Signaling restraint: international engagement and rebel groups’ commitment to international law

ABSTRACT Approximately 20% of contemporary rebel groups have expressed commitment to international law and signaled their intention to exercise restraint during wartime. Which rebel groups make these commitments and under what conditions? We argue that international engagement shapes the likelihood of rebel commitment to international law. Rebel groups with transnational non-military support and clear organizational structure are likely to speak the language of international law, especially near peace negotiations. We find support for our argument in a statistical analysis of international law commitments by rebel groups between 1974 and 2010. The analysis has implications for humanitarian engagement and promoting restraint in war.

From 1974 to 2010, approximately 20% of rebel groups expressed a commitment to international law through some 400documents. 1 These documents contain non-binding commitments to follow international rules and norms about the conduct of warfare. Some are unilateral commitments to international humanitarian law; others are internal codes of conduct, and include peace agreements that address humanitarian issues. In these documents, rebels promise to forego particular weapons, signal respect for human rights, and express commitment to treat non-combatants humanely. The documents provide a window to rebel groups' thinking, internal regulations, and strategic goals. Coupled with other signals rebel groups send either behaviorally or via social media, the documents produce rich materials to study the signals of restraint by rebel groups in civil conflicts around the world. Given its potential value as a signal, rebel commitment to international law can be fruitfully understood as an expression for their "Roots of Restraint," to follow the theme of this special issue (ICRC 2018).
The occurrence of rebel commitment to international law is an extraordinary phenomenon with implications for human security. Rebel commitment can sometimes lead to enhanced compliance, as Gleditsch et al. (2018) show this commitment-compliance connection in the case of antipersonnel mines. Moreover, rebel commitment is not cheap talk. Our analysis shows only 20% of rebel groups have made an expressive commitment. If international law commitment was completely costless, many more groups would have used it. And even if some commitments are turned out to be cheap talk, costless signals can have communicative value by altering the beliefs of political audiences (Kartik and Van Weelden 2019;Tingley and Walter 2011;Trager 2010). The overall "noisiness" of the signal likely varies in intensity, which has the implications for what commitment conveys. While most rebel commitments might remain cheap to talk with implementation problems and rampant reneging, signal heterogeneity ensures that the quality and volume of some signals might be distinguishable across different rebel groups.
Why would some rebel groups express commitment to international law, while others do not? Recent research shows that rebel groups in civil conflicts comply with international law to increase their legitimacy with key political audiences (Fazal 2018;Jo 2015;Stanton 2016). There is less emphasis on how 1 For broader dissident tactics and strategies, see the classic work by Scott (1985), and more recent work by Cunningham et al. (2017). rebel groups are socialized to address international law, however. Interactions between rebel groups and international actors, particularly their potential supporters, can explain whether and how rebel groups engage international law principles and norms. We argue and show in this paper that knowledge, capacity, and context help explain international law commitment by rebel groups. Rebels' knowledge about international law comes from transnational ties. Rebel capacity to interact with law-promoting international actors also enables international law talk. Finally, the context of peace talks increases the chance of rebel commitment to international law.
How do these three processes work? First, rebel interaction with international actors, such as international humanitarian organizations or diaspora populations, transmits knowledge of international law and the value of commitment. Second, rebel groups with strong internal command and control are likely to make international law commitments, partly because of their capacity to interact with international actors. Last, the context of peace talks accelerates rebel socialization into international law as groups use international law commitment to signal political viability and moderation in the presence of international actors.
We test our expectations by analyzing a new dataset of rebel commitment to international law. We find that rebel groups with non-military transnational support are more than twice as likely to express commitment to international law as those without. Rebel organizations with strong command structures are three to five times as likely to express international law commitment, compared to their counterparts with weak command structures. We also find that rebel commitment is nearly four times as likely near peace talks compared to other phases of conflict. Taken together, the evidence gives support for our argument that rebel groups' international engagement facilitates international law talk.
By explicating the process of rebel commitment to international law, we add international engagement to already-known mechanisms of rebel groups' international law-making, such as interaction with governments (Gleditsch et al. 2018), and rebel aims (Fazal and Konaev 2019). Our analysis complements existing studies of rebel groups' information and communication politics (Jones and Mattiacci 2019), while elucidating the political processes, motivations of key stakeholders, and political context of international humanitarian law engagement in conflict zones around the world.
Our analysis of rebel commitment to international law has broad implications for scholarship on international institutions and civil conflict, and also sheds light on the similarities and differences between state and non-state commitment to international law. With the similarities, both state and non-state actors experience socialization and employ the signaling functions of commitment while assessing the value of commitment in the shadow of compliance. Differences exist when non-state actors like rebels socialize with and signal to different actors for different purposes. While states socialize with each other in making international commitments (Goodliffe et al. 2012;Greenhill 2010;Mantilla 2018), some rebels socialize with a mix of specific international actors, but not others. Committing rebels tend to engage with transnational actors that provide non-military supports over other types of supporters.
