Bargaining with Insurgencies in the Shadow of Infighting

Bargaining with Insurgencies in the Shadow of Infighting Abstract Despite the long standing “no concessions” argument, scientific studies now suggest that governments can benefit from negotiating with militant insurgencies. However, despite government efforts, the leaders of insurgent movements often appear fanatical and unwilling to negotiate. This behavior presents a puzzle: If the leaders of insurgencies mobilize to create political change, and a government offers concessions, why do insurgent leaders refuse to negotiate? Using a game-theoretic model, we argue that insurgent leaders may rationally reject negotiation due to an internal commitment problem. Specifically, when leaders cannot credibly share the benefits of peace with their rivals, insurgent leaders may reject offers over fear of an internal conflict, which could leave the entire group vulnerable to government exploitation. However, the model demonstrates that insurgent leaders should negotiate if power in the insurgency is shifting in favor of their rivals, as it could help them maintain control of the movement. We illustrate these hypotheses using evidence from the Nigerian state's conflict with the Movement for the Emancipation of the Niger Delta (MEND) organization and Boko Haram. infighting, negotiation, insurgency For many years, policymakers argued that states should never negotiate with insurgencies due to the inability of these movements to form credible commitments. Negotiation would allow insurgencies to gain from armed conflict, which in turn would encourage these movements to continue using violence to blackmail governments in the future. Recent studies, however, indicate that, while negotiation is risky, there are conditions in which governments can form credible agreements with insurgencies to facilitate conflict termination and transition rebel fighters into normal politics.1 Negotiating further allows governments to avoid the costly use of force, which appears often to be counterproductive when used indiscriminately (Kalyvas 2006; Kocher, Pepinsky, and Kalyvas 2011; Weinstein 2007).2 Yet, despite this growing consensus, several insurgent leaders appear apocalyptic and seem to view violence as an end into itself (e.g., Wood 2015). These leaders show little interest in government efforts to negotiate peacefully. This puzzling behavior raises the question: Why are some insurgent leaders unwilling to negotiate, even when governments are open to making concessions? Using a game-theoretic model, we argue that negotiation creates an internal commitment problem for insurgencies, in which leaders cannot credibly refrain from monopolizing the peace dividend and exploiting their rivals in the movement. This commitment problem raises the risk of infighting, leading strategic insurgent leaders to reject negotiation outright. Strategic intransigence is often interpreted as extremism. However, despite the risk of infighting following peace, we demonstrate that insurgent leaders may accept negotiation if they face growing challenges from their rivals for control over the movement. Empirically, the model predicts that leaders will reject negotiation when their power is at relative parity with their rivals, but the rivals’ power is not increasing. On the other hand, insurgent leaders will negotiate if power in the insurgency is shifting in favor of their rivals, thereby increasing the probability of infighting. Intra-Insurgency Politics and the Problem of Negotiation Initial studies focused on whether states should attempt to negotiate with violent nonstate actors, given that aggressive policies appeared to be counterproductive. Many of these studies concluded that the presence of extremist factions3 in nonstate groups posed a key barrier to negotiation (Kydd and Walter 2006; Pearlman 2008/2009; Stedman 1997).4 These “spoilers” typically responded to negotiation by resorting to terrorism, or attacks against noncombatants, to prevent both the state and the moderates from fulfilling their obligations. This intransigence makes reaching credible deals with multifaction insurgencies very difficult. However, a key empirical problem is how to identify extremists prior to observing the outcome of negotiations. Formal studies suggest there are strategic incentives for all militant fighters to signal extremism to gain bargaining power (Lake 2002). These incentives may hinder efforts to separate moderates from extremists ex ante. For example, consider the Quetta Shura in Afghanistan. The foremost Taliban group embraces violence, signals unwillingness to compromise, and appears content to continue fighting the United States. Yet, privately, some Quetta Shura fighters, including some believed to be in more senior leadership positions, have signaled some willingness to negotiate. It is therefore difficult to definitively classify these factions as moderates versus extremists prior to observing their response to negotiation. If we can only identify factions as extremist following negotiation, it becomes tautological to use the presence of extremists to predict the failure of negotiation. Defining extremists as those who gain rents from conflict or have strong tastes for violence constitutes a post hoc explanation since we cannot observe whether extremists exist ex ante. Descriptively, violence often increases following negotiation, and those acting as spoilers appear vehemently opposed to peace. What causes some to engage in extremist opposition to negotiation? Let us first consider the makeup of insurgencies. These movements may be comprised of a diverse set of individuals with a variety of motivations for engaging in violence (Pearlman, 2008/2009; Cunningham, Bakke, and Seymour 2012; Humphreys and Weinstein 