mediators sharing a history of conflict and distrust will convey this relation into the mediation process. If states in multi-party mediations are connected via “negative” ties, it is rather unlikely that they are able to establish a culture of coordination, but face communication and co-ordination difficulties instead (Garb and Nan 2006). This makes these mediators less likely to be effective. Coming back to the Contact Group, for example, its initial mediation efforts have seen severe set-backs due to the aloofness of the Americans towards the Russians and vice versa. This argument enables me to formulate the next hypothesis:

Hypothesis 6.2a. The more conflicting interactions between states in a

mediating coalition, the less effective the mediation attempt.

Consequently and on the other hand, a coalition of countries sharing friendly and cooperative ties with each other is likely to be more effective in managing conflicts peacefully. Those “positive” ties between interveners should enhance communication flows and resource sharing as well as facilitate a common understanding of how to mediate effectively (Kelman 2006). Furthermore, states that previously shared cooperative relationships with each other are likely to have encountered similar negotiation situations as mediation bargaining processes before. They are therefore familiar with the counterpart’s intentions, preferences and incentives. This ultimately helps saving transaction costs between the mediators, increases the cooperative efforts and may well induce

112

synergy – despite possible collective action problems as outlined above (Keohane 1984; Axelrod and Keohane 1985). My hypothesis on cooperative interactions in multi-party mediation efforts argues accordingly:

Hypothesis 6.2b. The more cooperative interactions between states in a mediating coalition, the more effective the mediation attempt.

6.2.3. Regime Type The final core variable of the theoretical framework pertains to interveners’ regime type in a mediating coalition. Drawing upon the democratic peace literature (e.g. Oneal and Russett 1999), democracies have a common structural background and a common political culture, which gives these kinds of regimes more inclusivity, a higher level of transparency and it promotes a common understanding that induces shared norms. One of the most important norms of democracies is the peaceful resolution of conflicts (e.g. Dixon 1994; Raymond 1994; Bercovitch and Houston 2001; Mitchell 2002).27 This should make democratic belligerents not only more likely to settle their disputes peacefully, but also have an impact on the efforts of democratic mediators. For example, Dixon contends that democracies will be more successful in dispute settlement, since they operate under a norm of bounded competition that favours the use of compromise and non-violence (Dixon 1994). When intervening, democracies seek to follow these principles. Similarly, democratic interveners all share, practice and understand concepts such as limited governments, respect for judicial practices as well as constitutional constraints (Simmons 1999). Due to their previous, long-lasting practice and experience with these issues, democracies will be more likely to settle disputes effectively than (largely) autocratic coalitions, which are not familiar with them.

In addition, the inclusivity and transparency of democratic systems allows enhanced communication flows between these sorts of regimes, which facilitates the resolution of collective action problems in a mediating coalition and, ultimately, makes the mediators more effective in cooperating with each other. As indicated above, mediators having more information about each other will ultimately face less uncertainty and have more trust as well as higher reliability not only towards their fellow mediators, but also signal this to the antagonists (Mitchell, Kadera and Crescenzi 2005: 8). This is crucial against the background

27 Note that autocracies also share a common structural background, although they certainly lack the thereby induced shared norm of peaceful conflict settlement.

113

that effective conflict resolution depends on the “credible and truthful transmission of information” (Dorussen and Ward 2008a: 193) – especially to-wards the belligerents as they primarily demand mediation from states they can trust (Crescenzi et al. 2008: 13). On the opposite, promises of autocracies are less likely to be credible, since there is an “insecure hold of power and the absence of an heir” (Olson 1993: 571). Finally, increased flows of information allow democracies to coordinate their efforts and resources more effectively (Choi 2003: 145). Unlike autocracies, democracies tend to have transparent, open systems, which increase the likelihood that those regimes know about each others’ “internal evaluations, their intentions, the intensity of their preferences, and their willingness to adhere to an agreement even in adverse future circumstances” (Keohane 1983: 162).

In short, mediating coalitions that mainly comprise democratic states are characterised by high levels of communication, coordination, collaboration and integration. These are precisely the four factors that Strimling (2006) highlights for multi-party interventions to be successful. Above all, however, (largely) democratic coalitions should all share, practice and understand the norms of peaceful conflict management. My last hypothesis therefore claims:

Hypothesis 6.3. The more democratic a coalition of mediating states, the

more effective is the mediation attempt. 6.3. Research Design 6.3.1. Data Against the background of my original definition of mediation, I consider all kinds of voluntary, non-coercive third-party interventions of states such as facilitation, good offices or conciliation as mediation attempts. Cases that involve more than one third party via a coalition formation are treated as multi-party mediation attempts and I drop pure military interventions or bilateral conflict managements.28

Unlike in the previous chapters, however, I employ data from the Issues Correlates of War (ICOW) project (Hensel and Mitchell 2007; see also Hensel

28 Note, however, that this approach only excludes direct or “pure” military interventions. If con–flicts comprise both military interventions and third–party mediations, for example, I do consider these settlement attempts in my analysis.

114

2001; Hensel et al. 2008).29 The ICOW data cover territorial claims in the Western Hemisphere and Western Europe, river claims in the previous two regions plus the Middle East and maritime claims in the Western Hemisphere and Western Europe. Claims are identified according to explicit evidence of contention involving official representatives of two or more nation states over the issue type in question. My unit of analysis is a contentious issue claim with a third-party mediation attempt, while the temporal domain of my study is from 1965 to 2000 due to data constraints. To illustrate this with an example, the ICOW data list 16 mediation attempts – either single-party mediations or multi-party mediation attempts – for the Falklands War between Argentina and the UK. Thus, I consider all 16 settlement attempts for my analysis, while each mediation is a separate observation. In total, I obtain 151 contentious issues with settlement efforts over 55 territorial issues, 22 river claims and 74 dyadic maritime claims in the period under study.

The data are essentially cross-sectional. However, some temporal dependence may persist as interventions or settlement attempts are unlikely to be independent from previous conflict management efforts over a claim between a specific dyad. To address this problem, I cluster the standard errors on claims per dyad and consider a duration variable that is specified below. 6.3.2. Dependent Variable The dependent variable measures the effectiveness of third-party settlement attempts along five categories. First, the third-party mediation may be completely ineffective and there is no agreement among the belligerents. Second, the antagonists may be able to reach an agreement with the help of third parties, but at least one of the disputants does not ratify the treaty. Third, although there is an agreement that addresses the issues that led to the dispute in the first place, at least one of the belligerents does not comply with it. Fourth, although all of the warring parties comply with the reached agreement, this still may not end the claim. Finally, if all conflict states comply with an agreement and the claim is ended, then the effectiveness of a third-party intervention is considered to be at its maximum. Since the dependent variable follows an ordinal scale, I use ordered probit regression models to test my hypotheses. Table 6.1 gives an overview about Effectiveness.

29 The advantages of the ICOW data are twofold. First, they allow the comparison across various issues of contention. Second, since the data’s cases are not constrained to situations that become militarised at some point, we are able to circumvent the problem of selection bias to some extent.

115

Table 6.1. The Effectiveness of Third-Party Settlement Attempts, 1965-2000.

Frequency Percent No Agreement 73 48.34 Agreement 11 7.28 Ratification 5 3.31 Compliance 46 30.46 End of Claim 16 10.60 Total 151 100.00

6.3.3. Explanatory Variables The core variables of interest follow the lines of my theoretical framework. First, I consider a binary variable that indicates whether a conflict saw multi-party mediation or not. This not only gives me leverage for comparing single and multi-party conflict management interventions, but also tests the literature’s naïve, non-disaggregated claim on multi-party mediation effectiveness. In my sample, 52 settlement attempts (34.44%) saw multi-party mediation, while 99 cases (65.56%) were mediated by a single third party. Following my reasoning, I then disaggregate this variable to create an item counting the number of third parties that were simultaneously involved in an intervention. This variable ranges in the interval [1; 6]. I include its squared term in some models for testing my hypothesis on the inverted U-shaped relationship between coalition size and mediation effectiveness.

Second, in order to measure conflicting and cooperative interactions between states in third-party interventions, I rely on Crescenzi and Enterline’s dynamic model of interstate interaction (Crescenzi and Enterline 2001; see also Crescenzi, Enterline and Long 2008).30 The scholars capture the four dimensions of interstate interaction, i.e., accumulation, temporal distance, degree and the rate of change, with two equations. First, we obtain a growth function:

30 Note that there are also alternative operationalisations for conflicting and cooperative relation–ships. Previous studies, for example, rely on the similarity of alliance portfolios as measured by Kendall’s b (Bueno de Mesquita 1978; 1981) or the S–score (Signorino and Ritter 1999). With regard to the former, however, I believe that this item it theoretically too far from my claimed me–chanism, since b essentially measures states’ associations via alliances. With regard to the latter, Bennett and Rupert (2003) demonstrate that the S–score suffers from measurement errors and is arti–ficially skewed towards higher dyadic similarity. In other words, although Crescenzi and Enterline’s (2001) dynamic model of interstate interaction is the primary reason for my constrained temporal do–main, the potential bias and measurement error are likely to be lower.

116

t

tttg DistTemporal

Degreeii 1 (6.1)

where itg is the interaction level on the growth dimension – either cooperative or conflicting – for a particular dyad in a specific time unit t, it-1 signifies the dyad’s interaction level in the previous period and ß represents some pre-defined weight. Degree measures a dyad’s current extent of interaction, while Temporal Distt is the time elapsed since the previous interaction occurred (Crescenzi and Enterline 2001).