For the literature on civil wars, we demonstrate how rebel groups employ international law talk as a political and military strategy in civil war. 2 We conceptualize rebel commitment to international law as part of a set of nonviolent and violent strategies (White et al. 2015), ranging from governance (Mampilly 2011;Stewart 2018) and diplomacy (Coggins 2015;Huang 2016), to extreme violence (Fortna 2015;Walter 2017a). International law commitments are a useful tool to offset rebel weakness against the state, by delegitimizing opposing governments and claiming the moral high ground through expressions of restraint. In many cases, commitment to international law serves rebel groups' political, military, diplomatic, and social purposes. Those messages of commitment or restraint are communicated to international actors, although commitment is a noisy signal given the non-compliance that often follows. Our study thus responds to recent calls for more inquiry into information environments in conflict situations (Walter 2017b;Zeitzoff 2017), by demonstrating a way in which civil war includes political and social contests.
The paper proceeds as follows: First, we lay out our argument, which examines the process underlying rebel commitment to international law. We then derive three hypotheses based on our observations and assumptions about civil war environments. Subsequently, we test our arguments about international engagement using a new dataset of rebel commitment. We conclude by discussing the ramifications of our findings for scholarship and policy.

Characterizing Rebel Commitment
Rebel groups' expressions of international law are soft law commitments, as they lack precision, obligation, and delegation, 3 with a few more legalized exceptions. Few rebel commitment documents include specific and precise commitments, such as the concrete steps to eliminate the use of child soldiers, or to stop the use of anti-personnel mines. Some commitments declare intent to uphold humanitarian principles, but provide few specifics. For example, the African National Congress in 1980 made references to the Geneva Conventions but without any details. 4 In many instances, rebel groups do 2 For broader dissident tactics and strategies, see the classic work by Scott (1985), and more recent work by Sawyer, Cunningham, and Reed (2017). 3 See Abbott and Snidal (2000) for the characteristics of soft law. 4 Statement on signing declaration, on behalf of the ANC and Umkhonto We Sizwe, adhering to the Geneva Conventions of 1949 and protocol 1 of 1977, at the headquarters of international committee of the Red cross, Geneva, November 28, 1980. http://theirwords.org/media/transfer/doc/za_anc_mk_1980_02-fdcf14f6a39d5c73ebee6318a8568ad5.pdf not define civilians status in their documents or express what their understanding of civilians are. 5 The general lack of precision or specificity means that rebels use commitment to express respect for norms, but they retain flexibility in interpreting the rules.
Rebel commitment to international law also features low-level, non-binding obligations. Rebel expressions to adhere to the Geneva Conventions, or rebel pledges to respect civilian lives, do not carry the same force. Such non-binding obligations usually have less legal weight. 6 This lack of binding accountability makes rebel commitment an expression or a signal rather than a firm legal commitment. Such light obligations contrast with hard law in international treaties or conventions that have binding, ratified and domesticated provisions, whereby a national legislature gives an international agreement legal force.
Last, rebel commitment documents rarely contain provisions related to the delegation to international organizations. In a few cases, rebel groups may permit monitoring and verification by international organizations. For example, rebel groups allow international organizations to monitor de-mining after a commitment to forgo anti-personnel land mines. But such delegation to entrust enforcement to third parties infrequently occurs. The absence of monitoring and verification by third parties leads to an enforcement system relying on self-imposed restraints.
The defining features of soft law in rebel commitment documents, namely self-imposed rules with flexibility and little accountability built-in, have deep political and historical roots. Rebel groups have been mostly excluded from formal international law-making (Roberts and Sivakumaran 2012). Some rebel groups participated as observers during negotiations over the additional protocol of the Geneva Conventions (Provost 2012), but they did not have voting power. Rebel groups cannot sign or ratify international treaties, so adopting existing language and making soft commitments provides them one way to engage with conventionally-state-dominated international legal system. Rebel commitment to international law operates as a political medium and communication tool in response to their historically weak position in international law system. Rebel commitment to international law carries diplomatic elements, as studied in rebel diplomacy literature (Coggins 2015;Huang 2016). In drafting rebel commitment documents, rebel groups engage in statelike diplomatic conduct. Some rebel groups attend international humanitarian conferences or deposit their commitments with the International Committee of the Red Cross. However, rebel commitment to international law is qualitatively different from setting up rebel consular offices or making high-level visits to the United Nations or the United States' Congress. Whereas high-level diplomacy emphasizes political recognition from specific politicians or 5 Taliban's internal code of conduct, Layeha (Nagamine 2015) is an unusual exception because it hints at some definition of civilians. 6 See Krieger (2018) for an international law analysis of the state-centric legal framework.
organizations, issuing international law documents targets a broader political audience. Potential audiences of rebel commitment would include rank and file rebel soldiers, opposing governments, and international audiences. The phenomenon of rebel commitment to international law is also distinct from other political communication tools, such as social media messages. Issuing documents are different from social media because the rebel documents take relatively more official forms compared to social media posts. International law commitments by rebel groups are usually drafted and signed by political leadership, which gives them an official imprimatur. Real-time website messages or social media posts are used for other ends, such as announcing victories (Jones and Mattiaci 2019) or reaching out to potential recruits (Mitts 2019). Documents on the principles of international humanitarian law usually appeal to international rules and norms, by contrast.
Above, we characterized rebel commitment to international law as soft law and as a political communication tool. Below, using that characterization as a baseline, we explain how the phenomenon of rebel commitment comes about by explicating the process of international engagement leading to international law talk by rebel groups in civil conflicts.