2008; Peterson 2001; Wood 2003). Some may be true believers in a cause and may share the motivations of the leadership; others may be motivated to join by economic incentives, thrill-seeking, boredom, revenge, or issues tangential to the group's stated goals (Chai 1993). For example, Waldman (2010) demonstrates that the fighters and commanders within the Quetta Shura express widely different motivations for engaging in violence. The fighters disagree on how to reform the Afghan state, lack a common definition of Shari'a law, and express inconsistent conceptualizations of victory. While some fighters favor a fundamentalist state, others appear to prefer a limited central state with greater local autonomy. These numerous divergences indicate that victory as perceived by some may not be victory for others, and may even be a loss. Those seeking a strict interpretation of Islam enforced by violence are likely to challenge individuals seeking to limit the ability of the state to intervene in local issues. Similarly, success for Afghan nationalist fighters may conflict with success as defined by Pakistani-linked elements of the group, who may prefer to keep the government weak. Despite competing preferences, the fighters are united in their fight against the United States and the Afghan government. Other insurgencies similarly consist of multiple factions with different preferences and agendas.5 This diversity creates a substantial challenge for the leadership, which must unify the factions to maximize the insurgency's collective power.6 The variety of motivations also creates the potential for competition and rivalries, which may create sizeable challenges for negotiation (Bapat and Bond 2012; Christia 2012; Cunningham 2006; Fjelde and Nilsson 2012; Nygård and Weintraub 2015). If the insurgency is fractionalized, negotiations must appease both the leadership and the multitudes of fighters in order to achieve full disarmament. Ideally, the leadership would reach some agreement regarding distribution of the benefits to their followers after a settlement is reached. Once a leader receives concessions from the government, they must convince their rivals that they will share the benefits of the peace equitably. Unfortunately, given that the leader must first appease the elites that best help maintain their power, such a promise may not be credible (Acemoglu and Robinson 2009; Fearon 1998; Kalyvas 2000; Powell 2006). Instead, the leader may distribute a disproportionate share of the benefits to their own followers. Moreover, the leadership may engage in actions that will weaken its rivals’ political influence in the future. The leader might fund multiple infrastructure projects for their followers while providing minimal investment in their rivals’ zones of control. Over time, the wealth accumulated by the leader relative to their rivals may translate into greater political power, thereby leaving rival leaders vulnerable to further marginalization. This demonstrates the internal commitment problem produced by state negotiation: leaders cannot credibly refrain from marginalizing rivals following a settlement. This suggests that the leader's rivals should oppose settlements and strive to prevent implementation. We might therefore expect rivals to escalate incendiary terrorist attacks against the government's population or punish the leadership and its followers.7 Strategically, the purpose of these attacks may be to weaken the leadership and create public outrage, thereby making it politically impossible for governments to honor commitments under the negotiated settlement. This behavior fits the common description of extremist violence following peace negotiations. The internal commitment problem offers a rationalist explanation for extremism, in which violence is motivated by rivals’ fear of marginalization at the hands of the leader. However, a key problem with infighting is that it weakens the movement collectively, thereby leaving the entire insurgency vulnerable to “divide and conquer” tactics (Cunningham 2011, 276). Infighting destroys collective resources and turns the factions’ focus from fighting the government to fighting each other. Further, infighting leaves the leader's faction vulnerable to exploitation. If the government strikes a deal that reflects the movement's power at time t, but the movement weakens due to infighting at time t + 1, the government may use the infighting opportunistically to conquer the entire movement. Consider the Sudanese government's (GoS) negotiations with the Darfuri insurgency in 2006. The Sudan Liberation Army (SLA) consisted of two groups, one from the Fur group led by Abdel Wahid, and the other from the Zaghawa group, led by Minni Minawi. In early 2006, the two wings of the SLA and the Justice and Equality Movement (JEM) met in Abuja to discuss a possible peace agreement. Minawi agreed to disarm his group and was appointed to the fourth highest position in the government in return. The excluded groups branded Minawi and embarked on a campaign of violence to prevent implementation of the deal. Responding to the violence, Sudanese President Bashir announced in September that the GoS would no longer comply with United Nations Security Council (UNSC) Resolution 1706, authorizing United Nations (UN) peacekeeping in Darfur. Without consulting Minawi, Bashir announced that the GoS would deploy to Darfur to suppress the outlaws. In this case, negotiations led to infighting, which enabled the GoS to divide and conquer the insurgency. Given that infighting leaves the insurgency vulnerable to exploitation, leaders may fear and reject negotiations, giving the appearance of “extremism” and “irrationality.” However, we do empirically observe that several insurgent leaders do agree to negotiate, despite