On the other hand, interstate interaction is captured by a decay function:

1

)(

)( tHistoryEvent

DistTemporalEvent

td iei t

t

(6.2) where itd is the interaction on the decay dimension, Event History is the accumulation of occurrences of either conflict or cooperation between two states up to time t, Event Temporal Dist signifies the time that has passed since the last event and weighs the relative impact of the two factors.31 Combining the growth with the decay function results in the overall level of interaction between a dyad:

it=itd+itg (6.3)

With these equations, I construct variables for both conflicting and cooperative interactions between states in mediating coalitions. The variable that measures conflict relations uses militarised disputes as the negative shocks to a dyadic relationship (Crescenzi and Enterline 2001). It is theoretically bounded in [-1; 0] and lagged one year to account for temporal dependency. In my sample, this variable varies from -0.24 to 0.00. The item for cooperative interactions among dyads relies on the mutual membership in intergovernmental organizations (IGOs) (Pevehouse, Nordstrom and Warnke 2004; Crescenzi, Enterline and Long 2008). The joint decision of a dyad to join the same IGO(s) in a given year operationalises cooperative shocks. This variable is weighed by the number of other nations currently in the IGO, lagged by one year as well and I employ a centred version of this item due to its high pairwise correlation with Size. The final item, Cooperation, ranges from -0.76 to 0.35 in my sample. Since 31 Crescenzi, Enterline and Long (2001) emphasise that the form of this function constantly drives the value of itd towards zero over time; also, the exponential decay is accelerated by increases in Event Temporal Dist, but decelerated by increases in Event History.

117

both the variable on conflicting relations and the item on cooperative interactions measure in essence interstate interactions at the dyadic level, I operationalise the mean value for all possible dyads in a mediating coalition as the final variable. Mediation efforts that only saw one third party, obtain the highest possible value for both conflicting interactions and cooperative relationships.

Third, the overall level of democracy of a mediating coalition is measured by the average value of all intervening countries’ polity2 value from the Polity IV data (Marshall and Jaggers 2002), ranging from -10 (full autocracy) to +10 (full democracy).

Although I am primarily interested in the impact of multi-party mediation, I have to control for variables along the other two conceptual clusters that influence the effectiveness of intervention attempts, i.e. variables pertaining to the warring parties as well as the dispute context and its characteristics (Bercovitch and Langley 1993; Kleiboer 1996), in order to avoid omitting variables and thereby possibly biased results. Further, including other deter-minants of mediation effectiveness controls for possible selection effects as well, since it may be that only certain types of conflicts actually see multi-party mediation. Hence, I consider the following variables: first, I include the belligerents’ regime types in order to address one of the most crucial aspects of the antagonists. As elaborated above, democracies have common structural backgrounds and share the norm of peaceful conflict resolution, which will also increase the chances that fighting parties settle their disputes more effectively (e.g. Dixon 1994; Raymond 1994; Bercovitch and Houston 2001; Mitchell 2002). The data are taken again from Polity IV (Marshall and Jaggers 2002), using a weakest link specification where the least democratic country in a dyad determines the effectiveness of an outcome.

A second control captures another important characteristic of the fighting parties. Belligerents that approximate military parity, i.e. neither side has a clear military advantage are less likely to settle disputed issues effectively (Bennett and Stam 1996; Mason and Fett 1996: 550). Power discrepancies decrease the incentives for the most powerful country to agree to a peaceful settlement as it stands a better chance of winning the conflict militarily. Using data from the Correlates of War Project (Singer 1988), I include the natural log of the ratio of the stronger country’s capability to that of the weaker country (Singer, Brember and Stuckey 1972).

Third, the intensity or the importance of an issue claim is operationalised by Salience, which measures the characteristics of the issue claim. As Hensel (2001; see also Hensel et al. 2008) demonstrates, highly salient disputes are more difficult to resolve and are more likely to lead to a militarised dispute (Bercovitch and Gartner 2006; Quinn et al. 2006: 454). Following Hensel and

118

Mitchell (2007), the sample’s variable ranges in [2; 12], with higher values indicating greater salience. Finally, I control both for the context of disputes and existing temporal dependence by including the duration of an issue claim. Settlement attempts are less likely to be effective in long-standing conflicts that have “witnessed a long history of mutual attack and atrocity” (Bercovitch and Langley 1993: 362f). Hence, the longer the duration of a conflict, the less likely that it sees an effective settlement (Touval and Zartman 1985). 6.4. Empirical Findings Table 6.2 displays the results of my analysis. I employ four models. First, the focus is on the effect of multi-party mediation per se and the remaining core explanatory variables. Second, I include the controls in the right-hand side of the model. Finally, Models 3 and 4 follow the same procedure as Models 1 and 2, respectively, but here I replace the naïve approach using a binary item with variables on the size of an intervening coalition. To unveil the implied magnitude of my substantial findings, Table 6.3 reports the average predicted probability of the change across categories of Effectiveness as an explanatory variable changes from its minimum to its maximum, while holding all other variables at their means.

Before I will come to my core explanatory variables, I briefly discuss the results of the control covariates. In consistence with the previous literature that emphasized the importance of including those variables (Bercovitch and Gartner 2006; Kleiboer 1996), most of these items matter for mediation effectiveness. The signs of the belligerents’ level of democracy and the capability item have the expected direction, and are statistically significant at the 10% level at least. Furthermore, Issue Salience has the expected negative sign and is significant at the 5% level. When moving towards the maximum of Issue Salience, for example, the probability to reach the most effective settlement decreases by about 16.35% on average. However, the coefficient of Duration is approximately zero. The impact of this variable on the effectiveness of interventions is therefore only minor at best.

119

Table 6.2. The Effectiveness of Multi-Party Mediation, 1965-2000.

Model 1 Model 2 Model 3 Model 4 Multi-Party Mediation 0.285 0.704 (0.276) (0.320)** Size 0.896 1.630 (0.404)** (0.709)** Size2 -0.136 -0.287 (0.057)** (0.127)** Conflict -3.375 -3.721 -2.192 -4.158 (3.030) (3.144) (2.943) (3.326) Cooperation 1.443 2.154 1.698 2.318 (0.532)*** (0.650)*** (0.498)*** (0.645)*** Democracy 0.015 0.014 0.019 0.014 (0.021) (0.021) (0.022) (0.021) Belligerents’ Democracy 0.027 0.027 (0.014)* (0.015)* Capability Ratio -0.238 -0.244 (0.078)*** (0.082)*** Issue Salience -0.107 -0.107 (0.043)** (0.043)** Duration -0.000 -0.000 (0.000) (0.000) N 147 130 147 130 Log Pseudo Likelihood -178.960 -147.213 -176.958 -146.723 Wald 10.30 21.89 24.20 23.55 Prob> 0.04 0.01 0.00 0.01

Note: cut points not reported; clustered robust standard errors in parentheses; * significant at 10%; ** significant at 5%; *** significant at 1% (two-tailed).

Turning to my variables of main interest, I largely find support for this

chapter’s hypotheses. In particular, Multi-Party Mediation and Cooperation exert a positive influence on mediation effectiveness. Both variables are statistic-ally significant with the exception of Model 1, where Multi-Party Mediation does not even approach the 10% level. Those two variables also have a substantively important influence on the dependent variable. Moving from the minimum to the maximum value of Multi-Party Mediation changes the probability to obtain a specific category of Effectiveness by about 7.75% on average. Even more substantively strong, if interveners share good and friendly relations with each other, they tend to convey this culture of cooperation into the

120

mediation process. The probability of effective conflict resolution is raised by about 26.65% on average, if Cooperation changes from its lowest to its highest value. Hence, positive interactions among mediators, as measured by the joint decision of a dyad to join the same IGO(s) in a given year, are conveyed into multi-party mediations, making these efforts more effective.

If countries had hostile and conflicting relations with each other before intervening in a dispute, then the conflict management outcome should be lower. Conflict actually has a highly negative impact. On average, the chances to see the most effective settlement of a dispute decrease by about 12.48%. This substantive effect should be interpreted cautiously, however, due to the insignificance of this variable in all of my models. Comparing my findings for Conflict and Cooperation basically leads to the rather surprising conclusion that countries value cooperative relationships more than conflicting interactions. In other words, if states had negative encounters in the past, it is apparently less likely that these experiences are seen as an obstacle for effective multi-party mediation. On the contrary, my results show that these conflicting relations are unlikely to matter at all. Note, however, that the insignificance of Conflict may be the result of selection effects. First, countries sharing (extremely) hostile relations with each other may not agree on mediating a dispute simultaneously in the first place. Second, when belligerents ask for mediation, they seek to address states that could favour their interests (Crocker, Hampson and Aall 2001a). In other words, the antagonists seek to avoid a mediating coalition (or a single mediator) that is biased towards the dyadic counterpart as this would increase the chances that the mediation outcome might not favour them, but their opponent. Hence, fighting parties intend to establish a balance between mediators: the larger the bias of one mediator to a belligerent, the higher the likelihood that the opponent will ask for an additional state that represents its interest (Gent and Shannon 2007).32 For example, over the course of the conflict between Israel and the Palestinian Liberation Organization (PLO), the former frequently asked the U.S. to mediate, while the PLO feared the U.S. to merely mirror Israeli interests and, hence, also demanded that the USSR should mediate. As a result, the U.S. and the USSR formed mediating coalitions on various occasions such as during the 1991 peace conference in Madrid – despite their fairly uncooperative

32 In this context, Kydd (2003) argues that a mediator must be biased toward a belligerent if it is to communicate with it honestly. In a subsequent paper, Kydd (2006) discovers that if a mediator is to help resolve a conflict caused by mutual mistrust, it must be unbiased. On the other hand, however, Savun’s results (2008) suggest that biased mediators and interveners with relevant information yield more effective mediation outcomes. Due to the ambiguity of both theory and empirical evidence and since my explanation is of rather ad–hoc nature, I leave this issue and a more thorough investigation of it to further research.

121

interactions. However, the influence of these multi-party mediations was low as there was basically no interference from the U.S. and USSR sponsors, while the belligerents limited themselves to launch bilateral negotiations in the future (Choukri-Fishere 2008). Furthermore, recall that I argued that mediators sharing a history of conflict and distrust will convey this relation into the mediation process. Gent (2007) shows in this context that states with conflicting ties or diverging preferences are more likely to form a coalition in the case of military interventions. Under those circumstances, conflicting ties may well cancel each other out, thereby leading to the overall insignificance of this item.

Note: graph displays predicted probabilities for effective mediation attempts based on Models 3 (left) and 4 (right). All variables apart from Size and Size2 are held at their means.

Figure 6.2. Inverted U-Shaped Relationship between Coalition Size and

Mediation Effectiveness.