Knowledge, Capacity and Context
Three key factors and circumstances facilitate rebel commitment to international law. First, rebel groups acquire knowledge of international law through interaction with transnational actors. Second, greater rebel organizational capacity operates as an enabler, as it provides clear interlocutors for international engagement. Third, peace talks provide an opportune time where rebels have incentives to engage in international law activities to signal their moderation. In what follows, we explain how each of these three factors increases the likelihood of international law commitment by rebel groups.

Knowledge and international interactions
Interaction with international actors is an important source of rebel commitment. International humanitarian actors contribute to international norm transmission by providing necessary information. Without awareness of international rules and norms, rebel groups might not commit, and suppliers of international law facilitate this knowledge acquisition. Engagement with transnational actors allows rebel groups to acquire knowledge of international law and learn to express it (Fazal 2017).
International actors are also rebels' main audience for international law commitments, partly because many international audiences value international law. Not all international actors are pro-law constituencies, but normally, we can find these knowledge dissemination patterns from international humanitarian actors. International organizations and non-governmental organizations frequently train rebel groups on international humanitarian law and urge them to make international law commitments. For example, the United Nations Special Office for Children and Armed Conflict has negotiated action plans to stop the use and recruitment of child soldiers with 17 non-state-armed groups. 7 The Geneva Call's Deed of Commitment to eliminate anti-personnel mines includes more than 60 rebel groups (Gleditsch et al. 2018). The International Committee of the Red Cross has traditionally promoted international law with insurgent groups around the globe (ICRC 2018). The Center for Humanitarian Dialogue and UN mediators push states and rebels to include human rights and humanitarian provisions in peace agreements. In all these instances, the efforts of international actors provide a crucial supply of information that facilitates rebel commitment to international law.
Socialization may also allow rebel groups to recognize the value of commitment. 8 Many international organizations or non-governmental organizations that specialize in humanitarian affairs interact with non-state-armed groups to promote compliance with international law via persuasion either directly by invoking the law or indirectly by clarifying the situation and pointing out the benefits of compliance (Ratner 2011). Sometimes this international engagement leads directly to commitments by rebel groups. The Gates of Heaven Accord (Acuerdo de la Puerta del Cielo) is a noteworthy example of international interactions that produced rebel commitment. The document was published as a unilateral declaration by Ejercito de Liberacion Nacional (ELN), a rebel group in Colombia, but it was also the result of concerted efforts by German civil society, the Catholic Church, and the ELN (Jaramillo 2001, 197-199).
These socialization opportunities are a function of transnational networks. Diaspora populations are also part of such networks. The importance of diaspora populations is already part of research on rebel diplomacy (Huang 2016) and how external funders affect the course of civil wars (Sawyer, Cunningham, and Reed 2017). Whether diasporas help pacify civil conflicts is disputed, 9 but the transnational exposure provides an opportunity for international engagement. 10 Not all transnational actors promote international law, however. Some outside interactions with rebel groups in internal conflicts are meant to provide military support and win the war, with little regard for law. Charles Taylor's initial support for the Revolutionary United Front in Sierra Leone is 7 United Nations Office of the Special Representative of the Secretary-General for Children and Armed Conflict. "Action Plans" Available at https://childrenandarmedconflict.un.org/tools-for-action/action-plans/ 8 See the discussion in Hofmann and Schneckener (2011) as well as Caswell and Somer (2018). 9 Some argue the aggravating effect of diasporas financing and supporting violence (Collier 2000); others highlight the peacemaking efforts by diaspora population (Smith and Stares 2007). 10 Recent finding by Petrova (2019) regarding the positive role by diaspora support and rebel groups' non-violent tactics supports this engagement hypothesis. one case where outside supporters disregarded international law. More broadly, military support often reflects geopolitical interests of rebel supporters and places less emphasis on humanitarian concerns (Cunningham, Sawyer, and Reed 2017;San-Akca 2016). By contrast, actors without strong geopolitical motivations for intervention are relatively more likely to promote international law. Non-military support from NGOs and international organizations often attempts to uphold and strengthen international law. Of course, not all transnational actors offering non-military support are pro-law constituencies. Also, rebels may engage with transnational actors so they can transfer non-military support into military support. Rebels sometimes abuse humanitarian actors' good intentions (Terry 2002), so non-military support from transnational actors may not always promote international law leading to rebel commitment.
Recognizing the heterogeneity across external supporters for rebel groups, we expect that transnational supporters with non-military support agendas are more likely to be pro-law constituencies, relative to transnational actors with military support and geopolitical agendas. Here, we are circumspect on inferring pro/anti-law preferences from the type of support given. Some transnational actors simply do not have the capacity to provide military aid. Also, the same actor would have different law preferences if it provided military assistance in one case, but non-military assistance in another. With the caveat, our central contention is rather stated in average terms, by weighing the relative motivations of actors to engage in law promotion activities. Compared to outside supporters that provide military support to alter the military balance with little regard for law, rebel engagement with international actors who provide non-military support will likely provide rebel groups with information about international law as well as the benefits of commitment. This logic generates the following hypothesis: Hypothesis 1 (knowledge and the influence of transnational support): Rebel groups with non-military transnational support are more likely to commit to international law than those without such support.