the risk of infighting (Clarke and Paul 2014). The next step is to identify the specific conditions under which insurgent leaders refuse or are deterred from making peace and those under which leaders are willing to risk negotiation. To do so, let us turn to the formal model. Model Figure 1 presents a stylized model of a conflict between a government G and an insurgent movement M, consisting of a leadership A and their rivals B, who oppose G but maintain a different agenda than A. Assume both A and B initially agree to maintain cooperation, directing their violence at G. The payoffs to G, A, and B, respectively, for the status quo in which conflict occurs and there is no negotiation are equal to –w; pφw; (1 – pφ)w. Table 1 provides a summary of the model’s parameters. We represent gains made by the movement as w ∈ (0, 1). Since these gains come at G's expense, we represent its loss from conflict as –w. Since w represents joint gains by A and B, we assume the gains are divided based on the balance of power within the insurgency. The model represents the power of leader A relative to B as pφ ∈ (0, 1) and B's power relative to A as 1 – pφ. Although we assume A is more powerful than B at the start of the game (P > 1 – p),8 we assume two types of insurgencies: a static insurgency in which A is not facing a growing challenge from their rival B, and a competitive movement in which B is gaining strength and influence. To represent this distinction, we assume that φ = 1 in the former type, whereas φ = 2 in the latter. This indicates that A's power relative to B does not change in the former type since P1 = p, but A's power is declining in the latter type since P2 < p.9 Figure 1. View largeDownload slide Intrafactional conflict game with payoffs for G, A, and B, respectively. Figure 1. View largeDownload slide Intrafactional conflict game with payoffs for G, A, and B, respectively. Moves G begins the game by either continuing to fight or opening negotiations. If G negotiates, it offers a set of concessions w, which represents a level of concessions equal to what the insurgency would gain from continuing the conflict. Since A is in the leadership position, they decide whether to accept or reject G's offer. If leader A rejects the deal, the game ends in conflict, and A and B receive payoffs equal to (pφw, [1 – pφ]w), respectively.10 Notice that A's payoff is greater if the political situation in the movement is static (φ = 1) than if a leader is emerging in B to challenge their control (φ = 2). However, if leader A accepts G's offer, they take possession of the concessions G makes and are given the responsibility of distributing the concessions to the larger group. Let us assume that G receives a peace dividend if A agrees to negotiate, which is normalized to 1. If leader A accepts G's offer, the rival B next decides whether to accept the agreement or resort to terrorism to subvert it.11 If the rivals accept, we assume that they forfeit the ability to fight independently of leader A's faction.12 Leader A can then distribute the benefits of peace consistent with the balance of power in the insurgency or use their leadership position to seize control of the rivals’ gains.13 Despite any ex ante agreements, B will have no ability ex post to force leader A to share concessions; therefore, A will be unable to use promises of future concessions to sway B. Alternatively, the rivals may turn to terrorism in an effort to subvert the settlement before it consolidates. The attacks may be aimed selectively at civilians that support leader A or may be indiscriminate and target the government's civilians. Regardless, the purpose is to provoke the government into abrogating the agreement. Assume next that the direct costs of this terrorist violence are borne by both factions and that each one pays some cost c(A, B) ∈ [0, 1]. The provocative nature of the attacks also creates political pressure on G to retaliate and abandon the settlement. Assume that the political cost to G for exercising restraint and refusing to retaliate is represented by α ∈ [0, 1], which is an increasing function of the number of successful terrorist attacks B generates. Assume the number of successful terrorist attacks, k, is drawn from a Poisson distribution for which λ represents the mean number of successful attacks and is a function of the relative capabilities of the factions and the overall success of the movement.14 Formally, assume λB = 10(1 – pφ)w, which assumes that the maximum number of successful attacks is k = 10.15 This indicates that the expected number of successful terrorist attacks is a function of the rivals’ capability and the overall movement's success in fighting the government. The rivals produce more successful attacks if they are in a better military position vis-à-vis G (w→1) and if they are more powerful relative to A. Neither the government, leader A, nor rival B are certain how many successful attacks will materialize, but all are aware of the distribution from which the number is drawn.16 The use of N in Figure 1 denotes that this specific process—rather than a rival's strategic choice—determines the number of successful attacks. Figure 2 maps a hypothetical distribution of successful attacks initiated by B if G negotiates with A. When p = 0.4, A is relatively weaker than B. This means if G negotiates with leader A, it can expect that B can respond with a greater number of successful terrorist attacks. The mean number of successful attacks in this case is 4.2, indicating G should expect around four attacks. By contrast, if A is relatively stronger and p = 0.8, the mean number of attacks that G should expect in response to negotiations falls to 1.4. All else equal, excluding stronger factions from negotiation is likely to produce more