Civil conflict and agenda-setting speed in the United Nations Security Council

The United Nations Security Council (UNSC) can respond to a civil conﬂict only if that conﬂict ﬁrst enters the Council’s agenda. Some conﬂicts reach the Council’s agenda within days after they start, others after years (or even decades), and some never make it. So far, only a few studies have looked at the crucial UNSC agenda-setting stage, and none have examined agenda-setting speed. To ﬁll this important gap, we develop and test a novel theoretical framework that combines insights from realist and constructivist theory with lessons from institutionalist theory and bargaining theory. Applying survival analysis to an original dataset, we show that the parochial interests of the permanent members (P-5) matter, but they do not determine the Council’s agenda-setting speed. Rather, P-5 interests are constrained by normative considerations and concerns for the Council’s organizational mission arising from the severity of a conﬂict (in terms of spillover effects and civilian casualties); by the interests of the widely ignored elected members (E-10); and by the degree of preference heterogeneity among both the P-5 and the E-10. Our ﬁndings contribute to a better understanding of how the United Nations (UN) works, and they have implications for the UN’s legitimacy. positively associated with the existence of direct E-10 members’ interests in that conﬂict.


Introduction
The United Nations Security Council (UNSC) is designed to respond quickly to conflicts and crises. Its actions should be not only effective, but also "prompt" (UN Charter, Art. 24). To that end, the Council must function "continuously," and each Council member must be represented "at all times" at the seat of the organization (Art. 24, 28). The need for the Council to address conflicts early on is regularly emphasized by United Nations (UN) officials and policymakers, while scholars argue that the Council's capacity to act fast is a key feature that makes the institution attractive to UN member states (Abbott and Snidal 1998, 10).
However, for the Council to take (swift) action in a conflict, the conflict must first enter its agenda. Only once the Council members agree to discuss a conflict in a formal Council meeting-and thereby put it on the agenda-can they take specific decisions (in the form of resolutions), such as condemning the use of violence, imposing an arms embargo, or deploying a peacekeeping or peace enforcement operation.
With the end of the Council's Cold War paralysis, some conflicts have entered the Council's agenda with remarkable speed. For instance, Council members convened a formal meeting to discuss the situation in Libya only six days after the popular uprising against long-time ruler al-Qaddafi had turned into armed conflict and the rebels had taken Benghazi on February 26, 2011. The Council was also quick to put the conflicts in Kosovo, South Ossetia, and Mali on its agenda. In other cases, however, the pace of agenda setting in the Council was much slower. For instance, the civil conflict in Uganda entered the Council's agenda only in 2006-more than ten years after the onset of the conflict.
The North-South conflict in Sudan that escalated into a full-blown civil war in the mid-1980s was addressed by the Council only twenty years later (in 2003). Many conflicts never make it onto the Council's agenda.
What explains this variation in the Council's agendasetting speed? Why do some conflicts enter the Council's agenda faster than others? While UNSC decision-makingand intervention resolutions in particular-has received a great deal of academic attention, only a few studies focus on the Council's agenda setting, and so far no study has examined the Council's agenda-setting speed. This is unfortunate because whether, and if so how quickly, a conflict makes it onto the Council's agenda matters. For one, being on the Council's agenda is a necessary (but not sufficient) condition for subsequent Council action. Getting an item on the agenda quickly is a precondition for early action, and early action is widely considered to be more effective in that it is likely to reduce the number of victims and lower the risk of a conflict spilling over to other neighboring countries (Dowty and Loescher 1996, 42;Saleyhan and Gledtisch 2006). Moreover, even in the absence of subsequent action, putting a conflict on the Council's agenda is an important political act that imposes costs on the Council members (Allen and Yuen 2017). Unlike informal Council meetings that are held behind closed doors, in formal Council meetings public records are kept. As such, discussing a conflict in a Council meeting sends an important signal to various relevant audiences that the Council considers the conflict to constitute a threat to international peace and security. 1 Equally important, once the Council members agree to put a conflict on the agenda, it becomes "sticky" and cannot be easily removed from the agenda later on. Not surprisingly, then, Council and non-Council members attach a great deal of importance to UNSC agenda setting. When the Council sought to reform its working methods to make it easier to delete items from its formal agenda (the so-called Summary Statement), this was vociferously resisted by a number of UN members (Sievers and Daws 2014, 230). Finally, unlike Council decisions, agenda setting is covered by procedural vote, meaning that the five permanent members (P-5) have no veto right and cannot formally control the Council's agenda. Many different actors can call for an item to be placed on the agenda of a formal meeting, and once they do, the positive support of any nine members is sufficient.
Recent studies on UN intervention have examined two broad explanatory factors that we argue help explain, but do not fully account for, variation in agenda-setting speed. From the traditional realist perspective on international organizations (IOs), decisions in the UNSC are determined by the parochial interests of the powerful P-5 (de Jonge Oudraat 1996;Boulden 2006;Bosco 2009), whereas recent studies drawing on constructivism point to the importance of humanitarian considerations as well as the desire to fulfil the UN's organizational mission to preserve international peace and security (Gilligan and Stedman 2003;Mullenbach 2005;Fortna 2008;Beardsley and Schmidt 2012;Allen and Yuen 2014).
Building on these insights, this article adds to the existing scholarship in several important ways. First, it advances our understanding of how IOs work by going beyond the narrow focus on Council resolutions to examine the crucial agenda-setting stage. We examine agenda setting in the context of civil conflict because civil conflict presents a major challenge to the Council and makes up the bulk of its work (von Einsiedel, Malone, and Ugarte 2015, 6).
Second, the article proposes a novel theoretical framework that combines insights from realist and constructivist International Relations (IR) theory with lessons from institutionalist theory and bargaining theory-especially work on European Union (EU) decision-making-to focus on four drivers of agenda setting: (1) P-5 interests; (2) normative considerations and the UN's organizational mission; (3) the interests of the Council's ten elected members (E-10); and (4) the distribution of preferences in the Council. We argue and show that while the P-5 have no formal veto in the agenda-setting process, their parochial interests matter, and they matter strongly. However, they do not determine the Council's agenda-setting speed. Rather, P-5 interests are constrained by normative considerations and concerns for the Council's organizational mission arising from the severity of a conflict (in terms of spillover effects and civilian casualties); by the interests of the widely ignored E-10; and by the degree of preference heterogeneity among both the P-5 and the E-10. Overall, in terms of the relative explanatory power of different factors, we find that the effect of the E-10 on agenda-setting speed can be greater than that of the P-5, and that parochial interests and general preference heterogeneity appear to play a bigger role in UNSC agenda setting than do conflict characteristics and the UN's organizational mission.
Third, the article makes a methodological contribution by providing the first survival analysis (also known as "event history analysis") of UNSC agenda setting, which reveals insights about the speed with which the Council becomes involved in conflicts as well as the likelihood that, at any given point in time, a particular conflict will reach the Council's agenda at all. We apply survival analysis to an original dataset that includes information on the time that elapses between the onset of 276 episodes of civil conflict and the time those episodes were included on the Council's agenda.
Fourth, our findings also have implications for the Council's legitimacy. 2 For the Council's performance 2 Legitimacy denotes the right to rule, the normatively appropriate exercise of political authority. We use it here in its normative sense, meaning that an actor legitimacy-its ability to fulfil its mandate and to collectively solve problems (Scharpf 1999;Gutner and Thompson 2010)-timing is crucial because it arguably matters whether the Council addresses a deadly conflict after six days or only after six months (or never). Our finding that agenda-setting speed is positively associated with conflict severity supports the UNSC's performance legitimacy. By contrast, our findings have mixed implications for the Council's procedural legitimacy-fair procedures that reflect a broad range of UN member state interests (Tyler 2006;Binder and Heupel 2015). The finding that parochial P-5 interests skew agenda-setting speed undermines the Council's procedural legitimacy, whereas the finding that speed also depends on the interests of the E-10 increases the inclusiveness of the Council's agenda-setting process, thus improving procedural legitimacy.
We structure our analysis as follows. After a review of previous research on the UNSC to identify gaps in the literature, we introduce our theory and hypotheses about the determinants of UNSC agenda-setting speed. This is followed by a discussion of our data and methods. The final section presents the results and highlights the implications of our findings.