Capacity and the Role of Rebel Organization
Acquiring knowledge and information about international law through interaction with transnational actors is only one part of rebel commitment. Rebel groups themselves must have the capacity to form and articulate a commitment. A clear command structure can facilitate this process of using international law commitment for political benefits. When rebel groups decide to commit to international law, they might anticipate that engagement with international law will enhance their international standing. Rebel groups can commit as part of a competition for political legitimacy and reputation against national governments. Rebel commitment to international humanitarian law could also increase internal control with the expectation of improved codes of conduct that strengthen internal discipline within rebel organization.
Strong command and control structure of rebel organizations can also facilitate international engagement. 11 Dealing with a central authority is more straightforward than addressing disjointed components of a rebel organization. When a group has a clear structure and hierarchy, international actors can easily identify interlocutors within the rebel organization. In contrast, groups with weak command and control structures will have more difficulty in establishing international contacts.
International actors, for their part, may expect that commitment by groups with strong central command and control is more likely to lead to compliance. 12 Strong command and control could lead international actors to believe that rebels have ability to follow through on commitments. In contrast, weak command and control groups lack the ability to implement policy changes, which weakens the credibility of commitment. This does not mean that groups with strong command and control will always follow through on international law commitments, because rebels may prefer not to comply after the commitment, seeking their own military and diplomatic advantages. Rather, our claim indicates that the expectation of international actors with respect to future compliance might be greater for strong command and control groups, compared to the counterparts with weak command structures.
We suggest several pathways the strong rebel command might have pro-law preferences, compared to the groups with weak command. First, strong rebel command can commit to international law expecting the political benefit of rebel commitment to appease potential supporters, knowing that they could potentially renege on the promises. Second, strong rebel command might commit to international law if the leadership expects to get the military benefit of internal control when they streamline the internal code of conduct consistent with international law -by disciplining soldiers not to loot, for example, to garner civilian support.
Additionally, the strength of rebel command is important because strong internal control can minimize the potential costs imposed by commitments. Just like any other political promise, rebel commitment occurs in the shadow of compliance. By expressing commitment to humanitarian rules, rebel groups might have to consider forgoing military gains. Although commitment to international law can complement military efforts in some cases, "restraints" from the laws of war could limit warfighting options.
For rebel groups with strong command and control, these potential material costs may be offset by additional internal discipline or continued relationships with transnational actors. A law commitment by these strong groups should also send clearer signals of greater baseline capacity and ability to punish individual members. The noise around the signal from a strong command and control rebel group is, therefore, going to be lower, and hence the incentive to use commitment higher. Rebel groups with weak command and control, in contrast, might not be able to carry the costs of commitment, such as administrative cost of conducting international engagement or implementation cost of streamlining its internal code of conduct. They might even worry about the potential hands tying effects of commitments. Weak command and control groups therefore cannot credibly signal their compliance potential thru the use of rebel commitments, consequently producing a low probability for the commitment to international law.
The comparison of capacity of weak and strong rebel commands leads us to expect the following relationship between rebel organization and commitment probability.
Hypothesis 2 (capacity of rebel command): Rebel groups with strong command and control capacity are more likely to commit to international law than those with weak command and control capacity.

Context and the Proximity to Peace Talks
Besides rebel groups' transnational networks and internal organization, the political context plays a crucial role in rebel commitment. International actors are increasingly active in peace negotiations (Bell 2000;Howard and Stark 2018;Nilsson 2012). Given substantial international attention and involvement in peace talks, rebel groups should be exposed to international engagement. Peace talks create opportunities for rebel groups to make international law commitments. Near peace talks, rebel groups may have incentives to signal commitment to international law to increase their political legitimacy in the eyes of international actors. At the same time, committing to peace talks provisions on international law and human rights can increase rebel groups' stature with key actors in the negotiation. As international organizations and NGOs facilitate peace negotiations, international law commitment may allow rebels to signal their virtue to mediators and stakeholders for peace.
Going into peace talks, rebels must also signal their willingness to compromise. Rebel groups sometimes use shows of force to demonstrate their military power (Wood and Kathman 2014) especially when spoilers are present, but the core leadership conducting peace talks will have an incentive to appear politically moderate. International law commitment is one possible signal of moderation. Especially given multiple rebel groups, groups that signal their moderation can distinguish itself. Committing to international law and expressing restraints on violence might indicate that rebels will not seek every possible military advantage.
The above discussion suggests that the proximity to peace negotiations might provide opportunity for international engagement and makes rebels more willing to commit to international law, compared to the periods where peace negotiation is not in the offing. Therefore, we expect that rebel commitment to international law is more likely near peace talks, compared to other phases of conflicts.
Hypothesis 3 (context of near peace talks): Rebel groups are more likely to commit to international law near peace talks than away from peace talks.

Empirical Analysis of Rebel Commitment
We evaluate our theoretical expectations with a newly constructed dataset of rebel commitment to international law. Our sample of rebel groups includes all rebel groups from 1974 to 2010 in the Non-State Actor (NSA) dataset (Cunningham, Gleditsch, and Salehyan 2013). 13 The NSA dataset includes only rebel groups involved in a conflict that generated at least 25 battle deaths in a given year. The following analysis and the inferences we make will, therefore, focus on relatively militarily capable groups. 14 The resulting sample includes groups from every region of the world: 184 African, 35 American, 192 Asian/Oceanian, and 25 European.