terrorism, relative to excluding weaker factions. Figure 2. View largeDownload slide Distribution of number of successful terrorist attacks by B as a function of the balance of power in the movement (p). Note: Squares when p = 0.4, indicating A is relatively weaker than B; circles when p = 0.8, indicating A is relatively stronger. Figure 2. View largeDownload slide Distribution of number of successful terrorist attacks by B as a function of the balance of power in the movement (p). Note: Squares when p = 0.4, indicating A is relatively weaker than B; circles when p = 0.8, indicating A is relatively stronger. Assume G's payoff for refusing to retaliate and holding to a negotiated settlement is equal to 1 – w – kα. G receives the benefit of peace, minus the value of its concession, and minus the political cost –kα for failing to respond to the attacks, where k ≥ 0 represents the total number of successful attacks initiated by the rivals. After observing successful attacks, G may uphold the agreement, as the British upheld the Good Friday Agreement following the Omagh bombing in 1998. Doing so requires G to pay the political penalty α, which is increasing in the number of successful attacks. Alternatively, G can avoid paying the political price by abrogating its commitment to the collaborators. Abrogating the agreement at this stage is also advantageous due to the insurgency's infighting, which weakens its ability to fight and survive G's repression. We represent the power of the divided insurgency to resist G as w2. This indicates that G receives a payoff of –w2 if it resorts back to violence and a payoff of 1 – w – kα if it upholds the settlement. Solution Because the game is characterized by complete yet imperfect information, we solve it using the subgame perfect solution concept (SPE).17 This section characterizes the main insights behind the solution, leaving most of the formal details to the appendix. We begin at the game's final move (i.e., the government's decision to uphold or abrogate the settlement). Since negotiated settlements produce a peace dividend, the government strictly prefers to reach a settlement. However, if the rivals resort to terrorism, the government may be unable to fulfill its bargain for two reasons. First, the attacks may make it politically impossible for the government to refrain from violence (Bueno de Mesquita 2007). Second, infighting increases G's temptation to renege in favor of repressing the entire insurgency. Since G's payoff is decreasing in w, its payoff for conflict increases from –w against a unified movement to –w2 against a divided movement. Therefore, infighting may benefit the government, in that it fractures the movement and leaves it vulnerable to divide and conquer tactics.18 Formally, G abrogates peace in response to terrorist violence if –w2 > 1 – w – kα, which simplifies to   \begin{equation}{\rm{k}} > ( {1-{\rm{ w}} + {{\rm{w}}^2}} )/\alpha \end{equation} (1) We define k* = (1 – w + w2)/α. This indicates that the rival fails to stage enough successful attacks to sabotage the settlement if k ≤ k* and generates enough attacks to destroy the settlement if k > k*. If the government negotiates, the probability the peace deal holds is   \begin{equation}{\sigma _{\rm{B}}} = \Gamma ([ {{{\rm{k}}^*} + 1} ],{\lambda _{\rm{B}}})/[ {{{\rm{k}}^*}} ]!\end{equation} (2) The Rivals Rival B recognizes that accepting the deal will result in marginalization at the hands of A.19 Alternatively, resorting to terrorism has several potential drawbacks. First, terrorism is costly. Second, there is no guarantee B can stage enough successful attacks to scuttle the deal.20 Third, infighting weakens the entire insurgency, leading to fewer gains from conflict. B turns to terrorism and plays reject if   \begin{equation} {\sigma _{\rm{B}}} < ([1 - {{\rm{p}}^\varphi }]{{\rm{w}}^2} - {{\rm{c}}_{\rm{B}}})/([1 - {{\rm{p}}^\varphi }]{{\rm{w}}^2})\end{equation} (3) We define the critical probability that B fails to generate enough attacks to sabotage peace as σB* = ([1 – pφ]w2-c B)/([1 – pφ]w2). As σB decreases, B is increasingly likely to turn to terrorism. Comparative statics indicate that B is less likely to resort to terrorism if cB→1 and more likely to do so if cB→0. The cost of violence is often interpreted as a measure of resolve, or willingness to engage in conflict. In this case, cB may represent the animosity between the leader A and rival B. Empirically, if the leaders’ preferences are highly divergent, B is more likely to respond violently to a settlement between G and A. Condition 3 demonstrates that B responds with terrorism if the conflict is going relatively well for the movement (w→1). Infighting is therefore less likely when the larger insurgent movement is weak and insignificant. If these rivals were to succeed in sabotaging peace, which they are less likely to do given the lower value of w, infighting would effectively undermine the ability of these groups to combat the government. Moreover, governments are less likely to uphold peace deals with weaker movements, given that these deals are less costly to make. The balance of power between the leader and their rivals also influences B's decision to resort to violence.21 If leader A commands a significantly larger fighting force than B, and controls most of the insurgency's military power (p→1), B is likely to be deterred from terrorism. In such cases, attacks are costly, but unlikely to undermine the settlement. B is better off accepting the distasteful deal while sparing the cost of staging attacks. Formally, B is deterred and accepts G's deal with A if p ≥ 1 – (cB/w). However, if B maintains a larger share of military power, it is possible for B to resort to terrorism following negotiations. Leaders Leader A anticipates whether B