Prior Research
The standard realist informed view has been that Council decisions are largely determined by the parochial interests of the P-5 and that the E-10 are inconsequential (de Jonge Oudraat 1996;O'Neill 1996;Bosco 2009;Hosli et al. 2011). Yet recent research on the UNSC has begun to question the explanatory power of P-5 interests and the irrelevance of the E-10. Drawing on constructivist IR theory, many studies find that normative considerations and the organization's mission to address threats to international peace and security also matter, perhaps even more so than P-5 interests (Mullenbach 2005;Fortna 2008;Beardsley and Schmidt 2012;Binder 2017).
Only a few studies have focused on the Council's agendasetting process, but their results are inconclusive, and none have examined agenda-setting speed. Again, the predominant realist view has been that, just like any subsequent decisions, agenda setting in the Council is largely controlled by the P-5. Boulden (2006, 412) claims that "the simple existence of the veto, even if not formally exercised, gives the permanent members the ability to control the Council's agenda." Iwanami (2011) focuses on civil wars between 1946 and 1999 and finds support for the P-5 interests explanation. Her results suggest that civil wars that involve a P-5 ally or occur in a former P-5 colony are less likely to enter the Council's agenda, whereas those where a P-5 member has previously intervened are more likely to do so. Iwanami does not find support for the constructivist view, as the battle deaths variable tests out as insignificant. Frederking and Patane (2017) examine which conflicts "dominate" the UNSC's agenda (2017, 349) by counting the number of meetings per conflict each year over the period 1991-2013. Their findings challenge the realist view in that they find severe conflicts to be significantly more likely to dominate the Council's agenda, whereas P-5 interests in a conflict, as or institution has legitimacy if it conforms to external standards usually derived from democratic theory. We focus here on two such standards that are widely considered to be key to the legitimacy of IOs: the quality of their procedures and their performance (Scharpf 1999;Binder and Heupel 2015;Dellmuth and Tallberg 2015). But they are not the only ones. Others discuss various additional standards of IO legitimacy such as the provision of expertise or respect for human rights (e.g., Buchanan and Keohane 2006). measured by arms transfers and trade, test out as insignificant. Finally, Allen and Yuen (2017) focus on the time it takes the Council to act on a broad range of items that are already on its agenda, but they suspect that P-5 members keep issues directly tied to their interests off the agenda in the first place.
Overall, research on agenda setting in the Council is sparse and has produced inconclusive findings, and thus far no study has examined agenda-setting speed. To begin to fill this gap, we propose a theoretical framework for understanding UNSC agenda-setting speed that builds on insights from the realist and constructivist inspired literature on UN intervention and agenda setting. Yet because the P-5 have no formal veto at the agenda-setting stage and thus need to reach out to the E-10, and given the importance of preference heterogeneity for cooperation, we go beyond the dominant approaches in the study of the UN to also draw on institutionalist and bargaining theory.

Theories and Hypotheses
The Agenda-Setting Process How do civil conflicts enter the Council's agenda? According to the UN Charter, meetings will "be held at the call of the [Council] President at any time he deems necessary," and in practice the process is usually routine, since "Council members are planning in advance for each upcoming presidency," and most agenda items simply roll over from previous meetings (Sievers and Daws 2014, 193). But not all Council business is routine. When the Council is requested to take up a new item, such as addressing a civil conflict, the Council President needs to make available additional meeting time, either by scheduling a new meeting or by altering the agenda of an already planned upcoming meeting. To model UN Security Council agenda setting, we therefore need to focus attention on two things: decisionmaking rules and the preferences of relevant actors. For a start, we must identify which actors have the authority to set and modify the agenda, and any procedural rules that constrain that authority. A wide range of actors can bring a situation to the Council's attention. These include Council members themselves, other UN member states (UN Charter, Art. 35(1)), non-UN member states (Art. 35(2)), UN bodies (Art. 11(2)), as well as the UN Secretary General (Art. 99). If the Council President-rotating in English alphabetical order every calendar month-receives such a request by any of these actors, they shall call a meeting within fourteen days, in consultation with the other Council members. The Council President has some discretion here, since they can refrain from calling a meeting if they sense a lack of agreement among Council members (Sievers and Daws 2014, 196-7). In practice, if only some Council members agree on discussing an item, the Council President usually proposes to place the item on the provisional agenda of a new meeting or an upcoming meeting, and those members who disagree have the right to request a procedural vote. To adopt the agenda then requires nine positive votes without a right to veto for the P-5. 3 If Council members adopt the provisional agenda and discuss a matter during a formal meeting, it is then listed in the Summary Statement of matters of which the Security Council is seized. 4 Inevitably, the level of disagreement among Council members about whether to place a conflict on the agenda will vary across conflicts. Because we lack information on the precise preference location for each actor on each conflict, it is not possible to deploy a range of powerful analytical tools from procedural and bargaining models to identify the pivotal state(s), measure the precise size of the core and the winset, or compute an expected bargaining outcome (Tsebelis 2002;Thomson 2011;Drüner et al. 2018). However, one thing is certain: in the absence of agreement, the reversion point is the status quo where the conflict is not on the agenda.
The key theoretical question, therefore, is which factors tend to increase the likelihood that at any given point in time (at least) a nine-vote coalition will form in favor of changing the status quo and placing an item on the agenda. We develop hypotheses about four types of factors: (1) interests of the P-5; (2) normative considerations and the UN's organizational mission; (3) interests of the E-10; and (4) the level of general preference heterogeneity in the Council, among the P-5 as well as among the E-10.