Using the information on start and end dates of conflict dyads between a state government and rebel group, we construct a cross-sectional time series dataset in which the unit of observation is rebel group-year. This panel data is unbalanced dataset because some rebel groups are active for only a year but others fight for decades. Rebel-year observations are the unit of analysis because our third hypothesis focuses on changes in the political context. 15 The dependent variable is Rebel Commitment. It is a binary indicator, set equal to 1 if a rebel group made any international law commitment in a given year and 0 otherwise. 16 We utilized multiple sources to identify international law commitments. The primary source is the Directory of Armed Non-State Actor Humanitarian Commitments (Theirwords archive, hereinafter), 17 curated by the Geneva Call, a humanitarian organization based in Switzerland. The Theirwords directory is a comprehensive collection of public documents about agreements, unilateral declaration, and codes of conduct by rebel organizations. 18 Thematic areas covered in the database include children in armed conflicts, landmines, sexual violence, torture, and violence against civilians. Beyond the Geneva Call's directory, we also looked at secondary sources for cross-checks, especially the authoritative textbook by Sivakumaran (2012). 19 Other secondary sources provided further examples of rebel commitment, including internal codes of conduct from Bangerter (2011) and peace agreements from Bell (2000). We cross-checked further secondary sources as well, by examining reports from United Nations, Human Rights Watch, and International Committee of the Red Cross, for omitted commitments.
Using this range of sources, we collected 404 commitment documents by rebel groups between 1974 and 2010. While some rebel groups made individual commitments in a year, other documents were jointly signed by multiple rebel groups. 64 out of 364 rebel groups (about 18%) in our sample made at least one commitment to international law in a written document. At the rebel group-year level, 142 out of 1944 observations have a rebel commitment.
To ensure the representativeness of the sample, we considered the possibility that the Geneva Call selectively reached out to a certain set of rebel organizations, especially well-known rebels. Geneva Call operates in about 35 countries, including almost all the civil wars in the world, and the database includes some obscure rebels 20 that are not part of the NSA dataset. This alleviates the concern that we might be seeing only salient groups in the analysis. We also checked whether the Geneva Call's activities and engagement with non-state armed groups such as its programs for banning anti-personnel mines occupy our sample in great portions. We find that 14% of our observed commitment documents are the result of Geneva Call agreements, which does not dominate the sample. 21 We attempted to ensure the representativeness of the sample and proceeded with caution to be comprehensive in terms of rebel commitment documents, but it is still possible that we missed some international commitments by rebel groups -especially private commitments. For example, some UN action plans are kept confidential for security reasons (interview with Alec Wargo, a UN official, May 2016), and we might have missed other internal codes of conduct from obscure groups. As the scope of our theory deals with public messages rather than private ones, we proceed with a relative confidence that our data collected from publicly available sources are representative enough to test our theory.
Moving to the test of our three hypotheses, we now present our measurements for the three key correlates of international engagement for rebel commitment. We use a transnational support variable from the NSA dataset to measure interaction with international actors. Non-military Transnational Support is a dichotomous indicator of whether a rebel group received nonmilitary support from transnational non-state actors, such as international organizations, religious or ideological groups, ethnic or diaspora movements (Cunningham Gleditsch, and Salehyan 2013, 523-4). Connections between the Sudan People's Liberation Movement (SPLM) and Christian religious organizations in the USA are one example of non-military transnational support. Note that this variable purposefully excludes military support because we expect that military sponsors are inattentive to international law commitments. Although this measure is an imperfect proxy for pro-law preferences of international actors, we think this may be the best available measure due to a lack of specific and comprehensive data on which transnational actors rebel groups are engaged with. We later compare the effect of non-military support from transnational actors to the effect of military support from international actors to help validate this choice.
The second independent variable, Central Control Strength, measures the extent of central command within the rebel organization and the ability to control its members. We use a four-point scale from the NSA data, which classifies each rebel group as having 1) no central command, 2) low central command, 3) moderate central command, and 4) high central command. The majority of rebel years (60%) exhibit moderate level of central command, while no central command (6%), low level (15%), and high-level command (19%) comprise the rest of observations. There are 19 rebel groups (105 rebelyear observations) with no discernible command and control. These groups made no international law commitments, which suggests that a minimal level of command and control structure may be required for international law engagement.
The third independent variable, Near Peace Talks, is a binary measure of whether the rebel group and the government signed a peace agreement in the current or subsequent year. We include the year before peace talks because many negotiations take a couple of years to complete. The peace negotiations variable is based on data from Harbom, Högbladh, and Wallensteen (2006). 22 Our empirical models include multiple control variables that could be correlated with our independent variables and international law commitment. First, we account for whether transnational non-state actors provided a rebel group with military aid using a dichotomous variable (Military Transnational Support). Our theory applies to rebel groups that rely on non-military support, and we expect that rebel groups with foreign military support will be less likely to make international law commitments. We also account for aid from foreign state sponsors, which could be correlated with support by transnational nonstate actors and a rebel group's motivation to express commitments to international law. Two dummy variables, Non-military Support by Foreign Government and Military Support by Foreign Government, are added to the models, so the reference category is "no support by foreign government." We also control for the military capacity of rebel groups, as it can affect commitment motivation and is correlated with command and control. We employ a measure from NSA dataset, Military Strength, a five-scale ordinal measure of rebel group's military strength relative to government forces. This is a composite index primarily based on rebel groups' number of troops, adjusted for their mobilization capacity, ability to procure arms, and fighting capability relative to the government (Cunningham Gleditsch, and Salehyan 2013). The categories include 1) much weaker, 2) weaker, 3) parity, 4) stronger, and 5) much stronger. Most rebel groups are much weaker or weaker than the government forces, but about 13% of rebels are at parity or stronger. Another correlate of organizational capacity is Territorial Control, which is measured as a binary indicator of whether a rebel group has its own territory over which it exercises effective control. Approximately one-third of rebel groups have effective control at some point.