122

Most interestingly for my study, the results confirm the hypothesis of a non-linear relationship between the size of an intervening group and mediation effectiveness, thereby rejecting the naïve claim that more mediators are always more effective. Figure 6.2 illustrates this relationship graphically. While the x-axis signifies the values of Size in Models 3 and 4, respectively, the predicted probabilities of seeing either value 3 or 4 on Effectiveness are shown at the y-axis. Table 6.2 shows furthermore, for example, that the turning point associated with the size of an intervening group lies at around 3.29 here (Model 4: 2.84). In other words, both very small and very large groups of interveners perform worst in settling disputes peacefully. One intervener on its own or two third parties that mediate may not have the necessary amount of leverage to create incentives for the disputants. On the other hand, although very large groups of mediators may have the necessary leverage, they are more difficult to organise and have a greater heterogeneity of interests, making them ultimately also less effective. This finding is robust over a wide variety of model specifications. Adding or suppressing controls from the model does not alter this result.

Finally, I hypothesised that a higher level of democracy on the intervening side would increase the chances that disputes are mediated effectively. Although the variable is positively signed, my findings are inconsistent with that logic. Increasing the interveners’ overall democracy score from its minimum to its highest value slightly raises the likelihood of effectively mediated outcomes by 4.98% on average, but the variable is insignificant in each model. This finding clearly shows that we have to examine the supply side of multi-mediation more carefully, since there seems to be a lot more that has to be uncovered (Beardsley 2010). Two potential explanations appear plausible for this inconclusive finding, however. First, as argued above, democracies share norms of peaceful conflict resolution and have a common structural background. They are also tied to each other through higher flows of information. However, effective conflict resolution might rather depend on the mediators’ interests, their leverage capabilities, or mediation techniques (Princen 1992; Young 1972). And these factors are largely independent from the (aggregate) regime type of the intervener(s). Garriga (2009) argues in a similar way when stating that democracies are actually more constrained in their foreign policy due to domestic checks and balances. Eventually, this may cancel out the postulated positive impact of democracies’ common institutional backgrounds, preference similarities and enhanced information flows, ultimately leading to the observed insignificance of Democracy. Second, the provision of effective conflict settlement may entail collective action problems – especially for intervening democratic coalitions (Beardsley 2009; 2010; Crocker, Hampson and Aall 2001b: 58; Gent 2007). Although democracies may have an interest in peace and peaceful conflict

123

resolution, the “bonds of friendship” between democracies can cause that these kinds of states contribute less than their “fair share,” i.e. there might be a democratic tendency to free-ride in multi-party mediation efforts (Desch 2002; Olson and Zeckhauser 1970). In other words, the more democracies as third-party interveners in disputes, the more likely that these states will rely on their “fraternal obligation” and expect that other democracies will provide more leverage and resources for settling a conflict. This forces other interveners, including those that may be less democratic, to compensate for the lost resources and we conclude that the level of democracy in multi-party mediations does not play a major role for effective outcomes.

Table 6.3. Predicted Probabilities of Mediation Effectiveness, 1965-2000.

Model 1 Model 2 Model 3 Model 4 Multi-Party Mediation 4.5% 11.0% Size 39.9% 39.6% Size2 -28.9% -28.7% Conflict -12.6% -13.8% -8.3% -15.2% Cooperation 22.4% 28.9% 25.2% 30.1% Democracy 4.8% 4.5% 6.0% 4.6% Belligerents’ Democracy 8.6% 8.4% Capability Ratio -19.2% -19.6% Issue Salience -16.3% -16.4% Duration -0.8% -1.1%

Note: change of predicted probabilities of Effectiveness is shown for each independent variable when moving from the minimum to the maximum; all other variables held at their mean values.

124

6.5. Robustness In order to ensure the robustness of my findings, I changed a variety of model specifications and re-ran the estimates again. First, the dependent variable’s “ratification” category has only five observations, meaning that this value could be an outlier category that overestimates my results. In order to address this, I recoded Effectiveness by grouping the “agreement” and “ratification” categories together. Similarly, given my rather small sample of events, I also considered all models using a less stratified dependent variable by employing a dichotomous item for Effectiveness with the value of “0” matching categories 0-2 and the value of “1” matching categories 3 and 4 of the original variable. Both changes did not alter the substance of my findings.

Second, I stated above that mediation efforts and their composition are unlikely to be independent from previous attempts in a dispute. To further address this issue, I ran models with a simple count item on the number of previous mediation efforts in a conflict. Additionally, I also created a variable indicating whether it is a particular mediating group's first, second, third, etc. attempt in a specific dispute. Especially this latter approach accounts for “learning” and reciprocity over the course of mediations (Greig and Regan 2008). Although the “simple count item” is not significant in any model, the more informative “learning” variable is significant at the 1% level and negative. Thus, subsequent mediation attempts of the same mediator or intervening coalition are less effective, which may reflect more difficult conflicts or dispute issues in question. My core items stay the same, however.

Third, I then considered a couple of other variables that may influence both mediation effectiveness and my explanatory variables although I did not incorporate these items in the original design of this chapter. On one hand, I included a covariate for belligerents’ alliances in order to better capture the relationship between antagonists and a dummy item indicating whether a dispute broke out due to a territorial claim since these ones are likely to be the most intense conflicts (Hensel 2001; Hensel et al. 2008). However, both variables are highly insignificant in any model specification while the other findings stay the same. On the other hand, it could well be that the style of a mediation attempt and not the size of a coalition matter more for the prospects for peaceful settlement. Dixon (1996), for example, finds that more enforcing strategies seem to induce more effective outcomes (see also Beardsley at al. 2006). Against this background, the ICOW data include dichotomous variables on functional techniques (which attempt to address the use of the claimed territory but do not address sovereignty questions), procedural attempts (which address future efforts to settle a claim but do not address sovereignty directly) and substantive attempts

125

(which address sovereignty over part or all of a claimed territory). I therefore considered these variables in my estimations, but the results stay the same and there is no substantive difference between the different mediation strategies as such, i.e. all mediation strategy items are statistically insignificant while using functional attempts as a baseline category.

Fourth, one could also argue that it is not only the size of a mediating coalition but the type of the mediators, i.e. major powers versus small powers that influence the effectiveness of a third-party settlement attempt. For example, Favretto (2009) claims that major powers can rely on superior intelligence-gathering resources, a higher military effectiveness and are better equipped for intervening in the most difficult types of disputes. In sum, this enables more powerful states to solve issues of incomplete information, commitment problems and conflicts of indivisibility more effectively than less powerful actors, which, ultimately, should translate into higher mediation outcomes. Following the operationalisation of Democracy, I therefore created a variable for measuring the capabilities of a single mediator or a coalition by the average value of all intervening countries’ CINC score from the Correlates of War data (Singer, Bremer and Stuckey 1972). Including this variable in any model reveals that mediators’ capabilities have a negative, but insignificant impact on mediation effectiveness. In other words, size plays a more important role than pure power in the context of multi-party mediation, which again points to a disconnect between single-party mediations and cases of intervening coalitions.

Fifth, my theory indicates that there may be problems of selection bias. The decision to mediate – either unilaterally or as part of a coalition – is itself a strategic consideration. As demonstrated in Chapter 5, there are many underlying factors that are likely to influence both the decision to mediate and mediation effectiveness. Although I tried to mitigate the consequences of selection bias with multiple strategies in the research design, the empirical results indicate that these problems may persist for some covariates. In order deal with this issue in more depth, I ran alternative estimations using a Heckman selection model (Heckman 1979). For these calculations, the specifications are as follows. First, I employed the entire ICOW data, i.e. I did not drop cases without third-party mediation and, hence, used a contentious issue that may or may not see settlement efforts as my unit of analysis here. Second, the onset of mediation signified the dependent variable in the selection equation and I used Effectiveness as the dependent item for the outcome stage. Finally, the explanatory variables for the first stage comprised the control variables from above, which also represent usual factors that explain mediation onset, i.e. the intensity of a claim dispute, belligerents’ regime type and their capability scores

126

as well as variables for modelling time dependencies.33 I did not change the model specifications from my core estimations above in order to model the second stage. The results from these calculations, nevertheless, showed that it is unlikely here that unmeasured influences in the first stage associated with unmeasured influences in the outcome equation bias my parameter estimates. The parameter is statistically insignificant when employing the specifications from Model 4 above (Prob> 2=0.142). Furthermore, my core results stay the same, i.e. I still obtain evidence for an inverted U-shaped relationship between Size and Effectiveness, while Cooperation is positively signed and significant. Conflict and Democracy remain insignificant. 6.6. Discussion This chapter has sought to increase our understanding of the impact of multi-party mediation in third-party conflict resolution. In one of the first quantitative studies to examine this phenomenon, I disaggregated the coalition of intervening states and developed a theoretical framework that analytically examines the outcome of peaceful settlement attempts according to some of the mediators’ unique characteristics and interactions with each other.

Using data on states’ claims on various issues in 1965-2000, my findings are largely in line with the theory. In general, multi-party mediation may not necessarily be more effective than single third-party mediations in bringing conflicts to an end. Instead, I found an inverted U-shaped relationship between the size of the intervening group and effective outcomes. As elaborated above, the optimal size of a group trying to settle a dispute peacefully seems to be around three.34 As expected and in consistence with conventional wisdom are my findings on conflicting and cooperative interactions among mediators. Put simply, countries sharing good relations with each other are more likely to intervene effectively, while hostile states are less likely to create synergetic effects. More precisely, the latter may actually decrease the chances that a conflict will be brought to an end by peaceful means, but my results lack statistical significance. Finally, despite previous research claiming that a coalition of democratic intervening states may be more effective in settling issue

33 I replaced the duration item by a mediation–years variable and different sets of cubic splines (see Beck, Katz and Tucker 1998). 34 A caveat is here that six countries comprise the maximum size of an intervening coalition in the ICOW data.

127

claims, my findings show that the overall regime type of an intervening coalition rather is of minor importance.

The policy implications of this chapter are both crucial and straightforward. First, if belligerents seek to settle a dispute by peaceful means, then they should demand mediation from a coalition of states up to a certain size only. If antagonists ignore the size of a mediating coalition, then less effective outcomes automatically become more likely, hence questioning the demand of mediation in the first place. Second, when allowing a set of states to intervene, countries should be aware of these mediators’ previous relationships with each other – in particular with regards to their cooperative interactions. Otherwise, conflicting instead of synergetic negotiations may occur. Similarly, decision makers willing to mediate conflicts must be aware of whom they intervene with. My results found evidence that countries with less cooperative attitudes towards an intervener may not suit as partners in a mediating coalition. Finally, the regime type of intervening states appears to be a factor that is of little importance. Although I do not intend to reject that (some) democratic states may be helpful in a group of mediators, my research indicates that the chances for effective conflict resolution are not necessarily driven by regime type.