P-5 Interests
According to the standard realist view, IOs are nothing more than tools that serve powerful member states to advance their parochial interests (Mearsheimer 1994). Scholars argue that the possession of the veto over subsequent resolutions and (perhaps even more so) their economic and military power enable the P-5 to dominate the agenda-setting process. The P-5 can exercise their veto to prevent any Council decision they oppose (de Jonge Oudraat 1996;O'Neill 1996;Bosco 2009;Hosli et al. 2011), or they can deploy their vast resources to bribe the E-10 with side payments (Kuziemko and Werker 2006;Dreher, Sturm, andVreeland 2009a, 2009b; Allen and Yuen 2017, 12), or they can coerce them with threats to act outside the Council (Voeten 2001). According to Keating (2015, 146), the P-5 decide "what will be discussed and when. It is now common place for P-5 members…to control the Council and exclude discussion of items they find inconvenient." While many analysts agree that the P-5 dominate UNSC decision-making, including at the agenda-setting stage, they disagree over the causal direction of those interests. Some argue that civil conflicts are less likely to enter the agenda if they involve the interests of the P-5. On this view, the permanent members have an incentive to bypass the UN as they wish to exclude the interference of other members from areas under their spheres of influence (de Jonge Oudraat 1996;Boulden 2006;Fortna 2008). Others, by contrast, argue that conflicts that involve the interest of any permanent member are more likely to enter the agenda. This is because UN involvement provides burden-sharing, which reduces the costs of the permanent member trying to resolve the conflict themselves, and also provides legitimacy for new or ongoing intervention (Voeten 2001;Wallensteen 2002;Hurd 2007). These contrasting logics lead us to expect that civil conflicts that directly involve the parochial interest of one or more P-5 members will make it either harder or easier to build a nine-vote coalition and place a conflict on the agenda.

H1a/b: The speed with which a conflict enters the UNSC agenda is negatively/positively associated with the existence of direct P-5 member interests in that conflict. Normative Considerations and Organizational Mission
Drawing on constructivist IO scholarship, we expect that the speed with which a conflict reaches the Council's agenda is driven by humanitarian considerations and emerging norms of humanitarian intervention (Sandholtz 2002;Finnemore 2003;Carpenter 2005). In line with this theory, scholars have pointed to conflict severity-in terms of human suffering and destabilizing spillover effects resulting from a conflict-as a prominent explanation for UN action (de Jonge Oudraat 1996; Gilligan and Stedman 2003;Beardsley and Schmidt 2012;Hultman 2013;Binder 2017). Conflict severity operates through two distinct logics. First, if many people suffer and die in a conflict, this generates normative pressure on Council members to take rapid action to respond to that conflict. Observers of the Council point to a humanitarian "imperative" (Luck 2006, 92) that Council members experience when confronted with large-scale suffering and the violation of international humanitarian norms. Scholars argue that international societal actors, often in conjunction with the media, nongovernmental organizations, and domestic civil society organizations, play an important role in this process in that they mobilize around these norms to generate momentum for responding to situations of human suffering (Western 2002), in particular among civilians (Hultman 2013). This leads us to expect that severe civil conflicts generate more normative pressure on Council members, thereby making it easier to build a nine-vote coalition. H2a: The greater the number of civilian casualties in a civil conflict, the higher the speed with which it enters the Council's agenda.
Second, severe civil conflicts also tend to create significant negative externalities, in that they diffuse into wider regional wars and refugees spill over to destabilize neighboring countries (Salehyan and Gleditsch 2006;Kathman 2010). This poses a threat to the Council's core organizational mission-the maintenance of international peace and stability (Beardsley and Schmidt 2012). Not living up to the UN's key organizational mandate risks undermining the Council's legitimacy and, by extension, the support it receives from the UN membership. It also disincentivizes states from contributing to an organization that routinely fails to deliver the public good it was created to produce (Beardsley and Schmidt 2012, 39). This leads us to expect that strong spillover effects on neighboring countries, and the resulting threat to the Council's organizational mission, will tend to make it easier to build a nine-vote coalition. H2b: The stronger the spillover effects a civil conflict generates for neighboring countries or regions, the faster the conflict will enter the Council's agenda.

E-10 Interests
Several strands of institutionalist theory provide strong theoretical reasons to believe that the elected Council members are more than just placeholders and can have real influence on the work of the Council, including its agenda-setting process. 5 We theorize that E-10 influence on agenda-setting speed stems not only from procedural rules that make at least four of their members crucial for any nine-vote coalition, but also from the E-10's informal power resources, as well as from constraints on the P-5's subsequent veto. Rationalist institutionalist theory identifies procedural rules as a nonmaterial, formal source of small state power (Keohane and Nye 1977;Abbott and Snidal 1998). Procedural rules have enabled small states to play a significant role in shaping IO activity, for example multilateral development bank behavior (Lyne, Nielson, and Tierney 2006). In the UNSC, procedural rules have enabled the EU members of the E-10 to engage in issue linkage across IOs and thereby impact the Council's substantive work, promoting the security interests of other European states (Mikulaschek 2018). Since UNSC agenda setting is governed by the procedural vote, formally any nine members can place an item on the agenda. Of course, P-5 members might still try to dominate the agenda by buying E-10 votes, but that strategy faces serious limitations. Especially if the P-5 do not all agree about an agenda item, this would involve one or two permanent members trying to buy off up to seven elected members while facing potential outbidding from opposing P-5 members.
Institutionalists have also highlighted important informal constraints on the use of the P-5 veto right, including the threat to use it. Mikulaschek (2016) argues that great powers exercise informal strategic restraint in order to generate consensus and enhance compliance. Similarly, Krisch (2008, 139) contends that the P-5 strive for Council unanimity because this signals to a wide audience of states that UNSC decisions are desirable and appropriate. We contend that these informal constraints give elected members leverage beyond their formal voting power at the agenda-setting stage since E-10 efforts to shape the agenda are at least to some extent conditioned on anticipation of subsequent decisions.
Yet other variants of institutionalist theory posit that IOs are arenas for deliberation and public justification (Chayes and Chayes 1995, 125) in which the power of the better argument carries the day, so that small states can exploit a range of informal power resources to secure their objectives. Johnstone (2003, 452, 461) conceives of the UNSC as a "forum for justificatory discourse" to emphasize how, regardless of the P-5's formal veto power, processes of legal discourse, deliberation, and persuasion enable elected members to influence decisions. Accordingly, the E-10 can impact Council discussions with their regional and local knowledge about civil crises, as well as with the expertise and authority they gain by chairing the Council's subsidiary organs, 6 and they might also amplify their influence by focusing on a narrow range of issues. New Zealand is a prominent example of an influential elected member, but other elected members, including small countries such as Austria, Costa Rica, Denmark, Guatemala, or Uganda, have demonstrated that despite the P-5's power "it is nevertheless still possible to assert a leadership role in the Council" (Keating 2015, 152) by exploiting informal power resources such as technical expertise or diplomatic skill. Lessons from decision-making in the EU are particularly informative here. Just like the P-5 in the Council, the "big four" states of France, Germany, Italy, and the United Kingdom have a disproportionate amount of resources and power in the EU, yet they do not dominate EU decision-making, whether it be historic treaties or daily legislation, and there is no evidence that they manage to buy the the matter on the agenda. Likewise, in September 2006, China and Russia voted against the inclusion on the Council's agenda of the situation in Myanmar, yet the meeting was eventually held and the conflict was discussed. 6 Interview with former non-permanent Council member on October 11,