will resort to terrorism. If A is preponderant in the movement (p→1), B will acquiesce, and A strictly prefers to accept G's offer to negotiate, since w > pφw. However, if A's forces are not powerful enough to deter B from rejecting the deal, making peace is risky. Alternatively, if A refuses to negotiate, they receive a payoff of pφw for continuing the conflict. A therefore accepts if σB(w) + (1 – σB)(pw2) – cA ≥ pφw. Simplifying,   \begin{equation} {\sigma _{\rm{B}}} \ge ({{\rm{c}}_{\rm{A}}} + {\rm{w}} [{{\rm{p}}^\varphi } - {\rm{pw}}])/( {{\rm{w}} - {\rm{p}}{{\rm{w}}^2}} )\end{equation} (5) We define A's critical probability that a settlement holds as σB** = (cA + w[pφ – pw])/(w – pw2). A accepts G's offer if σB ≥ σB** and rejects it otherwise. We use comparative statics to examine what key factors affect leader A's decision to accept. Here, the critical variable is pφ, or the power A's forces will control if A refuses to negotiate. If power within the movement is static (φ = 1) and A's position is stable, the right-hand side is larger, and condition 5 is more difficult to fulfill. If A's power is declining (φ = 2), condition 5 is easier to fulfill. Figure 3 plots the critical probability σB** as a function of pφ. The solid line represents the case in which A's power is static, while the dashed line represents the case in which A's political influence is diminishing relative to B's. A is less likely to accept G's offer if their leadership will not be challenged, but more likely to do so if they face a political threat from a competitor in B. However, since A is presently more powerful within the movement than B,22 A has little reason to risk an internal conflict if B does not appear politically threatening. Alternatively, if A's power is diminishing relative to B's (φ = 2) and they reject negotiations, their gains from conflict are diminished since B is surpassing them in power. However, if A accepts, they may seize control of all gains from conflict. If they face a viable competitor and their power is in decline, leader A has the greatest incentive to strike a deal immediately. Figure 3. View largeDownload slide Threshold for insurgent leader A to accept as a function of its power relative to B (pφ). Figure 3. View largeDownload slide Threshold for insurgent leader A to accept as a function of its power relative to B (pφ). Equilibrium23 Figure 4 maps the equilibria as a function of the balance of power within the movement (p). The game has a unique SPE that varies according to different configurations of parameter values. Let us first consider the behavior of the players when A is preponderant. If leader A is dominant within the movement, they reject government offers so long as the movement is static, but accepts if power is fluid. On the other hand, leader A will accept G's offer if B is deterred from engaging in terrorism, regardless of whether power within the movement is static or fluid. Equilibrium 1: Governments and insurgent leaders can credibly negotiate without facing attacks from rival insurgents if the leader is sufficiently dominant over all other rivals within the insurgent movement.24 Figure 4. View largeDownload slide Equilibrium behavior as a function of p in static and dynamic cases. Figure 4. View largeDownload slide Equilibrium behavior as a function of p in static and dynamic cases. In the case where leader A is relatively weaker, B will resist any deal with violence. Therefore, if the movement is static and A's future gains will resemble current gains, A is deterred by B's threat and rejects if A is relatively weak. A is unwilling to risk conflict with B given that their payoff for continuing to fight together remains relatively higher. Interestingly, this suggests that if A does not have secure control over the movement, but does not believe that they will be challenged politically in the absence of peace efforts, they are likely to resist any efforts at peace. It follows then that insurgent leaders who do not face growing political competition and have moderate control over their movement should be the most intransigent in negotiation and may appear extremist or irrational by rejecting any efforts at peace. The model suggests this refusal to negotiate is not motivated by irrationality, but rather by a fear of infighting that will result from negotiation. This may explain the behavior of various terrorist leaders, such as the late Mullah Omar and Abubakr Shekau, who face little political competition but wield only moderate control over their movements. These leaders may fear that negotiation will cause an internal conflict that will leave their entire movement vulnerable, a perception that deters them from negotiation. However, if these leaders do face political competition, they may risk infighting in order to hold onto their power. Facing internal challenges, these leaders choose to bandwagon against their competitors by striking deals with the government. Equilibrium 2: Governments and insurgent leaders cannot reach negotiated settlements if no shifts in power are occurring and the leader is near parity with their rivals in the movement.25 Equilibrium 3: Governments and insurgent leaders reach negotiated settlements if power is shifting away from the leader and the leader is near parity with their rivals in the movement.26 Implications The model produces three hypotheses. First, leaders who maintain a preponderance of power within their movements are free to negotiate without fear of infighting. These leaders can reach Pareto efficient deals with governments and should seek to do so as peace is preferable to continued conflict. H1:Politically powerful leaders are more likely to negotiate. Alternatively, if leader A is only moderately powerful relative to their potential competitors, their