Hence, my empirical analysis demonstrates a great deal of empirical support for the theory and certainly increases our understanding of multi-party mediation, but many important questions remain. For example, the variables I employed in order to examine mediators’ interactions are largely of an ex-ante nature. The question that logically follows is if there are other specific variables pertaining to the interactions during a dispute that determine the relationships between interveners and ultimately the effectiveness of their effort. This issue rather is of rhetorical nature and should encourage future research to collect data and test respective claims empirically. Also, although my analysis indicates that coalitions’ regime types tend to be of minor importance in dispute settlement, this finding runs against my theoretical expectation and multiple other studies. Future research might want to address this and find evidence for or against the alternative explanations I provided above. Finally and related to the last avenue for further analyses, there is the issue of selection bias. Despite my efforts, doubts may persist that I was fully able to mitigate its consequences. I agree that one prospective way for more effectively addressing this problem requires us to go one step back by answering why multi-party mediation as compared to single third-party interventions occurs in the first place. On one hand, a supply side explanation could suggest that mediators benefit from shaping conflict outcomes, and hence more than one third party will try to be involved in a dispute settle-ment. On the other hand, a demand side explanation would argue that multiple third parties are necessary to bring in various sources of information and

128

leverage, or, as I argued above, to counteract any perceived biases such that each belligerent has a third party at the table that more closely represents their inter-ests. In light of these possible explanations, future research has to unveil the factors that determine multi-party mediation for dealing with existing selection problems more effectively than in this book.

129

7. Conclusion The four preceding chapters sought both to theoretically study and empirically test different aspects of actors’ interaction in the process of international mediation. While Chapters 3 and 4 examined the interactions of belligerents via various ties and their impact on mediation onset, Chapter 5 and Chapter 6 went on by clarifying whether multi-track diplomacy/multi-party mediation exerts a positive impact on mediation effectiveness.

The main findings are straightforward and worth knowing. First, the occurrence of mediation is more likely, the higher the interconnectedness between combatants via intermediaries. Warring parties that have many and dense links to actors outside the dyad have more possibilities to exchange information while bringing in a third party that may well have an interest in the conflict. Further, these links increase the availability of an actor willing to intervene on the mediation market. However, my findings suggested that direct links between antagonists do not notably increase the chances that a dispute will see mediation most prominently due to the fact that these direct avenues of communication become torn apart with dispute.

Second, when examining the interactions and ties between disputing parties, I demonstrated that we have to take into account both the characteristics and the underlying nature of those links. As Chapter 4 showed, the findings on belligerents’ ties from Chapter 3 will most likely change when we look at commercial interactions of states. Due to the inherent costs and benefits of trade ties, direct trade flows will increase the likelihood of mediation onset. The higher the commercial exchange between combatants before a dispute occurs and mediation might be attempted, the more dependent are these states on each other; further, the impact of sanctions from the dyadic counterpart is then highly detrimental as well. Ultimately, this increases the chances of mediation significantly. On the other hand, however, a fighting dyad’s trade links with states outside this dyad notably decrease the likelihood of mediation incidence. If many and dense trade with third parties exists, then belligerents are less dependent on the bilateral trade exchange and may also be able to circumvent any kind of sanctions. Furthermore, looking separately at bilateral trade between antagonists and trade with parties outside the dyad only tells us “one side of the story.” I theoretically outlined that we also have to be aware of the interaction of

T. Böhmelt, International Mediation Interaction, DOI 10.1007/978-3-531-92812-8_7,© VS Verlag für Sozialwissenschaften | Springer Fachmedien Wiesbaden GmbH 2011

130

those ties. In other words, even if strong bilateral trade ties do exist between disputing parties, the chances of mediation onset may still be low if at least one of the belligerents has strong trade exchanges with outside countries.

Chapters 3 and 4 set the foundation for the succeeding analyses, since the former examined the likelihood of mediation onset from the perspective of combatants’ interactions. In what followed, Chapters 5 and 6 changed the point of view and focused on how the interaction of mediators determines mediation outcomes. As a starting point, Chapter 5 employed the Tracks of Diplomacy framework (Ziegler 1984; De Magalhaes 1988; Diamond and McDonald 1996) as a suitable concept for analysis. I found evidence that official tracks of intervention are the most successful ones in settling conflicts peacefully due to higher leverage, more resources and better prospects for enforcement. This finding alone already contributes a lot towards clarifying “which mediators are more effective in settling disputes and why they are more effective” (Frazier and Dixon 2006: 391). More importantly, however, my results suggest that mediation efforts conducted by one third party perform worse than interventions that involve more than one mediator. Mediation attempts where both official and unofficial intermediaries acted, coordinated with each other and negotiated with the belligerents were by far more effective than efforts without such an interaction of mediating parties.

Nevertheless, Chapter 5 analysed this type of multi-track diplomacy or multi-party mediation from a rather naïve perspective using a dichotomous variable. As the last substantial chapter emphasized, the interaction of multiple mediators is far more complex. Here, I disaggregated multi-party mediation efforts into some essential factors, i.e. the size of the intervening coalition, conflicting and cooperative interactions as well as regime type and estimated the impact on mediation effectiveness again. The findings proposed that a naïve approach once again does not fully uncover what is going on in conflict managements involving multi-party mediation. More precisely, I found evidence that both small and large mediating coalitions are less effective dispute managers than coalitions of a rather moderate size. While very small groups are not able to create much leverage or only have a few resources, a set of many mediators suffers from high organisational costs and preference heterogeneity. Furthermore, although I obtained support for my claims that cooperative interactions between mediators facilitate an intervention, it seems that previous conflicting interactions between third parties are not much help in explaining variance in mediation effectiveness. Finally, I argued that more democratic coalitions of mediators are more effective in settling disputes, since they have a higher level of inclusivity, a shared culture of peaceful conflict resolution and increased communication flows. However,

131

my results show that the overall regime type of an intervening coalition is rather unlikely to determine if mediators are successful.

Against this background of the main findings of this book, in the remainder of this conclusion I first discuss and summarise what this manuscript adds to the academic debate in total and what kind of policy implications we can draw from my results. Here, I will also have a closer look at the control covariates that I employed throughout the chapters. Second, research only can make progress via a lively academic debate. Hence, my theoretical arguments and research designs as outlined above may be only one possible approach out of many. Consequent-ly, I briefly address some weaknesses and issues of my conducted research, i.e. simply trying to answer if there is anything that could have been done in another way. Finally, while the specific concluding sections in each chapter outlined some avenues for further research already, I summarise these suggestions and emphasize the most promising ones.

7.1. Academic Contribution and Policy Implications The main findings as summarised above are novel, rest on newly developed arguments in the theoretical literature of international mediation and, hence, each specific chapter as such certainly contributes to our understanding of both mediation incidence and effectiveness. From a broader point of view, however, the book also adds to the academic debate in the following five ways. First, as emphasized in various sections, mediation onset is a largely underexamined field of study. To address this, I theoretically developed and empirically tested a new approach to determining mediation onset via various linkages of belligerents. As stated in the introduction, understanding why mediation happens in the first place helps understanding mediation effectiveness. This book therefore contri-butes to have a more accurate grasp of the direct causal effects of mediation and mediation effectiveness, since it sheds further light on why some disputes get mediated while others do not see any third-party involvement at all.

Second, next to the theoretical arguments and the findings, the manuscript introduced two approaches to the study of international mediation that have largely been ignored so far. On one hand, to my knowledge Chapter 3 is one of the only thorough examinations of mediation using social network analysis. Social network analysis offers the possibility to draw inferences from analysing actors’ interactions, ties and relationships with each other. Given that inter-national mediation is an interactive process by definition and against the back-ground that the social network approach in this book proved to be beneficial, I believe that my book contributes to fostering social network analysis in the

132

academic study of international mediation. On the other hand, I argued that conflicts that actually see mediation are unlikely to be a random sample. Con-sequently, I addressed this selection problem in each chapter and – most prominently – treated mediation effectiveness as a function of a two-stage decision-making process using a Heckman selection estimator. However, explicitly recognizing this selection problem and properly addressing it still has to emerge in the study of international mediation. For example, Beardsley (2006), Beardsley et al. (2006), Gartner and Bercovitch (2006) and Schneider, Bercovitch and Selck (2006) are a few noteworthy exceptions here, but it is not yet conventional wisdom that ignoring the selection problem of mediation onset and effectiveness may either over- or underestimate their consequences, leading to biased results.

Third, recent research questions the effectiveness of international mediation per se in the long run (e.g. Beardsley 2008) or emphasizes that mediators with little leverage are only considered due to devious objectives and have a low impact at best on the peace process (Beardsley 2009). My book, however, shows not only that mediators can be effective (i.e. Chapters 5 and 6), but also that mediators without leverage, e.g. Track Two Diplomacy, can have a significantly positive influence on settling conflicts peacefully (i.e. Chapter 5) and that multi-party mediation generally can be very successful in settling disputes peacefully in the long run (i.e. Chapters 5 and 6). My robust findings are likely to stimulate the academic debate on that.

Forth, Beardsley and Greig (2009: 247) recently note that “we also do not yet have a clear understanding of the degree to which the presence of multiple mediators improves or detracts from the conflict management process.” In other words, there is a crucial lack of understanding and knowledge on multi-party mediation. In this respect, Chapter 5 to some extent and Chapter 6 are the first to offer a comprehensive theoretical rationale and empirical evaluation that future studies might want to address in further uncovering how one mediator shapes the behaviour and impact of others.