2018.
Downloaded from https://academic.oup.com/isq/advance-article-abstract/doi/10.1093/isq/sqaa017/5808900 by University of Reading user on 25 March 2020 votes of smaller states (Finke 2009;Golub 2012aGolub , 2012b. Informal power resources enable small states to punch well above their weight: they possess technical expertise in particular areas, and they focus on fewer issues so that when they do take a stand, their views have greater impact. Many studies have found that in terms of bargaining success over EU legislative outcomes, small states do just as well as, or even better than, the "big four" (Thomson 2011;Cross 2013). If these various mechanisms for E-10 influence operate, what are the observable implications for agenda-setting speed? The E-10 often champion thematic issues such as "women, peace, and security" or climate change (Allen and Yuen 2017, 11), but there is evidence that they use their term also to advance their national interests. For instance, Australia was successful in putting the downing of Malaysian Airlines Flight 17 (MH17) over eastern Ukraine on the Council's agenda. For Langmore and Farrall (2016, 65), this represents "a clear example of an elected member successfully influencing a Council outcome that served its own national interests." Likewise, Mikulaschek (2016) shows how elected African members use their influence to successfully push the Council toward deploying peacekeepers to civil conflicts in Africa. We expect that the E-10 use their influence to advance their conflict-specific parochial interests, and just as with the P-5, the direction of the effect of parochial interests could plausibly cut either way. While a negative effect is possible, it is arguably less likely than for the P-5, since the E-10 typically do not have spheres of influence from which they seek to exclude UN interference. In our view, it is more likely that E-10 states with a direct interest in a conflict will view UN involvement as positive, a source of burden-sharing that reduces the costs of them trying to resolve the conflict themselves.

H3:
The speed with which a conflict enters the UNSC agenda is positively associated with the existence of direct E-10 members' interests in that conflict.

General Preference Heterogeneity
Finally, we draw on bargaining theory to formulate hypotheses about how the distribution of preferences in the Council affects agenda-setting speed. Bargaining theorists have established that preference heterogeneity among decisionmakers increases distributional conflict and complicates cooperative efforts (Schneider and Urpelainen 2014). Heterogeneity increases the time needed to reach agreement and depart from the status quo, because in a world of imperfect information, it raises the transaction costs of identifying each actor's interests, as well as feasible log-rolls, package deals, and side payments (Sommerer and Tallberg 2016). This logic informs studies that predict that heterogeneity in the Council of Ministers will slow down EU decision-making (Klüver and Sagarzazu 2013;Drüner et al. 2018), that heterogeneity among political parties slows down coalition formation (Martin and Vanberg 2003), and that actor heterogeneity has a negative effect on the decision-making capacity of IOs more generally (Sommerer and Tallberg 2016).
Although they do not consider the issue of speed, recent studies have argued that UNSC decision-making depends not just on the parochial interests of the permanent members, but also on the extent to which their more general interests overlap. Allen and Yuen (2014) argue that a more heterogeneous P-5 makes it harder for the permanent members to agree on the mandate of a mission, resulting in less-specific mandates and more frequent mandate reviews. Similarly, Beardsley and Schmidt (2012, 45) argue that the UN is less able to act in international crises when there is greater P-5 preference heterogeneity, "when there is starker divergence between the two most divergent P-5 states." A straightforward extension of Beardsley and Schmidt's argument that follows from bargaining theory is that just as P-5 preference heterogeneity makes it harder for the UNSC to alter the status quo of "no involvement," it makes it harder and more time consuming for the UNSC to alter the status quo of a conflict not being on the agenda.

H4a:
A higher level of P-5 preference heterogeneity in the Council results in conflicts entering the agenda more slowly.
Formally, any nine members can place an item on the Council's agenda. More likely, though, is a scenario whereby a small number of P-5 states, typically the P-3 (France, the United Kingdom, and the United States), perhaps even just one, lead a winning coalition of nonelected members. But the impetus could also originate within the E-10 as in 1994 when New Zealand pushed to get a UNSC response to the genocide in Rwanda (Keating 2015), or in 2004 when Germany strongly advocated for a response in the Darfur conflict (Nitzschke 2015). Australia has played a leading role on numerous occasions, often despite the reluctance of various P-5 members, for example in regard to Afghanistan, humanitarian relief for Syria, and as mentioned above, the shooting down of Malaysian Airlines Flight MH17 (Langmore and Farrall 2016). Either way, we contend that the importance of heterogeneity should extend beyond just the P-5. If the E-10 members are trying to decide among themselves whether to place an item on the agenda, their heterogeneity should increase transaction costs and slow down agreement. And if one or several P-5 members are trying to build a winning coalition with the E-10, reaching out to craft a dealor more likely multiple deals-with a highly divided group will take longer than when that group holds more uniform views.
H4b: A higher level of E-10 preference heterogeneity in the Council results in conflicts entering the agenda more slowly.