Rebel aims may also change the need for international law commitment. To account for the possibility that rebels are more likely to commit when they want a state of their own, we include a dichotomous variable Secessionist. The variable is coded as 1 if a rebel group pursued secession or autonomy in a given civil conflict, and 0 otherwise, following the NSA dataset's classification of conflicts.
Additionally, domestic political conditions may be correlated with rebel group characteristics and incentives to make international law commitments. For example, an authoritarian government could block access to a rebel group by international actors by constraining political activities, which may in turn affect the likelihood of rebel groups' international law commitment. In this case, rebels would also have extra incentives to distinguish themselves from an oppressive government with international law commitments. We use two proxies for domestic political dynamics-rebel political activity and the host state's political regime. The Unauthorized Wings variable identifies rebel groups with a political party that is not recognized by the government, whereas Legal Wings indicates that a rebel group has a legal political party. These variables have a base category of rebel groups with no associated political party. In addition to a rebel group's political party, we control for the regime types of host countries, using an indicator of democracy provided by Geddes, Wright, and Frantz (2014). 23 Rebel military bases outside of the host country might also be related to transnational support and rebel commitment. Bases reflect state support and provide opportunities to interact with transnational actors. To control for this possibility, we include a dummy variable, Extraterritorial Bases, coded as 1 if a rebel group has some or extensive bases in foreign countries and 0 otherwise.
To examine broad trends in rebel commitment, we estimated the effects of the Cold War and the jihadist ideology of rebel groups. The end of the Cold War created systematic changes in the international environment for civil conflicts (Kalyvas and Barcells 2010). The Cold War variable is a binary variable, coded 1 for 1974-1989 and 0 for 1990-2010. The jihadist ideology variable captures a key trend after the Cold War (Walter 2017a). Ideology provides rebels with an alternative view of the world, and jihadists may prefer to avoid commitments to international law. Using several sources, 24 we coded rebel groups as jihadist as a 1 if we found that the group sought to wage a jihadist campaign, the implementation of sharia law, the formation of a strictly Islamist state for the entire country and/or the removal of Western or non-Islamist influences on society. 25 With this Jihadist Ideology variable, we are particularly interested in looking at whether jihadist groups are less likely to commit to international law.
Our final set of controls addresses the civil conflict environment and temporal dynamics of commitment. The presence of multiple rebel groups might spur competition through non-violent strategies, so we include Number of Rebel Groups -the number of other rebel groups in a country at a particular year. This variable ranges from 0 to 10. Rebel groups sometimes face only the government, but long-running civil wars (in the Democratic Republic of Congo, for example) involve many rebel groups due to splits or new groups. The other control variable is Age of Rebel Group, which counts the number of years a rebel group has been in civil conflict with the government. 26 We control for Time (Year), because rebel commitment to international law could be a function of global trends. Since our time frame starts in 1974, the Time variable is bounded between 1 and 37. To control for temporal dependence in rebel commitment, we include a cubic polynomial, suggested by Carter and Signorino (2010), to account for the possibility that each rebel commitment event is dependent on one another across time. 27

Results
In this section, we summarize the results of our analysis on the dataset of rebel commitment to international law. We start by showing the global trend of annual rebel commitments from 1974 to 2010 in Figure 1. The circles locate the number of commitment documents for each year and the trend line is estimated with loess regression.
We first note that the annual number of rebel commitment documents increases over time. The global increase in rebel commitments is consistent with our conjecture that international organizations and non-governmental organizations have increasingly engaged with rebel groups, thereby encouraging the rebels to socialize and learn about international laws. Moreover, commitment is fairly rare before 1990, but increases dramatically with the end of the Cold War. We suspect that this trend reflects diminished foreign military support for rebel groups and more activity by pro-law constituencies after 1990. The number of annual commitments by rebel groups likely reflects greater non-military engagement by foreign actors in civil conflicts. With the increasing use of social media and the changing space of humanitarian sector, the future trend is to be determined.
Before turning to parametric modeling, we offer cross-tabulations of commitment and our three key independent variables. These descriptive statistics provide preliminary support for our theoretical expectations. If socialization is salient, we expect to see that rebel groups with non-military transnational support are more likely to express commitment to international law than those without it. Consistent with our expectation, the data show that 27% of the rebel groups receiving non-military transnational support (28 groups out of 107) committed to international law at least once, whereas only 14% of the rebel groups without non-military support from transnational non-state actors (32 groups out of 233) made a commitment to international law during the same period. Groups with strong command are more likely to commit to international Clustered standard errors (by rebel groups) in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01, constants suppressed due to space constrains.