Finally and with regards to the control covariates I employed in the manuscript, it is worth noting that we largely find patterns that are consistent with the previous literature. There are two exceptions, however. First, there is a lively academic debate on whether mediation occurs in those conflicts that are relatively easy to mediate (e.g. Maoz and Terris 2006) or in those disputes that are considered to be relatively “hard” (e.g. Bercovitch and Gartner 2006a; Gartner and Bercovitch 2006). My work takes an alternative perspective by arguing that both perspectives comprise some elements of truth and that dispute intensity and mediation onset/effectiveness are characterised by an inverted U-shaped relationship. My findings support this claim and are largely robust

133

throughout the different chapters. Hence, although dispute intensity was not the actual focus of my project, the evidence for the curvilinear relationship is novel as well and may be able to contribute to the outlined academic discussion on conflict intensity and international mediation. Second, the majority of the empirical models included a variable on the incentives of the disputants as measured by combatants’ requests for third-party mediation. Surprisingly, and contrary to my expectations, this item had a negative influence on both me-diation incidence and effectiveness. Although it therefore seems that mediation does not occur or is not effective if third-parties lack sufficient incentives – despite a possible sufficient level of disputants’ incentives – I do not have a coherent explanation for this and leave it to future.

With respect to the policy implications of my book, the concluding sections of the different chapters emphasized a few already and highlighted how decision- makers may be able to use mediation more effectively as a conflict resolution instrument. Nonetheless, let me draw attention to the ones that I perceive to be most important. First, in order to make mediation as an effective conflict management tool to occur, decisionmakers have to consider the complex ties and interactions of belligerents more closely. On one hand, it is not sufficient to only examine their bilateral direct connections. Other linkages through various and multiple intermediaries, i.e. indirect links, tell us far more when estimating the chances that combatants will be willing to consider offers to mediate. On the other hand, policymakers who may want to intervene have to take into account the nature of those ties as well. As demonstrated, trade links are likely to have a different impact than interactions via IGOs. Having said that, the chances of mediation occurrence increase significantly if these two aspects are incorporated into the decision-making process of state leaders.

Second, mediation can be an effective tool for settling disputes. Hence, policymakers should continue to consider this instrument. Moreover, my findings suggest that more mediators are generally more likely to be successful in conflict resolution, but only to a certain extent. If a coalition of interveners becomes too large, my research demonstrates that it may well be that we observe an outcome that is contrary to the expectations in the first place and a settlement attempt could become ineffective or actually counterproductive. Put differently, decision makers have to be highly aware of the actual size and composition of a mediating coalition.

134

7.2. Discussion of Research Despite the striking findings that both contribute to the academic debate and comprise a set of crucial policy implications, the approach in this manuscript and the research designs of the various chapters present only one possible way out of many. First, the underlying rational approach to both conflict and dispute management and the quantitative designs are one alternative for studying mediation. However, other paradigms might apply as well and could have different views on some of the arguments I presented above. The most prominent other options are given by social constructivism that highlights the influence of perceptions, norms and identity constructs (see e.g. Schroeder 2006) or a qualitative research design.

Second, my book does rather not rely on the formal modelling of its theoretical arguments. However, extending the theory by including formal arguments and models could have increased their persuasiveness and, given that there are some existing mediation models in the literature (e.g. Kydd 2003; Terris and Maoz 2005; Kydd 2006; Favretto 2009), could have been applied here. Mostly due the focus on empirical models, I did not consider them in the book and will leave it upon succeeding studies to further develop the usage of formal models in international mediation research.

Third, I took into account the selection problem of international mediation in my research design. Among other strategies, I employed a zero-inflated logit estimator (Beger et al. 2009) in order to model that there are actually multiple reasons for not observing mediation in conflicts. As stated above, I only had the possibility to use a very preliminary version of this estimator that did not offer the possibility to calculate marginal effects or predicted probabilities for instance. However, due to its preliminary nature I only used it for simple robustness checks and did not employ it as a core model. A more developed version of a zero-inflated estimator is likely to be beneficial for future analyses.

Lastly, I described in detail the data sets I employed and presented reasons why I prefer these over others. Still, despite the rationale I had for mainly relying on the ICM project (Bercovitch 1999), depending on the research design a scholar may see the other data sets as more promising. Note that due to a different unit of analysis, time frame and variable operationalisation the results of this book could change, however.

135

7.3. Avenues for Future Research As presented in the introduction and the literature reviews of each chapter, the research on mediation has gone far towards getting a more accurate understanding of the conditions of mediation onset and effectiveness. This book seeks to contribute to this debate. Throughout my work, I was able to find empirical evidence for most of my hypotheses, but, as indicated, many other avenues for further research do exist.

First, Chapters 3 and 4 focused on the incidence of mediation. However, extending the logic to the outcomes of mediation may be worth studying, given that the majority of the literature focuses on mediation effectiveness. In a similar manner and with an appropriate research design, one can address the question of “who mediates” there as well.

Second, although Chapter 5 included conflicts with non-state actors and mediation attempts that were conducted by individuals or NGOs, a weakness of my manuscript is its state-centred approach. However, this limitation is due to the availability of data. Future research may therefore seek to collect new data on mediation onset, effectiveness and the various interactions and ties of the actors involved. More specifically, social network data involving non-state actors is necessary and researches also have to find ways for overcoming the caveats of the ex-ante nature of the existing data sets on international mediation.

Third, as demonstrated in Chapter 5, my results show that mediators in general have positive impacts on mediation effectiveness. At the same time, there may be specific dispute characteristics that make a mediator or a coalition of interveners less effective than under other circumstances. Although it was not the focus of this manuscript, future research has to unveil the interaction effects between conflict characteristics and specific mediators.

Finally, despite strong theoretical arguments claiming that a coalition of (largely) democratic states will be more effective than an autocratic intervening coalition, my analysis indicates that regime types tend to be of little importance for assessing interveners’ effectiveness. Further studies will have to address this again, perhaps by focusing on the alternative explanations I provided.

Among others, each of these suggestions represents important avenues for further research for scholars and students interested in international mediation onset and effectiveness – not only from the perspective of actors’ interactions.

137

Bibliography Agha, Hussein, Shai Feldman, Ahmad Khalidi and Zeev Scheff. 2003. Track-II Diplomacy: Lessons

from the Middle East. Cambridge, MA: The MIT Press. Ai, Chunrong and Edward C. Norton. 2003. “Interaction Terms in Logit and Probit Models.”

Economics Letters 80 (1): 123-129. Anderton, Charles H., and John R. Carter. 2001. “The Impact of War on Trade: An Interrupted Time-

Series Study.” Journal of Peace Research 38 (4): 445-457. Austen-Smith, David. 1992. “Strategic Models of Talk in Political Decision Making.” International

Political Science Review 13 (1): 45-58. Axelrod, Robert and Robert O. Keohane. 1985. “Achieving Cooperation under Anarchy: Strategies

and Institutions.” World Politics 38 (1): 226-254. Ban, Ki-moon. 2009. Report of the Secretary-General on Enhancing Mediation and its Support

Activities. United Nations Security Council Document S/2009/189. Barbieri, Katherine. 1996. “Economic Interdependence: A Path to Peace or a Source of Interstate

Conflict?” Journal of Peace Research 33 (1): 29-49. Barbieri, Katherine. 1998. Trade and Conflict: Assessing the Impact of Interdependence on

Militarized Conflict. Unpublished Manuscript. Barbieri, Katherine and Jack S. Levy. 1999. “Sleeping with the Enemy: The Impact of War on

Trade.” Journal of Peace Research 36 (4): 463-479. Barbieri, Katherine, Omar M. G. Keshk and Brian Pollins. 2009. “Trading Data: Evaluating our

Assumptions and Coding Rules.” Conflict Management and Peace Science 26 (5): 471-491. Beardsley, Kyle. 2008. “Agreement without Peace? International Mediation and Time Inconsistency

Problems.” American Journal of Political Science 52 (4): 723-740. Beardsley, Kyle. 2009. “Intervention without Leverage: Explaining the Prevalence of Weak

Mediators.” International Interactions 35 (3): 272-297. Beardsley, Kyle. 2010. “Pain, Pressure, and Political Cover: Explaining Mediation Incidence.”

Journal of Peace Research 47 (4): 395-406. Beardsley, Kyle. 2006. Politics by Means Other than War: Understanding International Mediation.

San Diego, CA: University of California. Beardsley, Kyle and J. Michael Greig. 2009. “Disaggregating the Incentives of Conflict

Management: An Introduction.” International Interactions 35 (3): 243-248. Beardsley, Kyle, David M. Quinn, Bidisha Biswas and Jonathan Wilkenfeld. 2006. “Mediation Style

and Crisis Outcomes.” Journal of Conflict Resolution 50 (1): 58-86. Beck, Nathaniel, Jonathan N. Katz and Richard Tucker. 1998. “Taking Time Seriously: Time-Series

Cross-Section Analysis with a Binary Dependent Variable.” American Journal of Political Science 42 (4): 1260-1288.

Beger, Andreas, Jaqueline H.R. DeMeritt, Wonjae Hwang and Will H. Moore. 2009. The Split Population Logit (SPopLogit): Modeling Measurement Bias in Binary Data. Unpublished Manuscript.

Bennett, D. Scott & Matthew C. Rupert. 2003. “Comparing Measures of Political Similarity: An Empirical Comparison of S versus b in the Study of International Conflict.” Journal of Conflict Resolution 47 (3): 367-393.

T. Böhmelt, International Mediation Interaction, DOI 10.1007/978-3-531-92812-8,© VS Verlag für Sozialwissenschaften | Springer Fachmedien Wiesbaden GmbH 2011

138

Bennett, D. Scott and Allan C. Stam. 1996. “The Duration of Interstate Wars, 1816-1985.” American Political Science Review 90 (2): 1260-1288.

Bercovitch, Jacob. 1999. International Conflict Management 1945-1995: Official Codebook for the International Conflict Management Dataset. Christchurch.

Bercovitch, Jacob. 1997. “Mediation in International Conflict: An Overview of Theory, a Review of Practice.” In William Zartman and J. Lewis Rasmussen, eds., Peacemaking in International Conflict. Washington, DC: U.S. Institute for Peace Press (125-153).

Bercovitch, Jacob. 1985. “Third Parties in Conflict Management: The Structure and Conditions of Effective Mediation in International Relations.” International Journal 40 (4): 736-752.

Bercovitch, Jacob, J. Theodore Anagnoson and Donnette L. Wille. 1991. “Some Conceptual Issues and Empirical Trends in the Study of Successful Mediation in International Relations.” Journal of Peace Research 28 (1): 7-17.