Research Design and Data
Given that the UNSC was virtually paralyzed during the Cold War, and that the fundamental dynamics of the Council changed dramatically thereafter, our hypotheses only apply to the situation after 1989. Our central puzzle is to explain variation in how long it takes for civil conflicts to reach the Council's agenda in the post-Cold War era. The set of conflicts come from the UCDP/PRIO (Uppsala Conflict Data Program/Peace Research Institute Oslo) Armed Conflict dataset. We focus on internal and internationalized internal conflicts, rather than on extrasystemic or interstate armed conflicts. We exclude conflicts when they involve only low activity, defined by UCDP criteria as fewer than twenty-five fatalities per year. A given conflict can last many years and have multiple episodes, reaching the twenty-five fatalities threshold, then subsiding back to low activity before eventually flaring up again. Our unit of observation is therefore the conflict-episode-day.
To analyze the determinants of agenda-setting speed, we employ survival analysis. Each conflict episode that begins after January 1, 1990 becomes at risk of being placed on the agenda once it rises above the level of low activity. Conflict episodes that began before the end of 1989 and continue into the post-Cold War era are coded as becoming at risk as of January 1, 1990 (or later if they have not already reached the "active" threshold). A conflict episode "survives" until it reaches the agenda (the "event") or until it is right-censored. Right-censoring occurs when the conflict dies back down to low activity, or when the conflict terminates, or on January 1, 2017 when we stopped tracking the Council's agenda.
The quantity of interest in survival analysis is the hazard rate, and both speed and outcome are intrinsically associated with this rate. Indeed, unlike models that ignore duration, the extra information about timing exploited by the survival approach is valuable precisely because it yields more accurate inferences about the pattern of eventual outcomes (Beck, Katz and Tucker 1998;Box-Steffensmeier and Jones 2004, 1). Thus, we can explain not only the relative speed with which conflicts reach the agenda, but also whether conflicts with particular covariate profiles are likely ever to reach the agenda. Because we have no strong theoretical predictions about the effect of time on agenda settingconflict episodes might be more likely to reach the agenda as soon as the episode starts, or perhaps pressure tends to build steadily, or maybe it follows a more complex pattern with waves of support-we fit Cox models. An important advantage of Cox models is that they do not require any assumptions about the shape of the baseline hazard (Box-Steffensmeier and Jones 2004, 47).

Dependent Variable
Our analysis uses original data to construct the dependent variable. For each conflict episode, we coded the time elapsed between when it became "active" (start_date2 in the UCDP/PRIO Armed Conflict dataset) and when it first entered the Council's agenda or was right-censored. The "active" threshold is widely used in published studies to distinguish signal from noise. Besides convention, we adopt this threshold also because we assume that given the scarcity of UNSC agenda time, only active episodes are at risk of reaching the agenda. We do not adopt a higher threshold because we assume that the UNSC often deems it important to discuss active civil conflicts before they escalate, potentially into full-blown wars (i.e., reach the 1,000 deaths in a given year threshold). 7 A conflict reaches the agenda if the Council agrees to hold a formal meeting to discuss that conflict. Information about formal Council meetings (day/month/year) and the agenda item discussed in each meeting (there is only one item per meeting) is available through UN websites. Specifically, we cross-checked two sources of information: the United Nations Digital Library and The Repertoire of the Practice of the Security Council. 8 The advantage of the Repertoire is that, in addition to the precise date a meeting was first convened by the Council, it also provides a more detailed description of each agenda item. Thus, it is possible to link the item that is being discussed at a meeting to a specific conflict episode. 9 The resulting dataset contains 145 conflicts and 276 conflict episodes (see online appendix Table S1). Just under 30 percent of the 7 In their analysis of which conflicts dominate the UNSC agenda, Frederking and Patane (2017) exclude all conflicts that never reached 1,000 deaths in a single year. In our view, this discards crucial information and also creates a hindsight bias. Nevertheless, we reran our analyses after dropping conflicts that never became wars. The results, which we report in the supplementary online appendix, confirm or even strengthen most of our original findings. 8 https://digitallibrary.un.org/ and https://www.un.org/securitycouncil/ content/repertoire/agenda-items-overview. 9 This is important because the formulation of agenda item titles is sometimes (deliberately) vague (e.g., "The situation in Liberia" or "Letter dated 10 August 1992 from the Permanent Representative of Bosnia and Herzegovina to the United Nations addressed to the President of the Security Council").
conflict episodes reach the UNSC agenda; the others are right-censored. On average, active conflict episodes survive for just over three years before reaching the UNSC agenda, although many are taken up immediately and some survive for over twenty-seven years. Figure 1 displays this variation in the form of a kernel density plot. Note that cases that are still ongoing but right-censored on January 1, 2017 are treated as having reached the agenda, which underestimates their survival time.

Independent Variables
To test our hypotheses about the effect of P-5 and E-10 interests, we obtained information about four different types of direct link between each Council member and the country experiencing the conflict. First, a state will have a stronger direct interest in a conflict when it has a formal alliance tie with the country experiencing the conflict (Mullenbach 2005, 452;Beardsley and Schmidt 2012, 40). We used Alliance Treaty Obligations and Provisions (ATOP) data (Leeds et al. 2002) to generate yearly ordinal measures of the total number of P-5 and the total number of E-10 states with alliance ties to the conflict state. For each of these variables, to improve tractability, we then aggregated categories with a small number of cases to produce recoded ordinal variables with either three or four categories.
Second, any state is likely to have a stronger direct interest in conflicts on its own doorstep. This might be because it views a nearby conflict as being within its sphere of influence (Fortna 2008;Beardsley and Schmidt 2012, 40) or because proximity increases the fear of contagion. We used version 3.2 of the Correlates of War (COW) Direct Contiguity dataset to code yearly dummies for whether each UNSC member and the conflict state were directly contiguous. We then aggregated these into two yearly ordinal measures: the total number of contiguous P-5 states and the total number of contiguous E-10 states. Due to small numbers in the highest categories we then recoded these into a P-5 contiguity dummy and an E-10 contiguity dummy.
Following Frederking and Patane (2017), we anticipate that the P-5 will have a stronger direct interest in conflicts when they transfer arms to or conduct trade with the conflict state. But given our expectation that E-10 states also matter to the speed of Council agenda setting, we extend the same reasoning to E-10 arms transfers and trade. We used the SIPRI Arms Transfers Database to compile yearly arms transfers from each UNSC member to each conflict state. We then aggregated these into two variables: the total yearly volume of arms transfers to the conflict state from P-5 states and the total yearly volume of arms transfers to the conflict state from E-10 states. We used the COW Trade Dataset 4.0 (Barbieri and Keshk 2016) to calculate total yearly trade between the conflict state and the P-5 states, and between the conflict state and the E-10 states.
To test our hypotheses about how conflict severity affects agenda-setting speed, we measured the level of human suffering and spillover effects. The extent of human suffering in each conflict in any given year is coded as the number of civilian deaths from version 18.1 of the UCDP Georeferenced Event Dataset (Sundberg and Melander 2013). To measure the spillover effects of each conflict in any given year, we used the number of refugees from the UNHCR Statistical Database 1977-2016 (http:// data.un.org/Data.aspx?d=UNHCR&f=indID%3AType-Ref).
To test our hypotheses about the effect of general preference heterogeneity, we used ideal point data based on states' voting behavior in the UN General Assembly (Bailey, Strezhnev, and Voeten 2017). We constructed two separate measures of the range of preferences within the Council each year, one for the P-5 and one for the E-10.