law (31%, 23 groups out of 74), compared to weak command and control groups (25%, 11 out of 44). Context appears to matter as well. Among the total 142 rebel commitments, 65 instances (46%) occurred near peace talks. 28 The statistical models in Table 1 estimate the association between our covariates and rebel commitment while adjusting for other factors. Because the dependent variable is dichotomous, we employ two logistic regression estimators. Model 1 fully pools observations, while Models 2 and 3 employ random effects. Pooled logit models estimate a common intercept for all rebel groups. Random effects logit models, or partial pooling or varying intercept models, account for unobserved heterogeneity among rebel groups by estimating an intercept for each rebel group and the correlation among those intercepts (King 2001). 29 Across all three models, the coefficients for our three key independent variables are all positive and statistically significant at conventional levels. Transnational support, strong central command, and proximity to peace talks all increase the probability of international law commitment by rebel groups. The estimated coefficients are fairly stable across these two models and other specifications in our robustness checks.
Because the coefficients in Table 1 are not directly interpretable, we calculate the substantive effects of our key independent variables. 30 Figure 2 plots the predicted probabilities of rebel commitment for each value of transnational nonmilitary support, along with the number of observations in each category. Consistent with our socialization expectation (Hypothesis 1), the probability of rebel commitment to international law increases with non-military transnational support. While the predicted probability of rebel commitment to international law is 11% when a rebel group receives non-military support from transnational nonstate actors, the probability decreases to 5% without such support.
Strong command and control of rebel organizations also increases the predicted probability of commitment, consistent with our expectation in Hypothesis 2 (rebel organization). Figure 3 plots the predicted probabilities of rebel commitment for each level of central command strength, along with the number of observations in each category. 31 In expectation, the probability of rebel commitment to international law increases by about seven percentage points (4% to 11%) as the rebel group's central command strength shifts from "low" to "high." The probability of commitment for rebel groups with "high" central control capacity is also about four percentage points higher than that for groups with "moderate" capacity.
28 For all of these comparisons, we can reject the null in a chi-squared test, which indicates a statistically significant difference in the frequency of commitment across these categories. 29 The "variance (Rebel group)" in Table 1 is the estimated variance of the rebel-group specific intercept parameters. 30 The estimates are from Model 2 in Table 1. To compute quantities of interest, we follow the suggestions of Hanmer and Kalkan (2013). Their "observed case" approach holds covariates at the observed values for each observation in the sample when calculating the relevant quantities of interest, and then averages the results over all the observations. We choose Model 2 to calculate substantive effects because this estimator fits the data better. See the online appendix for a comparison of model fit using separation plots. 31 Given the relatively small number of rebel groups in "high" category, we checked whether the results are driven by this potential outlier. The difference between moderate and high levels of command and control suggest that the results are not driven by outliers in the "high" category. Further checks on this are reported in the online appendix. Figure 4 shows the predicted probability of international law commitment for rebel groups near peace talks. Proximity to peace talks increases the predicted probability of rebel commitment to international law by about 12 percentage points. Furthermore, we observe that rebel commitments are made outside of the peace agreement venues. Approximately 60% (64 out of 110) of commitments made near peace talks are not part of peace-agreements. For example, the Frente Farabundo Martí para la Liberación Nacional (FMLN) made an agreement on securing human rights with the government of El Salvador in 1990 in addition to a separate peace agreement. 32 This multiple signals within and outside of peace agreements near peace talks support our interpretation that rebel groups have incentives to signal restraint and moderation near peace talks by expressing commitment to international law. 33  Table 1, we calculated the predicted probabilities of rebel commitment to international law, holding covariates at the observed values for each observation in the sample. Black circles are point estimates and black bars represent 95% confidence intervals of those point estimates. The shaded histogram plots the number of observations for each category of non-military transnational support variable. 32 As a robustness check, we also re-estimate the regression models after removing peace agreements from the dependent variable of rebel commitment. The alternative outcome is coded as 1 if a rebel group committed to any international law besides a peace agreement in a given year and 0 otherwise. We report the result in the online appendix. 33 Two examples in our dataset further illustrate the utility of our finding. The Moro Islamic Liberation Front (MILF) in the Philippines made a commitment in 2000 (anti-personnel mines) and a made peace agreement in 2001. The Alliance for the Re-liberation of Somalia/Union of Islamic Courts ARS/UIC in Somalia made commitments in 2006 (civilian protection) and 2008 (human rights) and as well as a peace agreement in 2008. In a recent example, although outside of the temporal scope of our data , Taliban issued a statement in May 2018 regarding "humanitarian issues" and "civilian casualties", in addition to "need for peace" (the full transcript captured in Roggio 2018). Note that this is the period when the talks with the US was in the offing.
Turning to the discussion of other variables, we do not find statistically significant associations between rebel commitment to international law and several control variables. The Number of Other Rebel Groups does not appear to have a discernible impact on rebel groups' international law commitment decision. Political Wings, External Bases, and Military Support by Non-State Actors also produce statistically insignificant coefficients. Similarly, we find a negative but statistically insignificant association between secessionist aims and the probability of international law commitment. Age of Rebel Group on the other hand seems to matter: longer running rebel groups are more likely to make soft law commitments than nascent groups. While the predicted probability of commitment by a rebel group that has been in conflict for 12 years is 7%, the probability almost goes up to 13% when a group has lasted for 30 years. 34 To a certain extent, the result is consistent with our socialization story: as rebel age increases, there are more socialization opportunities, which can lead to rebel commitment.