rival B escalates violence in response to any attempt by leader A to negotiate. Since the rivals are not weak, their campaign is more likely to provoke G into abrogating the agreement, thereby allowing G to fight a weakened, fractured movement. Given this possibility, A may be deterred from striking an efficient deal with G. If A is only moderately powerful, but the movement is static and lacks an alternative to the leader, A will refuse to negotiate with G. A has no incentive to risk infighting when their political control is uncontested, yet vulnerable to prospective challengers. However, if leader A faces competition from B, inaction risks a future loss of control over the movement. The model therefore predicts that, while moderately powerful but politically unchallenged leaders will adopt hardline stances, those facing competitors may strike deals to lock in control over their movements. H2:Moderately powerful leaders are less likely to negotiate if their rivals are not increasing in power. H3:Moderately powerful leaders are more likely to negotiate if their rivals are increasing in power. The Case of Nigeria To illustrate, we examine Nigeria's conflicts with two prominent insurgencies, Boko Haram and the Movement for the Emancipation of the Niger Delta (MEND).27 These cases allow us to hold constant both the target government's capacity and time period, as both groups began in the early to mid-2000s. Also, foreign state influence is not a factor in either conflict. In both movements, factions have used violence in similar ways. We therefore contend there are no moderate factions in either movement.28 Government attempts at negotiation, however, have met different levels of success. Both Boko Haram and MEND have geographic influence. Boko Haram's presence extends to many northeastern states in Nigeria, including Borno, Yobe, Katsina, and Bauchi. MEND's core influence is in the South, largely throughout the Niger Delta. The two groups maintain a fairly diffuse power structure, and power is not equally distributed among their factions. Within Boko Haram, Abubakar Shekau leads a thirty-member Shura council that directs multiple scattered cells, many of which never interact (Walker 2012, 8). The International Crisis Group (ICG) reports that, of Boko Haram's six factions, Shekau's is the largest and, at the time of the report, had been responsible for the majority of a recent spate of attacks (ICG 2014). MEND similarly consists of multiple factions with no real hierarchical structure, though some factions appear more powerful. In contrast to Boko Haram, the organization does not have a single readily identifiable leader (Courson 2011, 31). During the heavy fighting that occurred between 2006 and 2009, major leaders in MEND included Henry Okah, Tom Polo in the Delta State, and General Boyloaf in the Bayelsa state (Amadi, Imoh-Itah, and Obomanu 2016, 176). These three leaders appeared to be the most significant in the organization. However, MEND also consisted of many smaller factions, such as those led by Ateke Tom, Egberi Papa, and Asari Dokubo (Amadi, Imoh-Itah, and Obomanu 2016, 177). Power in MEND was neither equitably distributed nor concentrated in any one leader. Instead, power fluctuated between factions as fighters’ loyalties shifted. Both Boko Haram and MEND fall into the “moderately powerful” category of armed organizations, where some factions are somewhat more powerful than others, but none is preponderant. The two groups adhere to very different ideologies. While Boko Haram is Salafi Jihadi, the bulk of MEND's activity revolves around resisting the operations of Western oil companies (Courson 2011, 33). The group's individual factions have a range of diverse objectives and ample access to economic resources through oil bunkering, ransom, protection money, and gun running (Akinola 2011, 76). According to Weinstein's (2007) logic, this suggests that, relative to Boko Haram, MEND is more likely to recruit and attract opportunistic insurgents who, having been initially attracted by the promise of personal enrichment, are more inclined to shift their loyalties to a rival faction or otherwise challenge leadership when doing so promises material gain. Not only does Boko Haram's relative lack of resources limit opportunism, but Berman (2009) argues that radical religious organizations are better equipped to screen out opportunists and build loyalty to the group and its leadership among their cadre or recruits. According to Berman, groups like Boko Haram should see greater cohesion and membership stability because they require members to make very costly initial sacrifices that screen out opportunists likely to defect to other factions, while including those who are loyal to the group and its leadership, and because members become more dependent on the group as they are separated from the outside world both physically and through adopting the extreme lifestyle dictated by the group's religious prohibitions (2009, 81). Thus, if we assume that these recruitment mechanisms remain fairly constant across factions, we should expect fewer rapid power shifts within Boko Haram, since its members are less likely to challenge leadership, defect to opposing factions, create their own opposition forces, or seek material rewards. We should expect therefore to see more stable factions and few leadership challenges within Boko Haram and, conversely, more shifts among the MEND factions where material incentives to switch exist, loyalty to faction leaders is expected to be lower and opportunism is expected to be higher. Indeed, leadership stability should be much higher in Boko Haram, compared to MEND. Boko Haram: No Viable Political Competition By most accounts, Boko Haram is considered a cohesive organization that operates