Results
Our estimation strategy is to keep each model as simple as possible, avoiding a "garbage can" or "kitchen sink" approach, and then explore the robustness of the findings. We believe this is a wise strategy since it is unlikely that we have managed to identify the complete and true model, and thus adding more variables can do more harm than good (Clarke 2005). We also take steps to avoid post-treatment bias, caused by controlling away for the consequences of a covariate and "soaking up" some of its effects. The most likely culprits here are the four different types of parochial interests: alliance ties, arms transfers, trade, and contiguity. States that are physically closer to each other are more likely to form alliances, to trade, and to have more arms transfers with each other (Gibler and Wolford 2006;Disdier and Head 2008;Ackerman and Seim 2014). We therefore fit four main models, one for each type of parochial interest. The two trade variables are highly correlated (r = 0.88) so we entered them separately. The two measures of conflict severity present another potential source of post-treatment bias, since the number of cross-border migrants is arguably at least partly a function of the number of civilian casualties. However, recent research suggests that these two variables are actually unrelated, and since the weakness of a correlation is borne out in our dataset (r = 0.09) we include both of them simultaneously in our models. Table 1 presents the results of our main Cox models. As recommended by Box-Steffensmeier and Jones (2004, 137), we tested each covariate for violations in the proportional hazards assumption and accommodated violations by adding an interaction term of the form B*ln(t). Positive coefficients indicate a higher hazard rate and thus the likelihood that conflict episodes will reach the agenda more quickly, whereas negative coefficients indicate a lower hazard rate and slower agenda-setting speed. The hazard ratio directly compares the hazard rate for two cases with different covariate profiles. Profiles associated with extremely low hazard ratios are far less likely to ever reach the agenda. For covariates that also have a B*ln(t) term, the overall effect of the covariate changes over time and one cannot simply interpret the constituent terms-one needs to examine the combined effect and the corresponding confidence interval (Licht 2011).
We hypothesized that when one or more P-5 members had a parochial interest in a conflict that this would significantly affect the speed with which the conflict entered the Council's agenda, but that the direction of the effect could cut either way depending on whether P-5 parochial interests reflected a concern to keep the UN out of their spheres of influence or a desire to involve the UN as a burden-sharing device. The results strongly support the conclusion that P-5 parochial interests matter but that they cut only one way, to keep the UN out by slowing down the rate at which conflicts reach the UNSC agenda. Model 1 examines the effect of alliance ties on UNSC agenda-setting speed. When only one P-5 member has an alliance with the conflict state (which occurs in 22 percent of the observations), the hazard ratio is time-dependent. The effect of one P-5 ally is negative but only materializes after about six months and then intensifies very slowly. For the first 190 days of a conflict, the effect is insignificant, after which it reduces the hazard ratio by 63 percent. When more than one P-5 member has an alliance tie with the conflict state (49 percent of the observations) the effect is immediate and even stronger, reducing the hazard rate by about 70 percent.
Model 2 examines the effect of contiguity on UNSC agenda-setting speed. When a P-5 member is contiguous with a conflict state (54 percent of the time), it reduces the hazard ratio by 80 percent. Model 3 examines the effect of arms transfers. A one standard deviation increase in P-5 arms transfers to a conflict state reduces the hazard ratio by 89 percent. Model 4 shows that a one standard deviation increase in P-5 trade with the conflict state has no effect on  Notes: Robust standard errors in parentheses, clustered by conflict. *** p < 0.01, ** p < 0.05, * p < 0.1. the hazard ratio for the first week, but reduces it by 78 percent after a month, by 88 percent after three months, and by 94 percent after a year. Our findings are consistent with the widely held realist belief that P-5 parochial interests are key drivers of UNSC decision-making, including at the agenda-setting stage where they have no formal veto power. If P-5 interests are at stake in a conflict, this significantly reduces the speed with which the civil conflict makes it onto the Council's agenda. The negative effect of alliances supports Iwanami's (2011) study on UNSC agenda setting, but unlike Iwanami, we also find support for the negative effect that P-5 contiguity has on agenda-setting speed. At the same time, our findings run counter to Frederking and Patane's (2017) surprising finding that P-5 interests (in terms of arms transfers and trade) have no significant effect on agenda setting in the Council. Finally, our results are in line with recent studies on UN intervention and peacekeeping (Fortna 2008;Beardsley and Schmidt 2012), suggesting that the P-5 seek to keep the UN out of their respective spheres of influence, rather than involving it in burden-sharing.
We also hypothesized that more severe conflicts would reach the UNSC agenda more quickly. The results mostly support our expectations. For the civilian deaths variable, across all five models, the sign is in the expected direction and the estimated coefficient is statistically significant in four of the five models. A one standard deviation increase in the number of civilian deaths (212) raises the hazard ratio by about 67 percent. This aligns with constructivist arguments suggesting that Council members respond to a humanitarian "imperative" in situations of large-scale human suffering (Gilligan and Stedman 2003;Mullenbach 2005;Hultman 2013).
For the migrants variable, the results suggest that the UN's organizational mission to prevent regional destabilization matters. After a brief delay, an increase in cross-border migration speeds up the process of getting a conflict onto the agenda. In most of the models, the effect of this variable is time-dependent. A one standard deviation increase in migrants (256,000) has an insignificant effect for the first few weeks, after which it increases the hazard ratio by about 17 percent, then by 30-40 percent after three months, and by 37-58 percent after one year. This lends support to the claim that spillovers have become a ground for international action, and it is consistent with both Iwanami and Frederking and Patane. However, the Council's organizational mission seems to matter less than its humanitarian mandate, in that civilian deaths have a larger and more immediate effect than migrants on agenda-setting speed. Furthermore, while our findings suggest that both P-5 interests and conflict severity matter, unlike some UN intervention studies (Beardsley and Schmidt 2012) we do not find that agenda setting is better explained by normative considerations arising from conflict severity or the UN's organizational mission. To the contrary, at the agenda-setting stage, we find the effect of parochial interests to be much stronger than humanitarian or mission concerns.
Drawing on insights from institutionalist theory, we hypothesized that the E-10 matter, and that UNSC agendasetting speed could not be explained simply by looking at P-5 interests and the level of human suffering. We also hypothesized that the parochial interests of E-10 members would have similar effects as P-5 parochial interests, but that the direction of the effect was more likely to be positive (burden-sharing objectives) than negative (spheres of influence concerns). The results strongly support our main claim that E-10 states matter, but there is evidence that the effect of their parochial interests cuts both ways.