Bercovitch, Jacob and Scott S. Gartner. 2006a. “Empirical Studies in International Mediation – Introduction to a Special Issue of International Interactions.” International Interactions 32 (4): 319-328.

Bercovitch, Jacob and Scott S. Gartner. 2006b. “Is There Method in the Madness of Mediation? Some Lessons for Mediators from Quantitative Studies of Mediation.” International Interactions 32 (4): 329-354.

Bercovitch, Jacob and Allison Houston. 2000. “Why Do They Do It Like This? An Analysis of the Factors Influencing Mediation Behavior in International Conflicts.” Journal of Conflict Resolution 44 (2): 170-202.

Bercovitch, Jacob and Richard Jackson. 1997. International Conflict: A Chronological Encyclopaedia of Conflicts and Their Management, 1945-1995. Washington, DC: Congressional Quarterly.

Bercovitch, Jacob and Jeffrey Langley. 1993. “The Nature of the Dispute and the Effectiveness of International Mediation.” Journal of Conflict Resolution 37 (4): 670-691.

Bercovitch, Jacob and Su-Mi Lee. 2003. “Exploring the Relevance and Effectiveness of Directive Strategies in Mediation.” International Journal of Peace Sciences 8 (1): 1-19.

Bercovitch, Jacob and Patrick M. Regan. 1999. “The Structure of International Conflict Management: An Analysis of the Effects on Intractability and Mediation.” International Journal of Peace Studies 4 (1): 1-19.

Bercovitch, Jacob and Gerald Schneider. 2000. “Who Mediates? The Political Economy of International Conflict Management.” Journal of Peace Research 37 (2): 145-165.

Böhmelt, Tobias. 2009. “International Mediation and Social Networks: The Importance of Indirect Ties.” International Interactions 35 (3): 298-319.

Boehmer, Charles, Erik Gartzke and Timothy Nordstrom. 2004. “Do Intergovernmental Organizations Promote Peace?” World Politics 5 (7): 1-38.

Boehmer, Charles R., Bernadette M.E. Jungblut and Richard J. Stoll. 2009. Tradeoffs in Trade Data: Do Our Assumptions Affect Our Results?. Unpublished Manuscript.

Brambor, Thomas, William R. Clark and Matt Golder. 2006. “Understanding Interaction Models: Improving Empirical Analysis.” Political Analysis 14 (1): 63-82.

Braumoeller, Bear F. 2004. “Hypothesis Testing and Multiplicative Interaction Terms.” International Organization 58 (3): 807-820.

Brecher, Michael and Jonathan Wilkenfeld. 2009. International Crisis Behavior Project, 1918-2006. Ann Arbor, MI: Inter-University Consortium for Political and Social Research.

Buchanan, James M. and Gordon Tullock. 1962. The Calculus of Consent: Logical Foundations of Constitutional Democracy. Ann Arbor, MI: University of Michigan Press.

Bueno de Mesquita, Bruce. 1978. “Systematic Polarization and the Occurrence and Duration of War.” Journal of Conflict Resolution 22 (2): 241-267.

Bueno de Mesquita, Bruce. 1981. The War Trap. New Haven, CT: Yale University Press.

139

Carter, Jimmy. 1995. Talking Peace: A Vision for the Next Generation. New York: Dutton Children’s Books.

Chamberlain, Gary. 1980. “Analysis of Covariance with Qualitative Data.” Review of Economic Studies 47 (1): 225-238.

Chamberlain, John. 1974. “Provision of Collective Goods as a Function of Group Size.” American Political Science Review 68 (2): 707-716.

Chataway, Cynthia J. 1999. “The Evolution of Diplomacy: Coordinating Tracks I and II.” In A. Walter Dorn, ed., World Order for a New Millennium: Political, Cultural and Spiritual Approaches to Building Peace. New York: St. Martin’s Press (139-146).

Chataway, Cynthia J. 1998. “Track II Diplomacy: From a Track I Perspective.” Negotiation Journal 14 (3): 269-287.

Chay, Kenneth and Dean Hyslop. 1998. Identification and Estimation of Dynamic Binary Response Panel Data Models: Empirical Evidence using Alternative Approaches. Unpublished Manuscript. Available at: http://www.irle.berkeley.edu/cle/pdf/5.pdf (accessed December 2009).

Chigas, Diana. 2003. “Track II (Citizen) Diplomacy.” In Guy Burgess and Heidi Burgess, eds., Beyond Intractability. Boulder, CO: Conflict Research Consortium, University of Colorado.

Chigas, Diana. 1997. “Unofficial Interventions with Official Actors: Parallel Negotiation Training in Violent Intrastate Conflicts.” International Negotiation 2 (3): 409-436.

Choi, Ajin 2003. “The Power of Democratic Cooperation.” International Security 28 (1): 142-153. Choukri-Fishere, Ezzedine. 2008. Against Conventional Wisdom: Mediating the Arab-Israeli

Conflict. Oslo: Centre for Humanitarian Dialogue. Clarke, Kevin A. 2005. “The Phantom Menace: Omitted Variable Bias in Econometric Research.”

Conflict Management and Peace Science 22 (4): 341-352. Cox, Robert W. and Harold K. Jacobson. 1973. The Anatomy of Influence. New Haven, CT: Yale

University Press. Crescenzi, Mark J. and Andrew J. Enterline. 2001. “Time Remembered: A Dynamic Model of

Interstate Interaction.” International Studies Quarterly 45 (3): 409-431. Crescenzi, Mark J., Andrew J. Enterline and Stephen B. Long. 2008. “Bringing Cooperation Back In:

A Dynamic Model of Interstate Interaction.” Conflict Management and Peace Science 25 (3): 264-280.

Crescenzi, Mark J.C., Kelly M. Kadera, Sara McLaughlin Mitchell and Clayton L. Thyne. 2008. A Supply Side Theory of Mediation. Unpublished Manuscript.

Crocker, Chester A., Fen O. Hampson and Pamela Aall (eds), 1999. Herding Cats: Multiparty Mediation in a Complex World. Washington, DC: U.S. Institute for Peace Press.

Crocker, Chester A., Fen O. Hampson and Pamela Aall. 2001a. “Is More Better? The Pros and Cons of Multiparty Mediation.” In Turbulent Peace: The Challenges of Managing International Conflict, Chester A. Crocker, Fen O. Hampson and Pamela Aall (eds.). Washington, DC: U.S. Institute for Peace Press.

Crocker, Chester A., Fen O. Hampson and Pamela Aall. 2001b. “A Crowded Stage: Liabilities and Benefits of Multiparty Mediation.” International Studies Perspectives 2(1): 51-67.

Crumm, Eileen M. 1995. “The Value of Economic Incentives in International Politics.” Journal of Peace Research 32 (3): 313-330.

Davidson, William D. and Joseph Montville. 1981. “Foreign Policy According to Freud.” Foreign Policy 45 (4): 145-157.

De Magalhaes, José C. 1988. The Pure Concept of Diplomacy. New York: Greenwood Press. Desch, Michael C. 2002. “Democracy and Victory: Why Regime Type Hardly Matters.”

International Security 27 (2): 5-47. Deutsch, Morton. 1973. The Resolution of Conflict. New Haven, CT: Yale University Press. Diamond, Loise and John W. McDonald. 1996. Multitrack Diplomacy – A Systems Approach. 3rd ed.

West Hartford, CT: Kumarian Press.

140

Dixon, William J. 1994. “Democracy and the Peaceful Settlement of International Conflict.” American Political Science Review 88 (1): 14-32.

Dixon, William J. 1996. “Third-Party Techniques for Preventing Conflict Escalation and Promoting Peaceful Settlement.” International Organization 50 (4): 653-681.

Dorussen, Han. 2006. “Heterogeneous Trade Interests and Conflict.” Journal of Conflict Resolution 50 (1): 87-107.

Dorussen, Han and Hugh Ward. 2008a. “Inter Governmental Organizations and the Kantian Peace – A Network Perspective.” Journal of Conflict Resolution 52 (2): 189-212.

Dorussen, Han and Hugh Ward. 2010. “Trade Networks and the Kantian Peace.” Journal of Peace Research 47 (1): 29-42.

Dorussen, Han and Hugh Ward. 2008b. “Trade Networks and the Politics of Cooperation and Conflict.” Paper prepared for the 104th Annual Meeting of the American Political Science Association, August 28-31, Boston, MA.

Doyle, Michael W. and Nicholas Sambanis. 2000. “International Peacebuilding: A Theoretical and Quantitative Analysis.” American Political Science Review 94 (4): 779-801.

Favretto, Katja. 2009. “Should Peacemakers Take Sides? Major Power Mediation, Coercion and Bias.” American Political Science Review 103 (2): 248-263.

Fearon, James D. 1998. “Bargaining, Enforcement, and International Cooperation.” International Organization 52 (2): 269-306.

Fearon, James D. 1995. “Rationalist Explanations for War.” International Organization 49 (3): 379-414.

Fearon, James and David Laitin. 2003. “Ethnicity, Insurgency, and Civil War.” American Political Science Review 97 (1): 75-90.

Filson, Darren and Suzanne Werner. 2002. “A Bargaining Model of War and Peace: Anticipating the Onset, Duration, and the Outcome of War.” American Journal of Political Science 46 (4): 819-837.

Fisher, Ronald J. 2007. “Assessing the Contingency Model of Third-Party Intervention in Successful Cases of Prenegotiation.” Journal of Peace Research 44 (3): 311-329.

Fisher, Ronald J. and Loraleigh Keashley. 1991. “The Potential Complementarity of Mediation and Consultation within a Contingency Model of Third Party Intervention.” Journal of Peace Research 28 (1): 29-42.

Ford, Lester R. and Delbert R. Fulkerson. 1956. “Maximal Flow through a Network.” Canadian Journal of Mathematics 8 (3): 399-404.

Frazier, Derrick and William Dixon. 2006. “Third-Party Intermediaries and Negotiated Settlements, 1946-2000.” International Interactions 32 (4): 385-408.

Frohlich, Norman and Joe A. Oppenheimer. 1970. “I get by with a Little Help from my Friends.” World Politics 23 (1): 104-120.

Garb, Paula and Susan A. Nan. 2006. “Negotiating in a Coordination Network of Citizen Peacebuilding Initiatives in the Georgian-Abkhaz Peace Process.” International Negotiation 11 (1): 7-35.