through a cellular structure, wherein Shekau passes instructions via trusted leaders (BBC 2014). The organization's roots are in the Yusufiyya, a movement founded in 2002 by Mohammed Yusuf that is often called the Nigerian Taliban for its aim to purify Islam in the country (ICG 2014, 9). The Yusufiyya took on a state-like role, engaging in the care of refugees and religious policing. Around 2007, it began to conduct bombings and targeted assassinations, and Yusuf was eventually captured and killed in 2009. At that time, Shekau took control of the movement, and it became known as Boko Haram (ICG 2014).29 As members who were exiled following Yusuf's death returned to Nigeria in 2010, they raided the homes of those who had cooperated with the police, and killed families who refused to leave the confiscated dwellings of Boko Haram agents. Boko Haram also began to supplement its cash flow through robbery. However, rather than shifting its focus toward personal enrichment of leaders or members, Boko Haram goes to great lengths to free captured operatives, provides for widows, and pays and supplies active fighters more consistently than the Nigerian armed forces (Campbell 2014). Although Boko Haram supplements its income through crime, the group seems to have no significant external funding sources (Campbell 2014).30 While Boko Haram eventually gained resources through theft, its limited initial resource access made it more dependent on member sacrifices. Evidence suggests that early members of the Yusufiyya and Boko Haram were anything but opportunistic and were instead highly loyal to their leaders. Walker writes that a senior member explained “more than anything else, it was what Yusuf revealed to him about the Quran that convinced him to throw his lot in with the group, give up his job, and bring his family to live in the mosque” (2012, 9). The combination of limited initial resources, radical religious motivations, heavy initial sacrifices, and isolation from outside influences is expected to limit the level of opportunism among group members and produce a more cohesive group with little shirking. Indeed, Boko Haram's ability to carry off sophisticated attacks, such as a 2010 jailbreak freeing more than seven hundred inmates, speaks to the commitment of its members (Sergie and Johnson 2014). Following Yusuf's death, internal divisions formed between nationalist elements under Shekau and factions more focused on transnational activities that also maintained ties to regional groups. A splinter group by the name of Ansaru emerged following a Boko Haram attack that killed nearly two hundred Nigerians in a Muslim-majority area in January 2012. Its formation was in part a reaction to Boko Haram's practice of killing other Muslims, particularly defectors, and also the result of the growing nationalist-transnational divide within the organization (Bey and Tack 2013; Zenn 2013). While Ansaru initially distanced itself from Shekau and his practice of killing other Muslims, a subsequent Ansaru leader, al-Barnawi, began to cooperate with Shekau (Zenn 2013; ICG 2014). Ansaru and Boko Haram also coordinated joint kidnapping and ransom efforts, with Shekau providing the manpower (Zenn 2013; ICG 2014, 22).31 A report from the ICG also goes on to say that, while Shekau “nominally controls fighters led by rival leader Mamman Nur ... most commanders and foot soldiers are more loyal to Nur ... but [Nur] chooses to maintain the status quo because of Shekau's ruthlessness” (ICG 2014, 22). These dynamics highlight Shekau's vulnerability to potential infighting. Furthermore, “both al-Barnawi and Nur lack sufficient grassroots networks in Borno, where Boko Haram carries out more than 80 percent of its attacks,” for them to be able to lead the group were Shekau to be killed or captured (Zenn 2013, 29). Although there are no direct challenges to Shekau's position of power, his loyalist faction does not maintain absolute control over the organization. Our model predicts that the loose structure of the organization, combined with the absence of challengers, should cause Shekau to adopt a hardline, intransigent view toward negotiation. This appears correct, as Shekau often seems to lack a clear set of demands and appears more interested in continuing violence, rather than negotiating. Citing an interview with an intelligence officer, the ICG reports that Shekau orders the killing of members who propose negotiations, thereby “silencing other prodialogue individuals” (ICG 2014, 21). Though Shekau makes every effort to appear powerful, his behavior suggests that he maintains less control over the organization than many analysts believe. This is evident through both the presence of multiple factions within Boko Haram and a 2012 video in which Shekau embraces al Qaeda. The latter in particular is indicative of a shift away from the Shekau faction's previously exclusive focus on domestic dynamics and toward one of a more regional nature embraced by rival factions (ICG 2014, 28). This shift may have been intended to prevent Ansaru, who aligned with al Qaeda in the Islamic Maghreb (AQIM), from gaining ground. These shifts in organizational focus indicate that if Shekau were to accept an offer of amnesty or settlement from the Nigerian state, Boko Haram may become vulnerable to infighting. Given these risks, Shekau is more likely to prefer avoiding a potentially divisive settlement. Absent outright challenges to his power, Shekau appears free to continue his terrorist campaign while avoiding risky settlements with the Nigerian state. The unwillingness of Boko Haram to come to the negotiating table, as well as the group's practice of cracking down on those who reach out to the state without the leadership's authorization to do so, is consistent with our