Turning again to Model 1, the effect of E-10 alliances on agenda-setting speed is the opposite as for P-5 alliances and fits the burden-sharing perspective. As with the P-5, the effect of alliance ties depends on the number of allies and is often time-dependent, but the more important lesson is that overall the effect of E-10 alliances is arguably greater than that of P-5 alliances. When only one or two E-10 members have an alliance tie with the conflict state (25 percent and 24 percent of the time), this has an insignificant effect on the hazard ratio. But when three E-10 members have alliance ties with the conflict state (17 percent of the observations), the effect materializes after just twenty-one days, at which point the hazard ratio increases by 136 percent, then by 182 percent at three months, and by 236 percent at one year. With four or more E-10 allies (just over 7 percent of the observations), the hazard ratio increases by 214 percent and the effect is immediate.
In Model 2, when an E-10 member is contiguous with the target state (49 percent of the time) this reduces the hazard ratio by 40 percent. This is yet further evidence that E-10 states matter, but we were somewhat surprised that the direction of the effect was the same as for the P-5, as it suggests that E-10 states close to a civil conflict want to keep the UN out of their backyard. Perhaps E-10 members, too, consider that they have spheres of influence, just with smaller radii. That India wants to keep the conflict in Sri Lanka, and even more so the Kashmir conflict, off the UN agenda illustrates this point. In Model 3, the sign on the E-10 arms variable is negative, pointing in the same direction as P-5 arms transfers, but the coefficient is statistically insignificant. In Model 4, a one standard deviation increase in E-10 trade with the conflict state has no effect on the hazard ratio for the first week, but reduces it by 89 percent after a month, and by 95 percent after three months. As with contiguity, for trade the direction of the effect was the same for the E-10 as for the P-5, reinforcing our finding that E-10 members, too, have perceived spheres of influence.
Our findings challenge the conventional wisdom that the E-10 have little to no influence in the Council; or that they entirely trade away their influence in exchange for bribes. Rather, our results suggest that, in terms of agenda-setting speed, the E-10 have considerable influence they use to advance their interests. This stands in sharp contrast to almost all other systematic empirical studies on UN agenda setting, peacekeeping, and intervention that have ignored the role of the E-10 (the exception being Mikulaschek 2016).
Finally, the distribution of preferences in the Council has so far not attracted much scholarly attention. We drew on bargaining theory to hypothesize that general preference heterogeneity among both the P-5 and the E-10, measured by the range of UN ideal points, would make reaching agreements more difficult and increase the time needed to place a conflict on the Council's agenda. Results from all five models support our hypotheses, but the preference heterogeneity of the E-10 appears to matter even more than that of the P-5, the effect of which wears off over time in all but one of the models. In Model 1, a one standard deviation increase (0.36) in P-5 heterogeneity reduces the hazard rate by 33-60 percent during the first eighteen days of a conflict but is statistically insignificant thereafter. In Models 2 and 3, it reduces the hazard rate by 53-71 percent during the first week, by about 40 percent after one month, by 30 percent after two months, and the effect wears off entirely after three-and-a-half months. In Model 5, it wears off entirely after just forty-six days. For the E-10, by contrast, in most models a one standard deviation increase (0.35) in preference heterogeneity has an immediate and time-constant effect, reducing the hazard ratio by 35 percent. In Models 4 and 5, the effect of E-10 heterogeneity grows over time, reducing the hazard ratio by about 38 percent during the first week, then by 40-42 percent thereafter. That the E-10 have more actors, which likely requires higher transaction costs to form large coalitions, would explain why heterogeneity among this group has a stronger effect on agenda-setting speed than does P-5 heterogeneity.
To test the robustness of our findings, we fitted a number of additional models that included potential omitted variables or altered the operationalization of our main independent variables. First, we considered two additional characteristics of the conflict state: its geographic location and its regime type. Work on UN intervention has found that the deployment of peacekeepers is biased in favor of certain regions of the world (Mullenbach 2005;Fortna 2008;Beardsley and Schmidt 2012). And while the argument is theoretically ambiguous in terms of the direction of the effect (Gilligan and Stedman 2003, 40), some have claimed that UN intervention depends on whether or not the target state is a democracy (Fortna 2008). We coded the location of the conflict state using the set of "region" dummies from the UCDP/PRIO Armed Conflict dataset, and its level of democracy using the well-known Polity2 measure. A word of caution is in order. There is a clear risk of post-treatment bias here, since the variables used to capture the parochial interests of UNSC members (especially arms transfers and alliance ties) are almost certainly related in part to the region and regime type of the conflict state.
We also considered two temporal factors that might affect our results. Ten percent of our conflict episodes began during the Cold War and lasted beyond its end without having reached the UNSC agenda. Perhaps there is something distinctive about these episodes and the fact that before becoming at risk of making the agenda they spent time frozen out by the UNSC's Cold War paralysis. To assess this possibility, we created a "span Cold War" dummy similar to Gilligan and Stedman's (2003, 42) use of a dummy to account for civil wars that began during the Cold War but were still active after it ended. We noted earlier that some conflicts experienced multiple episodes. But how this "previous episode" dummy might affect agenda-setting speed is not obvious. On the one hand, we might expect that the squeaky wheel gets the oil, so that the second (or subsequent) time a conflict flares up, the more likely it is to draw attention and thus the faster it will reach the UNSC's agenda. Alternatively, if a conflict has already breached twenty-five fatalities per year and then waned without the UNSC putting it on their agenda, those wanting to revisit the matter when the conflict flares up again might face the problem of issue fatigue. If so, the existence of a previous episode could slow down the rate at which subsequent episodes made the UNSC agenda.
We also considered an alternative operationalization of our alliance ties and human suffering variables. Instead of counting all forms of alliances, we restricted the focus to just defense pacts, as some previous studies of UN intervention have done (Beardsley and Schmidt 2012). This produced the ordinal measures P5dpacts and E10dpacts. And as an alternative measure of human suffering, we replaced our civilian casualties variable with the broader measure of total battle deaths, taken from the UCDP Battle Related Deaths dataset.
We mentioned earlier that whichever state holds the rotating UNSC presidency exercises some discretion over the agenda. Not explicitly modelling this discretion may affect our findings. However, as we explain more in the supplementary online appendix, we know of no available theories to guide the analysis, and expectations would likely be indeterminate or idiosyncratic to the particular state and conflict. Nevertheless, we undertook exploratory analyses and found that none of our findings changed when we controlled for several systematic aspects of the UNSC presidency: whether the president is from a permanent or elected member, whether the president's state has an alliance with the conflict state, the geographic distance between the president's state and the conflict state, and the political difference between the president's state and the conflict state as measured by their respective UN ideal points and Polity regime scores.
Tables A1-A9 in the appendix present the results of these robustness checks. All of our main findings hold up well, apart from the E-10 contiguity variable, which in some of the models retains the negative sign but drops below conventional levels of statistical significance. It is noteworthy that, unlike civilian deaths, the battle deaths variable was consistently insignificant. This supports existing studies on agenda setting in the Council in which the number of battle deaths tests out as insignificant (Iwanami 2011;Frederking and Patane 2017).

