APPENDIX II: EXTENDED DISCUSSION OF CODING METHODOLOGY

APPENDIX II: EXTENDED DISCUSSION 1 OF CODING METHODOLOGY GREGORY H. FOX, KRISTEN E. BOON, AND ISAAC JENKINS TABLE OF CONTENTS I. Conflicts Coded... A2 II. Binding Versus Non-Binding Obligations... A4 III. Accounting for the Council s View of Conflicts... A6 IV. Frequency of Council Intervention as an Indication of Law-Making... A7 In order to determine whether the Security Council (the Council ) imposed a consistent set of norms on parties to non-international armed conflicts ( NIACs ), we collected original data that allows us to draw conclusions about aggregate Council action imposed over time and across conflicts. No other data set of which we are aware codes for the substance of Council resolutions on legal issues related to armed conflict. Most conflict data sets do not code Council resolutions at all. The only one that has done so the International Peace Institute s Security Council Compliance Project investigated compliance with Council demands made to parties in civil wars. 2 It did not ask if those 1. For the Article accompanying this Appendix, see Gregory H. Fox, Kristen E. Boon & Isaac Jenkins, The Contributions of United Nations Security Council Resolutions to the Law of Non-International Armed Conflict: New Evidence of Customary International Law, 67 AM. U. L. REV. 649 (2018). Please cite to this Appendix as: Gregory H. Fox, Kristen E. Boon & Isaac Jenkins, Appendix II: Extended Discussion of Coding Methodology (Feb. 2018) (unpublished appendix), https://law.wayne.edu/pdfs/fox_boon_jenkins _appendix_ii_1-17-18.pdf. 2. See JAMES COCKAYNE ET AL., INT L PEACE INST., THE UNITED NATIONS SECURITY COUNCIL AND CIVIL WAR: FIRST INSIGHTS FROM A NEW DATASET 4 (2010), https://www.ipinst.org/wpcontent/uploads/publications/ipi_rpt_unsc_and_civil_war_epub.pdf. A1

A2 APPENDIX II Feb. 2018 demands create legal obligations for the parties addressed or assess the compatibility of the actions demanded with existing international law. I. CONFLICTS CODED We created a data set comprising all Council resolutions on armed conflicts from 1990 to 2013, including those that were ongoing as of 1990. 3 We take 1990 to be an important transition point in world history, which ended U.S.-Soviet proxy wars in developing countries and ushered in a new focus on human rights in the international community. Relevant scholarship has also shown that the Council became more willing to address conflicts with assertive resolutions as the sweeping effects of the end of the Cold War triggered civil wars around the world. 4 From 1945 to 1989, the Council passed 646 resolutions, averaging about fourteen resolutions per year. From 1990 through the end of 2014, the Council passed 1549 resolutions, averaging about sixtytwo resolutions per year, 4.4 times as many. Following common practice in the quantitative conflict literature, we define an armed conflict as one involving at least 1000 battle-related deaths occurring in at least one year of the conflict. 5 This definition applies to both international and non-international conflicts. While the 1000 battle-related deaths threshold is not uncontested in the literature, 6 we selected this threshold because these severe conflicts are most likely to merit inclusion in the Council s limited agenda time. Recent research has shown that the Council tends to be involved in the 3. We also consider a limited number of thematic resolutions that are unrelated to any individual conflict. These include S.C. Res. 1265, at 1 (Sept. 17, 1999) and S.C. Res. 1894, at 1 (Nov. 11, 2009) on the Protection of Civilians; S.C. Res. 1325, at 1 (Oct. 31, 2000) on Women in Armed Conflict, S.C. Res. 1612, at 1 (July 26, 2005) on Children and Armed Conflict; and S.C. Res. 1502, at 1 (Aug. 26, 2003) on Protection of UN Personnel. These resolutions are cited frequently in conflict-specific resolutions and we code for those citations. 4. See Lisa Hultman, UN Peace Operations and Protection of Civilians: Cheap talk or Norm Implementation?, 50 J. PEACE RES. 59, 60 (2013) (noting the global security shift toward protecting civilians after the Cold War); Peter Viggo Jakobsen, National Interest, Humanitarianism, or CNN: What Triggers UN Peace Enforcement after the Cold War?, 33 J. PEACE RES. 205, 205 (1996) (discussing the rise in U.N. peace enforcement interventions following the end of the Cold War). 5. See Lotta Harbom & Peter Wallensteen, Armed Conflicts, 1946 2009, 47 J. PEACE RES. 501, 501 (2010) (categorizing warlike conflicts as those in which there were at least 1000 battle-related deaths). 6. For a discussion of the contention surrounding the 1000 battle-related deaths threshold, see Fox, Boon & Jenkins, supra note 1, at 659 n.32.