At face value, our secessionist estimate appears to be inconsistent with the importance of secessionism in promoting desirable behavior by rebel groups, including less civilian targeting (Fazal 2017;Jo 2015) and child soldiering (Lasley and Thyne 2015), as well as greater diplomatic engagement (Coggins  Table 1, we calculated the predicted probabilities of rebel commitment to international law, holding covariates at the observed values for each observation in the sample. Black circles are point estimates and black bars represent 95% confidence intervals of those point estimates. The shaded histogram plots the number of observations for each category of central control strength. 34 Results reported in the online appendix. 2015; Huang 2016). However, we think our results are actually compatible with prior work and consistent with our theoretical explanation. 35 Our results indicate that secessionists are not more likely to make international law commitments than non-secessionists. Although secessionists stand to benefit from international law commitment in their quest for statehood, they may also be especially sensitive to the consequences of non-compliance. Secessionist rebels may also pursue different legal strategies other than publication of soft laws. For example, the Polisario filed an objection to the EU-Morocco association agreement with the European Court of Justice. 36 The result suggests scholars should pursue a more nuanced understanding of secessionists' political strategies toward international law. There are substantial differences between the soft law commitments we cover here and state-like diplomatic acts.
In terms of the broad trend in rebel commitment, the insignificant coefficient of the Cold War variable shows that there is no systematic difference between the Cold War period and post-Cold War period in terms of rebel commitment behavior. The non-result may be due to the fact that we controlled for other factors that are produced by the Cold War, such as the prevalence of  transnational support and other factors. The coefficient for Jihadist Ideology shows that jihadist rebel groups are less likely to commit to international law, compared to other groups with non-jihadist ideologies. The finding is consistent with recent work on the extremist's advantage in civil wars (Walter 2017a) in which jihadists gain advantages by distinguishing themselves ideologically. Now, to further investigate the implications of our theoretical expectations, we also estimated several interactive models. Recall that our theoretical mechanisms of socialization and international engagement suggest some interactive effects of these variables. The presence of transnational support might augment the effect of rebel command and groups with strong central command may be more likely to be contacted by transnational actors. Also, proximity to peace agreements may complement the presence of transnational support. We find evidence of an interactive effect between transnational support and rebel command, but not for transnational support and peace agreements. 37 The results indicate that transnational support and strong rebel command are complementary sources of commitment, but that transnational support does not modify the effect of proximity to peace talks. We interpret this result to mean that additional transnational support beyond whatever internal engagement rebels experience near peace talks does not increase the probability of commitment.
In general, our empirical results match our theoretical expectations about the correlates of international law commitment by rebel groups. Three core aspects -knowledge, capacity, and context -determine how international engagement drives rebel groups' soft commitments to international law.

Conclusion
Without power, authority, or influence over formal international law-making, rebel groups engage with international law by developing soft law commitments. The results have implications for humanitarian action. Political engagement with rebel groups is inevitable during civil conflict as states and international actors seek humanitarian access, mediation, and conflict resolution. Our study provides suggestive evidence of when rebel groups engage with international law, which may facilitate humanitarian gains. Specifically, our findings suggest that peace talks provide a window of opportunity for international actors to engage rebel groups, as recent research finds that providing and implementing humanitarian provisions may make a more durable peace (Joshi and Quinn 2017).
At the same time, we recognize that our research delves into the issue of rebel groups' words, not their deeds. Commitment does not automatically lead to compliance. Since rebel commitments to international law produce a noisy signal of compliance and non-compliance, identifying when rebel commitment 37 Results reported in the online appendix. leads to rebel compliance is a crucial next step for scholars and practitioners. In devising engagement strategies with violent armed groups, it would be useful to know what leads rebel groups to follow through or renege on their promises. If some rebels follow through on their soft law commitments under certain circumstances, encouraging commitment using particular social mechanisms in specific institutional contexts could reduce human suffering in civil conflict.
The patterns and conditions for rebel commitment and compliance deserve more scholarly attention, because they have important consequences for human security. This is especially so as the rebel messaging space gets wider beyond traditional venues, including social media as well as at international courts. Such research should also consider further diversity in our non-state armed actor profiles, which encompass rebels, terrorist groups and criminal gangs. How the map of commitment and compliance differs, depending on rebel organizational structures, as well as their sources of revenue and political support, will help us devise effective humanitarian actions for engaging difficult actors in global conflicts.
In the context of multiple armed groups, the role of strategic competition may have a prominent role in conditioning rebel commitment to international law. When a rebel group receives foreign support, it might matter if a foreign supporter's international rivals are simultaneously intervening in the conflict. Depending on the identity of the international rival and foreign supporters, the said insurgent group might intensify its effort to signal its international law commitment as part of lobbying effort to obtain continued outside support. In these cases, armed groups will try to leverage the value of commitment to international law, which may then alter the security policies of potential interveners into civil conflicts.
Future work can also examine the diversity of these rebel commitments to international law and their effects. The rebel commitment data in our study includes internal codes of conduct, peace agreements, and unilateral declarations. Considering how each type of commitment translates into restrained violence will advance our understanding on conflict and humanitarian outcomes, going beyond the general determinants of rebel commitment we identified in this paper. recommendations in this article are those of the author and do not represent the views of the funding agency.