Garriga, Ana Carolina. 2009. “Regime Type and Bilateral Treaty Formalization.” Journal of Conflict Resolution 53 (5): 698-726.

Gartner, Scott S. and Jacob Bercovitch. 2006. “Overcoming Obstacles to Peace: The Contribution of Mediation to Short-Lived Conflict Settlements.” International Studies Quarterly 50 (4): 819-840.

Gartzke, Erik. 1999. “War Is in the Error Term.” International Organization 53 (3): 567-587. Gasiorowski, Mark and Solomon W. Polachek. 1982. “Interdependence and International Conflict:

Some Cross-National Evidence.” International Studies Quarterly 30 (1): 23-28. Gent, Stephen E. 2007. “Strange Bedfellows: The Strategic Dynamics of Major Power Military

Interventions.” Journal of Politics 69 (4): 1089-1102.

141

Gent, Stephen E. and Megan Shannon. 2007. It’s Not Who but How: Third Parties and the Peaceful Settlement of Territorial Claims. Paper prepared the Annual Meeting of the American Political Science Association, August 30-September 2.

Gleditsch, Kristian S. 2002. “Expanded Trade and GDP Data.” Journal of Conflict Resolution 46 (5): 712-724.

Gleditsch, Kristian S. and Kyle Beardsley. 2004. “Nosy Neighbors: Third-Party Actors in Central American Conflicts.” Journal of Conflict Resolution 48 (3): 379-402.

Goertz, Gary and Paul F. Diehl. 1993. “Enduring Rivalries: Theoretical Constructs and Empirical Patterns.” International Studies Quarterly 37 (2): 147-171.

Gowa, Joanne. 1989. “Bipolarity, Multipolarity, and Free Trade.” American Political Science Review 83 (4): 1245-1256.

Gowa, Joanne and Edward D. Mansfield. 1993. “Power Politics and International Trade.” American Political Science Review 87 (2): 408-420.

Greig, Michael J. 2005. “Stepping into the Fray: When do Mediators Mediate?” American Journal of Political Science 49 (2): 249-266.

Greig, Michael J. and Paul F. Diehl 2006. “Softening Up: Making Conflicts More Amenable to Diplomacy.” International Interactions 32 (4): 355-384.

Greig, Michael J. and Patrick M. Regan. 2008. “When Do They Say Yes? An Analysis of the Willingness to Offer and Accept Mediation in Civil Wars.” International Studies Quarterly 52 (4): 759-781.

Hafner-Burton, Emilie and Alexander Montgomery. 2006. “Power Positions – International Organizations, Social Networks, and Conflict.” Journal of Conflict Resolution 50 (1): 3-27.

Heckman, James. 1979. “Sample Selection Bias as a Specification Error.” Econometrica 47 (1): 153-161.

Hensel, Paul R. 2001. “Contentious Issues and World Politics: The Management of Territorial Claims in the Americas, 1816-1992.” International Studies Quarterly 45 (1): 81-109.

Hensel, Paul R. and Sara McLaughlin Mitchell. 2007. Issues Correlates of War Project: User Manual for ICOW Data, Version 1.1. Available at <http://www.icow.org>.

Hensel, Paul R., Sara McLaughlin Mitchell, Thomas E. Sowers and Clayton L. Thyne. 2008. “Bones of Contention: Comparing Territorial, Maritime, and River Issues.” Journal of Conflict Resolution 52 (1): 117-143.

Jackson, Richard. 2000. “Successful Negotiation in International Violent Conflict.” Journal of Peace Research 37 (3): 323-343.

Kathman, Jake. 2007. The Geopolitics of Civil War Intervention. PhD Dissertation. University of North Carolina, NC.

Kaye, Dalia. 2001. “Track Two Diplomacy and Regional Security in the Middle East.” International Negotiation 6 (1): 49-77.

Keashly, Loraleigh and Ronald J. Fisher. 1996. “A Contingency Perspective on Conflict Interventions: Theoretical and Practical Considerations.” In Jacob Bercovitch, ed., Resolving International Conflicts. Boulder, CO: Lynne Rienner (235-262).

Kelman, Herbert C. 2006. “The Role of an International Facilitating Service for Conflict Resolution.” International Negotiation 11 (1): 209-223.

Keohane, Robert O. 1984. After Hegemony: Cooperation and Discord in the World Political Economy. Princeton, NJ: Princeton University Press.

Keohane, Robert O. 1983 “The Demand of International Regimes.” In Stephen D. Krasner, ed., International Regimes. Ithaca, NY: Cornell University Press: 162-163.

Keshk, Omar M. G., Brian M. Pollins and Rafael Reuveny. 2004. “Trade Still Follows the Flag: The Primacy of Politics in a Simultaneous Model of Interdependence and Armed Conflict.” Journal of Politics 66 (4): 1155-1179.

142

Kim, Hyung Min and David L. Rousseau. 2005. “The Classical Liberals Were Half Right (or Half Wrong): New Tests of the ‘Liberal Peace’, 1960-88.” Journal of Peace Research 42 (5): 523-543.

King, Gary, Robert O. Keohane and Sidney Verba. 1994. Designing Social Inquiry: Scientific Inference in Qualitative Research. Princeton, NJ: Princeton University Press.

Kleiboer, Marieke. 1996. “Understanding Success and Failure of International Mediation.” Journal of Conflict Resolution 40 (2): 360-389.

Kressel, Kenneth and Dean G. Pruitt. 1989. “Conclusion: A Research Perspective on the Mediation of Social Conflict.” In Mediation Research, eds. Kenneth Kressel and Dean G. Pruitt. San Francisco, CA: Jossey-Bass: 241-222.

Kydd, Andrew. 2006. “When Can Mediators Build Trust?.” American Political Science Review 100 (3): 449-462.

Kydd, Andrew. 2003. “Which Side Are You On? Bias, Credibility, and Mediation.” American Journal of Political Science 47 (4): 597-611.

Kydd, Andrew. 2010. “Rationalist Approaches to Conflict Prevention.” Annual Review of Political Science 13: 101-121.

Kydd, Andrew and Duncan Snidal. 1997. “Progress in Game-Theoretical Analysis of International Regimes.” in Andreas Hasenclever, Peter Mayer and Volker Rittberger, eds., Theories of International Relations, Cambridge, MA: Cambridge University Press (112-135).

Lederach, John P. 1997. Building Peace: Sustainable Reconciliation in Divided Societies. Washington, DC: United States Institute of Peace.

Levy, Jack S. and Katherine Barbieri. 2004. “Trading with the Enemy during Wartime.” Security Studies 13 (3): 1-47.

Li, Quan and David Sacko. 2002. “The (Ir)Relevance of Militarized Interstate Disputes for International Trade.” International Studies Quarterly 46 (1): 11-43.

Lieberfeld, Daniel. 2002. “Evaluating the Contributions of Track-Two Diplomacy of Conflict Termination in South Africa, 1984-1990.” Journal of Peace Research 39 (3): 355-372.

Mansfield, Edward D. 1994. Power, Trade, and War. Princeton, NJ: Princeton University Press. Mansfield, Edward D. and Rachel Bronson. 1997. “Alliances, Preferential Trading Arrangements,

and International Trade.” American Political Science Review 91 (1): 94-107. Maoz, Zeev. 2009. “The Effects of Strategic and Economic Interdependence on International Conflict

across Levels of Analysis.” American Journal of Political Science 53 (1): 223-240. Maoz, Zeev, Ranan D. Kuperman, Lesley Terris and Ilan Talmud, 2006. “Structural Equivalence and

International Conflict – A Social Networks Analysis.” Journal of Conflict Resolution 50 (5): 664-689.

Maoz, Zeev and Lesley Terris. 2006.”Credibility and Strategy in International Mediation.” International Interactions 32 (4): 409-440.

Mapendere, Jeffrey. 2000. Defining Track One-and-a-Half Diplomacy: Its Complementarity and the Analysis of Factors that Facilitate its Success. Ottawa: Royal Roads University.

Mapendere, Jeffrey. 2005. “Track One and a Half Diplomacy and the Complementarity of Tracks.” Culture of Peace Online Journal 2 (1): 66-81.

Marshall, Monty G. and Keith Jaggers. 2002. Polity IV Project: Political Regime Characteristics and Transitions, 1800-2002. Dataset User’s Manual. College Park, MD: University of Maryland.

Mason, David T. and Patrick J. Fett. 1996. “How Civil Wars End: A Rational Choice Approach.” Journal of Conflict Resolution: 40 (4): 546-568.

Mitchell, Sara McLaughlin. 2002. “A Kantian System? Democracy and Third-Party Conflict Resolution.” American Journal of Political Science 46 (4): 749-759.

Mitchell, Sara McLaughlin, Kelly M. Kadera and Mark J.C. Crescenzi. 2005. “Practicing Democratic Community Norms: Third Party Conflict Management and Successful Settlements.” In

143

Empirical Studies in International Mediation, eds. Jacob Bercovitch and Scott Sigmund Gartner. New York: Routledge.

Montville, Joseph V. 1991. “Track Two Diplomacy: The Arrow and the Olive Branch: A Case for Track Two Diplomacy.” In Vamik D. Volkan, Joseph P. Montville and Demetrios A. Julius, eds., The Psychodynamics of International Relations: Vol. 2, Unofficial Diplomacy at Work, Lanham, MD: Lexington Books (161-175).

Nan, Susan A. 2003. “Track I Diplomacy.” In Guy Burgess and Heidi Burgess, eds., Beyond Intractability. Boulder, CO: Conflict Research Consortium, University of Colorado.

Nan, Susan A. 2002. Unofficial Conflict Resolution as a Complement to Diplomacy: A Case Study of the Georgian-South Ossetian Peace Process Highlighting ‘Track One and a Half Diplomacy. Paper prepared for the annual meeting of the International Studies Association.

Nan, Susan A. and Andrea Strimling. 2006. “Coordination in Conflict Prevention, Conflict Resolution and Peacebuilding.” International Negotiation 11 (1): 1-6.

Nan, Susan A. and Andrea Strimling. 2004. “Track I – Track II Cooperation.” In Guy Burgess and Heidi Burgess, eds., Beyond Intractability. Boulder, CO: Conflict Research Consortium, University of Colorado.