Feb. 2018 APPENDIX II A3 most deadly and destructive conflicts. 7 In the use of peacekeepers, for example, the Council selects the most difficult conflicts to solve, when measured by both battle deaths and civilian deaths. 8 The Council has also been shown to intervene most in conflicts with high levels of civilian victimization, especially after the 1999 decision to focus U.N. peacekeeping efforts on civilian protection. 9 The Uppsala Conflict Data Program ( UCDP ) from which we draw our list defines an armed conflict as one involving only twenty-five battle-related deaths. Conflicts with at least 1000 deaths are designated as wars. We have chosen to retain the term armed conflict despite the different terminology used by UCDP because it is the prevalent phrase in international law. War was once an important term of art in international law, but most scholars believe it has lost much significance in an era when declarations of war are almost non-existent, and the dominant legal paradigms speak primarily instead of the use of force (jus ad bellum) and armed conflict (jus in bello). 10 In contrast, UCDP distinguishes between wars and minor armed conflicts for purposes of gauging their intensity. 11 We employ a commonly accepted threshold in the international conflict literature but use the terminology of international law. What we describe as armed conflicts UCDP would 7. See Martin Binder, Paths to Intervention: What Explains the UN s Selective Response to Humanitarian Crises?, 52 J. PEACE RES. 712, 713 (2015); Hultman, supra note 4, at 60 (noting that the Council seems to intervene in conflicts with high levels of casualties, but explicitly instructs peacekeepers to protect civilians). 8. See VIRGINIA PAGE FORTNA, DOES PEACEKEEPING WORK?: SHAPING BELLIGERENTS CHOICES AFTER CIVIL WAR 19 (2008); Scott Gartner & Jacob Bercovitch, Overcoming Obstacles to Peace: The Contribution of Mediation to Short-lived Conflict Settlements, 50 INT L STUD. Q. 819, 835 (2006) (noting that the presence of peacekeepers in a country are a sign of an intransigent conflict likely to flare up into violence ); Michael J. Gilligan & Ernest J. Sergenti, Do UN Interventions Cause Peace? Using Matching to Improve Causal Inference, 3 Q.J. POL. SCI. 89, 91 92 (2008) (analyzing the effectiveness of peacekeeping missions at maintaining a lasting peace after serious conflicts); Mark J. Mullenbach, Deciding to Keep Peace: An Analysis of International Influences on the Establishment of Third- Party Peacekeeping Missions, 59 INT L STUD. Q. 529, 530 (2005) (explaining that most peacekeeping missions take place in states experiencing internal conflicts). 9. See Hultman, supra note 4, at 60. 10. See Christopher Greenwood, The Concept of War in Modern International Law, 36 INT L & COMP. L. Q. 283, 305 (1987) ( It is doubtful... whether it is still meaningful to talk of war as a legal concept or institution at all. ). 11. See UPPSALA CONFLICT DATA PROGRAM, UCDP/PRIO ARMED CONFLICT DATASET CODEBOOK 9 (2012), http://www.pcr.uu.se/digitalassets/124/c_124920-l_1-k_codebook_ ucdp_prio-armed-conflict-dataset-v4_2012.pdf.