Conclusion
This article has, for the first time, examined the speed with which civil conflicts enter the Council's agenda. Agendasetting speed matters: including a conflict on the Council's agenda is a necessary condition for further Council action; it sends a costly signal to relevant audiences; once included on the agenda, conflicts tend to become "sticky"; and Council members themselves attach a great deal of importance to the Council's agenda-setting processes. Furthermore, unlike the decision-making stage, the five permanent members cannot use their veto to block items from reaching the Council's agenda, at least not formally. In our view, this warrants an analysis of agenda-setting speed in its own right.
To explain variation in the speed with which civil conflicts reach the Council's agenda (if they reach it at all), we proposed a theoretical framework that combines realist and constructivist theory with insights from institutionalist theory and from bargaining theory. We find that in situations of civil conflict, parochial P-5 interests matter, and they matter strongly. However, P-5 parochial interests do not determine the Council's agenda-setting speed. Rather, P-5 interests are constrained in three important ways. First, agenda-setting speed is affected by the conflict's severity in terms of spillover effects. The UN is more likely to respond quickly to conflicts that threaten its organizational mission (preventing regional destabilization) and, even more so, to those that involve large-scale civilian deaths. Second, E-10 parochial interests-almost entirely ignored in previous UNSC research-are an important part of the story. When the E-10 have alliance ties with a country experiencing civil conflict, they tend to seek burden sharing, speeding up the agenda-setting process. Yet when conflicts occur in their backyard, they seem to be no different from the P-5 and slow it down. Third and finally, P-5 parochial interests are constrained by the distribution of preferences both among the P-5 and among the E-10. When preference heterogeneity is low, this makes it easier for Council members to swiftly include a civil conflict on the Council's agenda.
Importantly, while all of our hypotheses enjoyed support, we also found that some factors have more explanatory power than others, in that the effect of the E-10 on agendasetting speed can be greater than that of the P-5, and that parochial interests and general preference heterogeneity appear to play a bigger role in UNSC agenda setting than do conflict characteristics and the UN's organizational mission. Our results account for why it takes more than three years on average for a conflict episode to be put on the agenda, and why only 30 percent of these episodes ever reach the agenda: episodes have a dramatically reduced hazard ratio and are thus far less likely to ever reach the agenda (or only after a long delay) if they involve few civilian deaths or migrants, if any P-5 member has an alliance tie with or transfers arms to the conflict state, if any UNSC member is contiguous with the conflict state, if P-5 or E-10 members have strong trade links with the conflict state, or if E-10 members hold highly diverse general preferences. When several of these conditions obtain at once, the hazard ratio is miniscule.
Our findings also have important implications for the Council's legitimacy, in terms of both performance legitimacy and procedural legitimacy. Using a research design that is different from ours, Frederking and Patane (2017) conclude that the Council has a high level of legitimacy because the most devastating conflicts dominate the Council's agenda. But they look at the number of formal Council meetings per conflict each year, whereas we measure the time elapsed between when a conflict episode became active and when it first entered the Council's agenda. Thus, in our conceptualization the UN is more legitimate if devastating conflicts reach the agenda faster. For the UNSC's performance legitimacy, it makes a big difference whether a crisis reaches the Council's agenda rapidly and is discussed in public, or whether a crisis goes on for a long time without any exposure at a UNSC meeting and no chance of further action. Our findings are consistent with a legitimate Council that takes its mandate seriously, addressing more promptly conflicts that produce substantial human suffering and massive negative externalities for neighboring countries.
But our findings also help us assess an important element of the Council's procedural legitimacy by demonstrating how issues reach the agenda. Do parochial interests of the powerful permanent members skew agenda-setting speed? Surprisingly, Frederking and Patane (2017) do not find that P-5 interests determine which conflicts dominate the Council's agenda, suggesting legitimate Council procedures. Our findings, by contrast, yield a more nuanced picture. We show that narrow P-5 interests definitely affect the speed of agenda setting in important ways, but also that the Council does not appear to be an entirely illegitimate P-5 dominated elite club in which the elected members trade away their influence in exchange for bribes. Instead, the P-5 need to reach out to the E-10, rendering the agenda-setting process more inclusive of a larger set of interests.
In this article, we have taken a first stab at analyzing agenda-setting speed in the UNSC. There are at least three avenues for further research. First, while we have argued for the importance of studying the agenda-setting stage in its own right, we fully recognize that the story does not end here. Our work invites attention to the speed of what comes next, to the treatment of conflicts that manage to make it onto the Council's agenda. Second, agenda setting matters not only in the UN, but in other international institutions too. There are a few studies now on the International Monetary Fund (IMF) (Moody and Saravia 2013; McDowell 2017), but overall our knowledge of agenda setting in IOs remains limited. More comparative research is needed to explain potential differences in agenda setting across IOs. Finally, we have assumed that-on average-quicker agenda setting is likely to prevent conflict escalation, but we have not tested this assumption. Future research should investigate whether there is a systematic link between the speed with which conflicts reach the UNSC's agenda and the effect this has on the UN's success in terms of conflict resolution, peacekeeping, or sanctions. Pursuing these avenues would contribute to better understanding of how IOs work, their effectiveness, and their legitimacy.