North, Douglass. 1990. Institutions, Institutional Change and Economic Performance. Cambridge, MA: Cambridge University Press.

Notter, James and John McDonald. 1996. “Track Two Diplomacy: Nongovernmental Strategies for Peace.” U.S. Foreign Policy Agenda – USIA Electronic Journal 1 (19): 32-36.

Olson, Mancur. 1965. The Logic of Collective Action – Public Goods and the Theory of Groups. Cambridge, MA: Harvard University Press.

Olson, Mancur. 1993. “Dictatorship, Democracy, and Development.” American Political Science Review 87 (3): 567-576.

Olson, Mancur and Richard Zeckhauser. 1970. “An Economic Theory of Alliances.” In Julian R. Friedman, Christopher Bladen and Steven Rosen (eds), Alliance in International Politics. Boston, MA: Allyn and Bacon.

Oneal, John R. & Bruce Russett, 1999. ‘The Kantian Peace: The Pacific Benefits of Democracy, Interdependence, and International Organizations, 1885-1992’, World Politics 52 (1): 1-37.

Ott, Marvin C. 1972. “Mediation as a Method of Conflict Resolution: Two Cases.” International Organization 26 (4): 595-618.

Pearson, Frederic S. 2001. “Dimensions of Conflict Resolution in Ethno-political Disputes.” Journal of Peace Research 38 (3): 275-287.

Pevehouse, Jon, Timothy Nordstrom and Kevin Warnke. 2004. “The Correlates of War 2 International Governmental Organizations Data Version 2.0.” Conflict Management and Peace Science 21 (2):101-119.

Polachek, Solomon W. 1980. “Conflict and Trade.” Journal of Conflict Resolution 24 (1): 57-78. Polachek, Solomon W., John Robst and Yuan-Ching Chang. 1999. “Liberalism and Interdependence:

Extending the Trade-Conflict Model.” Journal of Peace Research 36 (4): 57-78. Polachek, Solomon W. and Judith A. McDonald. 1992. Strategic Trade and the Incentive for

Cooperation. In Disarmament, Economic Conversion, and Management of Peace, eds. Manas Chatterji and Linda R. Forcey. Westport: Praeger.

Pollins, Brian M. 1989a. “Conflict, Cooperation, and Commerce: The Effect of International Political Interactions on Bilateral Trade Flows.” American Journal of Political Science 33 (3): 737-761.

Pollins, Brian M. 1989b. “Does Trade Still Follow the Flag?.” American Political Science Review 83 (2): 465-480.

Princen, Thomas. 1992. Intermediaries in International Conflict. Princeton, NJ: Princeton University Press.

144

Quinn, David, Jonathan Wilkenfeld, Kathleen Smarick and Victor Asal. 2006. “Power Play: Mediation in Symmetric and Asymmetric International Crises.” International Interactions 32 (4): 442-470.

Raymond, Gregory A. 1994. “Democracies, Disputes, and Third-Party Intermediaries.” Journal of Conflict Resolution 38 (1): 24-42.

Regan, Patrick M. 2002. “Third-Party Interventions and the Duration of Intrastate Conflicts.” Journal of Conflict Resolution 46 (1): 55-73.

Regan, Patrick M. and Aysegul Aydin. 2006. “Diplomacy and Other Forms of Intervention in Civil Wars.” Journal of Conflict Resolution 50 (5): 736-756.

Regan, Patrick M., and Allan C. Stam. 2000. “In the Nick of Time: Conflict Management, Mediation Timing, and the Duration of Interstate Disputes.” International Studies Quarterly 44 (2): 239-260.

Reuveny, Rafael and Heejoon Kang. 1996. “International Trade, Political Conflict/ Cooperation, and Granger Causality.” American Journal of Political Science 40 (3): 943-970.

Rothman, Jay and Marie L. Olson. 2001. “From Interests to Identities: Towards a New Emphasis in Interactive Conflict Resolution.” Journal of Peace Research 38 (3): 289-305.

Rowe, David M. 2001. Manipulating the Market: Understanding Economic Sanctions, Institutional Change, and the Political Unity of White Rhodesia. Ann Arbor, MI: University of Michigan Press.

Rubin, Jeffrey Z. ed., 1981. The Dynamics of Third-Party Intervention: Kissinger in the Middle East. New York: Praeger.

Russett, Bruce and John R. Oneal. 2001. Triangulating Peace: Democracy, Interdependence and International Organizations. New York: Norton.

Said, Abdul A. and Charles O. Lerche. 1995. Concepts of International Politics in Global Perspective. Englewood Cliffs, NJ: Simon and Schuster.

Sanders, Harold. 1991. “Officials and Citizens in International Relations.” In Vamik D. Volkan, Joseph P. Montville and Demetrios A. Julius, eds., The Psychodynamics of International Relations: Vol. 2, Unofficial Diplomacy at Work, Lanham, MD: Lexington Books (41-69).

Savun, Burcu. 2008. “Information, Bias, and Mediation Success.” International Studies Quarterly 52 (1): 25-47.

Schneckener, Ulrich. 2002. “Making Power-Sharing Work: Lessons from Successes and Failures in Ethnic Conflict Regulation.” Journal of Peace Research 39 (2): 203-228.

Schneider, Gerald, Jacob Bercovitch and Torsten J. Selck. 2006. The Dispute or the Mediator? Structural and Personal Factors in Mediation. Paper presented at the Annual Meeting of the International Studies Association, San Diego, CA.

Schroeder, Michael. 2006. Getting the ‘Right’ Agreement? International Norms and the Social Construction of Mediator Preferences in Bosnia-Herzegovina, 1991-1995. Paper presented at the Annual Meeting of the International Studies Association, San Diego, CA.

Schwarzer, Gudrun. 1998. “The Peaceful Settlement of Interstate Conflict: Saar, Austria and Berlin.” Journal of Peace Research 35 (6): 743-757.

Sherman, Frank L. 1995. SHERFACS: A Cross-Paradigm, Hierarchical, and Contextually-Sensitive International Conflict Dataset, 1937-1985. ICPSR Version. Syracuse, NY: Syracuse University.

Sigelman, Lee and Langche Zeng. 1999. “Analyzing Censored and Sample-Selected Data with Tobit and Heckit Models.” Political Analysis 8 (2): 167-182.

Signorino, Curtis S. and Jeffrey M. Ritter. 1999. “Tau-b or not Tau-b: Measuring the Similarity of Foreign Policy Positions.” International Studies Quarterly 43 (1): 115-144.

Simmons, Beth A. 2002. “Capacity, Commitment, and Compliance – International Institutions and Territorial Disputes.” Journal of Conflict Resolution 46 (6): 829-856.

145

Simmons, Beth A. 1999. “See you in ‘Court? The Appeal to Quasi-Judicial Legal Processes in the Settlement of Territorial Disputes.” In Paul F. Diehl (ed), A Roadmap to War: Territorial Dimensions of International Conflict. Vanderbilt University Press: 205-237.

Singer, David. 1988. “Reconstructing the Correlates of War Dataset on Material Capabilities of States, 1816-1985.” International Interactions 14 (2): 115-132.

Singer, David, Stuart Bremer and John Stuckey. 1972. Capability Distribution, Uncertainty, and Major Power War, 1820-1965. In: Peace, War, and Numbers, ed. Bruce Russett. Beverly Hills, CA: Sage.

Slikker, Marco and Anne van den Nouweland. 2001. Social and Economic Networks in Cooperative Game Theory. Dordrecht: Kluwer.

Slantchev, Branislav L. 2003. “The Principle of Convergence in Wartime Negotiations.” American Political Science Review: 97 (4): 621-632.

Snidal, Duncan J. 1994. “The Politics of Scope: Endogenous Actors, Heterogeneity, and Institutions.” Journal of Theoretical Politics 6 (4): 449-472.

Stein, Kenneth W. and Samuel W. Lewis. 1996. “Mediation in the Middle East.” In Chester A. Crocker, Fen O. Hampson and Pamela Aall, eds., Managing Global Chaos: Sources of and Responses to International Conflict. Washington, DC: United States Institute of Peace (463-473).

Strimling, Andrea. 2006. “Stepping Out of the Tracks: Cooperation between Official Diplomats and Private Facilitators.” International Negotiation 11 (1): 91-127.

Terris, Lesley and Zeev Maoz 2005. “Rational Mediation: A Theory and a Test.” Journal of Peace Research 42 (5): 563-583.

Touval, Saadia and William I. Zartman. 1985. “Introduction: Mediation in Theory.” In International Mediation in Theory and Practice, eds. Saadia Touval and William I. Zartman. Boulder, CO: Westview Press: 7-17.

Wagner, R. Harrison. 2000. “Bargaining and War.” American Journal of Political Science 44 (3): 469-484.

Wall, James A., John B. Stark and Rhetta L. Standifer. 2001. “Mediation – A Current Review and Theory Development.” Journal of Conflict Resolution 45 (3): 370-391.

Walter, Barbara F. 2002. Committing to Peace. Princeton, NJ: Princeton University Press. Walter, Barbara F. 1997. “The Critical Barrier to Civil War Settlement.” International Organization

51 (3): 335-364. Ward, Hugh. 2006. “International Linkages and Environmental Sustainability: The Effectiveness of

the Regime Network.” Journal of Peace Research 43 (2): 149-166. Wassermann, Stanley and Katherine Faust. 1997. Social Network Analysis: Methods and

Applications. 2nd ed. Cambridge, MA: Cambridge University Press. Wilkenfeld, Jonathan, Kathleen Young, Victor Asal and David Quinn. 2003. “Mediating

International Crises: Cross National and Experimental Perspectives.” Journal of Conflict Resolution 47 (3): 279-301.

Wilkenfeld, Jonathan, Kathleen Young, Victor Asal and David Quinn. 2005. Mediating International Crises. New York: Routledge.

Young, Oran R. 1972. “Intermediaries: Additional Thoughts on Third-Parties.” Journal of Conflict Resolution 16 (1): 51-65.

Young, Oran R. 1967. Intermediaries: Third Parties in International Crises. Princeton, NJ: Princeton University Press.

Ziegler, David W. 1984. War, Peace, and International Politics. Boston, MA: Little, Brown and Company.