A4 APPENDIX II Feb. 2018 classify as wars. We do not believe the terminology reflects differences about the nature of the conflicts under study. To create the data set we relied on UCDP s conflict data, which is widely used by conflict scholars. We used the customized reports function on their website 12 to select conflicts categorized as having at least one year with 1000 or more battle-related deaths, using the Battle-related deaths in categories (conflict) variable to select all categories at 1000 or more. We then restricted the data set to 1990 to 2013, using the Year conflict variable. The conflicts included are those that crossed the 1000 deaths threshold in a year prior to 1990 but not in the period of concern. This Uppsala data set itself is restricted to conflicts in which at least one party was a government, so conflicts solely between non-governmental parties are excluded. II. BINDING VERSUS NON-BINDING OBLIGATIONS For our hypothesis, which involves evidence of changes in law, a clear distinction exists between binding and non-binding Council statements. In coding the resolutions we took account of this distinction by denominating Council statements according to a three point scale: (1) clearly non-binding statements; (2) statements that may be binding; and (3) statements that are clearly binding. The list is set out in Appendix 1. 13 We used this scale to determine when operative paragraphs of Council resolutions impose binding obligations. 14 Several clear markers helped us to answer this question. Article 25 of the U.N. Charter (the Charter ) provides that decisions of the Council are binding on all member states. 15 Thus, when the Council explicitly cites Article 25 as the basis for an action or decides to take a particular action, there is little disputing that member states are legally bound. In addition, Article 48 of the Charter effectively replicates 12. Since we generated our data set, UCDP has substituted the customized report function for an application program interface (API) that allows users to make custom data calls on the site. See Uppsala Conflict Data Program, Dep t of Peace and Conflict Research, Application Programing Interface, UPPSALA UNIVERSITET, http://ucdp.uu.se/apidocs (last visited Feb. 26, 2018). The API allows users to replicate our data set or create their own. 13. See Fox, Boon & Jenkins, supra note 1, at 731 app. 1. 14. We recognize that in some circumstances binding obligations may be contained in preambular paragraphs to resolutions and in Presidential Statements issued by the Council. These are likely to be rare and so the current project will not miss many codable statements by their omission. An expanded version of the project in the future may include these statements. 15. U.N. Charter art. 25 ( The Members of the United Nations agree to accept and carry out the decisions of the Security Council in accordance with the present Charter. ).

Feb. 2018 APPENDIX II A5 Article 25 for decisions issued under Chapter VII of the Charter. 16 Finally, recent practice demonstrates that when Council members have wanted to signal that a resolution is to be understood as mandatory, the Council has almost always either specifically cited Chapter VII or included wording which clearly implies reliance on that chapter. 17 We therefore coded as legally binding any operative paragraphs of resolutions that (i) invoke Article 25; (ii) are denominated as decisions ; (iii) invoke Article 48; or (iv) invoke Chapter VII. 18 Importantly, much of the data we rely on in the Article for making claims of frequency of Council obligations counts only Chapter VII obligations, the clearest form of Council expression. Finally, based on a review of numerous Council resolutions and consultations with members of the UN Secretariat, employees of national missions and scholars who study the Council, we developed a list of commonly-used verbs that introduce resolutions operative paragraphs. As shown in Appendix I, we grouped these verbs into three categories: non-binding, potentially binding, and binding. 19 Thus, both paragraphs introduced by adjectives denominated as binding and resolutions/paragraphs operating under Articles 25 and 48, and Chapter VII were coded as binding. Whether the Council imposes binding obligations in any other circumstances is a much more difficult question. The International Court of Justice has held that the inquiry must begin with the language of the resolution but may then proceed to more contextual factors. 20 16. U.N. Charter art. 48(1) ( The action required to carry out the decisions of the Security Council for the maintenance of international peace and security shall be taken by all the Members of the United Nations or by some of them, as the Security Council may determine. ). 17. LORAINE SIEVERS & SAM DAWS, THE PROCEDURE OF THE UN SECURITY COUNCIL 389 (4th ed. 2014) (emphasis added). Dissenting in the Kosovo case, Judge Koroma appears to have adopted a similar view. Accordance with International Law of the Unilateral Declaration of Independence in Respect of Kosovo, Advisory Opinion, 2010 I.C.J. 402, 10 (July 22) (Koroma, J., dissenting) ( That resolution was adopted by the Security Council pursuant to Chapter VII of the Charter of the United Nations and is thus binding pursuant to Article 25 of the Charter. ). 18. The Council has recently started applying Chapter VII only to portions of resolutions, meaning that in those cases some portions of the resolution will be coded as binding and others as non-binding or possibly binding. 19. See Fox, Boon & Jenkins, supra note 1, at 731 app. 1. 20. Legal Consequences for States of Continued Presence of South Africa in Namibia (South West Africa), Notwithstanding Security Council Resolution 276, Advisory Opinion, 1971 I.C.J. 16, 114 (June 21) ( The language of a resolution of the [Council] should be carefully analyzed before a conclusion can be made as to its

A6 APPENDIX II Feb. 2018 We recognize that the verbs themselves may not provide definitive evidence of the Council s intentions and that the Council sometimes uses the same terms to describe different actions or different terms to describe the same actions. But for the present project, we were satisfied that the Category 3 verbs listed as binding were used in that manner in virtually all cases. Future researchers may wish to consult debates in the Council and/or Secretary General reports to get a clearer sense of the Council s intentions. The uncertainty concerning the degree to which obligations are binding can be taken into account when the coded data is analyzed. Our analysis measures the frequency of legally significant Council actions. Both overinclusion and undercounting would affect our results. The challenge in determining when to include obligations as binding when they are outside the universally recognized pool of verbs leads to three potential standards of inclusion: (1) only Category 3; (2) Categories 2 and 3; or (3) Category 3 and Category 2 on a case-by-case basis. We choose to include both Category 2 and 3, all binding and potentially binding clauses, but we also analyze exclusively Category 3 as a robustness check and find similar results. The data set thus measures both the frequency and legally binding character of Council actions in the legal categories we examine. Used in combination, these two indices will allow us to assess whether the evidence is consistent with our theory about the nature of legal change brought about in law related to NIAC. The unit of analysis in the data set is Council resolution, and each resolution features basic variables for the conflict it addresses, including year, duration, the presence of peace agreements, battle-related deaths, and other characteristics. Each resolution is coded for the seven categories of possible legally binding obligations, comprising a total of fifty-one individual variables coded on the 0 to 3 scale, of which twenty-four are used in this analysis. III. ACCOUNTING FOR THE COUNCIL S VIEW OF CONFLICTS The UCDP conflict data, from which the set of conflicts included in our data is drawn, is organized in dyads, meaning each observation represents the conflict between two actors in a given year. For example, the data binding effect. In view of the nature of the powers under Article 25, the question whether they have been in fact exercised is to be determined in each case, having regard to the terms of the resolution to be interpreted, the discussions leading to it, the Charter provisions invoked and, in general, all circumstances that might assist in determining the legal consequences of the resolution of the [Council]. ).

Feb. 2018 APPENDIX II A7 divide the Bosnian War (1992 1995) into nine dyad-years, and a typical observation within this period is the year 1993, in which the Government of Bosnia-Herzegovina was in conflict with Bosnia Serb forces. This format allows for quantitative empirical examination of many conflict dynamics, but it does not reflect how the Council addresses multiparty civil conflicts. For example, the Council considered the Bosnian War as a whole, and within the broader context of the dissolution of Yugoslavia. 21 The Council passed ninety-seven resolutions addressing Bosnia during this period, and all of them target all three main actors government forces, Serb forces, and Croat forces. We thus concluded that coding conflict dyads would diverge in a fundamental way from the Council s more holistic view of the conflicts it addresses. This practice led us to aggregate simultaneously occurring conflict dyads into territorially-based conflicts, to more accurately reflect how the Council addresses conflicts. We list the Bosnia conflict, for example, as a single conflict despite the presence of nine UCDP dyads in the UCDP data set. Simultaneous civil conflict dyads are aggregated into territorially-based conflicts, but international conflicts and different periods of conflict are kept distinct (for example, the Gulf War and the Iraq War are treated as distinct, as are India-Pakistan and India-Naxalites). This method presents the most reasonable approximation of how the Council has regarded armed conflicts in the past, and this is basis of the results reported in our findings. As a robustness check, we repeat our analysis with conflict dyads as the unit of observation and find consistent results. The data features 1071 unique resolutions. There are fifty-three conflicts in the territorial configuration, and sixty-three conflicts in the dyadic configuration. IV. FREQUENCY OF COUNCIL INTERVENTION AS AN INDICATION OF LAW-MAKING To understand how the Council has affected international law, it is necessary to understand how the Council has intervened in conflicts. We perform a descriptive analysis of the data, focusing on key variables of interest, to understand how, where, and when the Council has made binding decisions. We use the three point scale for each variable of interest to calculate relative frequencies of binding decisions in each category and for each conflict. Generating the relative frequencies 21. For a description of the U.N. Protection Force mission to the former Yugoslavia, see Former Yugoslavia UNPROFOR, UNITED NATIONS (last updated Aug. 31, 1996), https://peacekeeping.un.org/mission/past/unprof_p.htm.

A8 APPENDIX II Feb. 2018 involves taking a simple proportion of observations that meet a specified criteria over a relevant number of observations drawn from the population. We investigate both the proportion of observations coded as Category 2 or 3 for a particular variable and solely observations coded as Category 3, for a stricter assessment. We selected the denominators for each proportion based on a principle of applicability. Obligations in resolutions are not relevant for all circumstances and would not plausibly be used by the Council outside relevant contexts. An obligation in a resolution to cooperate with the U.N. mission, for example, is only relevant where a U.N. mission exists. Likewise, an obligation to abide by the terms of a peace agreement is only relevant where a peace agreement has been signed. In assessing the frequency with which modifications to the mandate of a peacekeeping mission are made, we use only those conflicts in which a peacekeeping has already been mandated. We therefore select the applicable group of observations to serve as the denominator for those proportions. For more universal obligations, such as the treatment of detainees and the rights of women and children, all cases were considered relevant. We also investigate additional denominators, theoretically motivated but more expansive, as a robustness check and find consistent results. The observations then feed into two descriptive analyses: resolutionbased and conflict-based. We analyze the resolutions based on geographic, temporal, conflict, and other variables to create a portrait of trends in Council usage of obligations over time and across space. We also use the maximum score of each variable for each conflict to determine how frequently the Council applies obligations across conflicts. This aggregate assessment is applicable because the data constitutes legal impositions that endure over time. Though the Council may alter, escalate, or repeat obligations for a variety of reasons, a single instance of an obligation nonetheless carries the force of a Council resolution. We also categorize the data by an important, theoretically-driven variable: Chapter VII authorities. The Council does not address all conflicts immediately. Because of the escalatory nature of conflict, the responsive nature of the Council, and the presence of other mediation and resolution efforts in the international community, we do not expect the Council to respond to all conflicts with the same vigor. Further, the Council can work cooperatively with parties to conflict, or can overcome lack of cooperation through the use of Chapter VII authority. We therefore employ Chapter VII references as a form of categorization that the Council itself has chosen to use against noncooperative parties. We expect that conflicts for which the Council

Feb. 2018 APPENDIX II A9 refers to Chapter VII authority will be the most likely targets of obligations; we do not necessarily expect the Council to use its most assertive obligations to target cooperative parties. Therefore, we also analyze the proportions of obligations within the set of conflicts with Chapter VII authority invoked at least once during the course of the conflict, with Chapter VII references coded as Category 3 on our scale. Finally, it is important to note that we attempt neither predictive analysis nor multivariate correlation. Council processes are complicated. Predicting even when the Council passes resolutions, let alone when it imposes individual obligations, requires a multi-stage theory of the Council process. We do not attempt to provide such a theory, which would have to account for target conflict dynamics, the political influence of the Council membership, and the deterministic influence of the Council s five permanent members. 22 Instead, we investigate frequency through descriptive statistics, without imposing parametric assumptions or providing theoretical predictors as covariates. We leave a predictive theory and its corresponding empirical analysis for future work. 22. For more on these dynamics, see generally Bruce Bueno de Mesquita & Alastair Smith, The Pernicious Consequences of UN Security Council Membership, 54 J. CONFLICT RESOL. 667 (2010), and Barry O Neill, Power and Satisfaction in the United Nations Security Council, 40 J. CONFLICT RESOL. 219 (1996).