model's characterization of a moderately powerful leadership. These circumstances suggest that, while various factions may not totally embrace a group leader's control, there are no viable challengers that threaten the overall internal balance of power.32 MEND: A Competitive Political Environment Unlike Boko Haram, which is characterized by the presence of weak factions, more powerful factions within MEND that fear their own future decline should be more likely to seek opportunities to negotiate with the state, at the expense of weaker factions. MEND coalesced in 2005 from an assortment of opposition groups in the Niger Delta with the stated goals of ensuring the return of a larger share of the proceeds from the region's oil wealth to its people and the withdrawal of government troops from the region (Courson 2011). MEND factions were united in their opposition to oil companies, but divided in their individual objectives. The lucrative nature of MEND's activities attracts both ideological and greed-motivated fighters (Akinola 2011, 76). The combination of low entry barriers, great resource wealth, and numerous autonomous factions results in dynamic power shifts among factions. MEND has identifiable faction leaders, but no central command structure. Formed from rival groups, MEND maintains many factions operating with a high degree of autonomy, which cooperate when it is mutually beneficial to do so (Courson 2011). In 2009, when faction leader Henry Okah faced capital charges, the Nigerian government found itself facing a large violent movement with no faction sufficiently powerful to deter others from cooperating with the government. Taking advantage of this opportune moment in the internal structure of the armed group, President Yar'Adua initiated negotiations with Bayelsa state faction leaders and announced an amnesty program. On August 7, Bayelsa's leader Boyloaf became one of the first to accept amnesty, citing the government's release of Okah and the offer of development programs as justification. Boyloaf's acceptance prompted other MEND leaders to announce his replacement. According to a Vanguard report from July 24, 2009, Boyloaf charged Ijaw youth leaders in the area with attempting to persuade terrorist leaders that the government offer was insincere, and the news outlet reported on August 10, 2009, that eleven MEND leaders in Delta State had disowned Boyloaf (Boyloaf accuses Ijaw 2009; Cracks in militants’ camp 2009). By year's end, many other major leaders had negotiated amnesty agreements. These agreements offered large payouts and protection contracts to critical leaders including Ateke Tom (leader of the Icelanders gang and the Niger Delta Vigilante Movement in Rivers State), Tompolo (leader of a Delta State faction), and Boyloaf. Each was paid to secure the oil lines they had previously attacked (Amadi, Imoh-Itah, and Obomanu 2016, 178). Job training and smaller payments, often to be distributed by commanders, were offered to lower-ranking fighters. However, these concessions did little to satisfy MEND's initial demands and were thus rejected by some factions. While a few top commanders like Tompolo and Boyloaf became increasingly wealthy and powerful, many in the region, including former terrorists, did not benefit. Why were the leaders willing to negotiate despite the risk of infighting? Theoretically, the model predicts that moderately powerful leaderships will negotiate if they face political challengers. Since the power of MEND commanders was solely based on their control of territory and wealth associated with the oil trade, the fluidity of power guaranteed that other commanders would crop up to challenge the old guard. Therefore, in the MEND case, the presence of oil inhibits the sort of leadership stability that we argue leads to intransigence in other cases.33 This is because oil creates material incentives for opportunistic individuals to join or form militant groups to which they do not have strong loyalties. Opportunistic rebels in turn are more likely than those that feel personal loyalty to the group and its leaders to defect from one group and join or form a rival if doing so promises material gains. Even as the amnesty concluded, the old guard faced challenges from individuals who claimed that the deal did nothing to stop oil pollution or address income inequality in the South. Several of these individuals claimed that they would go back to the creeks and resume the campaign of violence (Ajibola 2015). Had Boyloaf, Tompolo, and Tom continued with insurgency, there was no guarantee that they would not have been replaced by the myriad other “generals” within the MEND organization. Therefore, each of these leaders strategically decided to negotiate before losing all of their power to potential competitors. While the payments and job training programs calmed the violence temporarily and included an estimated twenty-six thousand former militants as of 2011 (Canada 2011), the amnesty did little to address underlying issues of the insurgency, and many former militants remained unsatisfied. Consistent with the model's predictions, the excluded militants who did not benefit from the program responded by escalating terrorist violence, breaking MEND's 2009 truce with attacks specifically targeting amnesty discussions in the Delta State. A March 15, 2010, BBC report quotes a Delta State spokesperson as responding with the following: “I think the intention is obvious, just to scuttle the talks and make it seem as if Warri in Delta State is not safe” (qtd. in Explosions hit Nigeria 2010). A Reuters report from the same day notes that, if MEND had lived up to its threats of “renewed attacks on oil installations in the coming days,” this could have derailed the amnesty and Nigeria's economy (Tattersall 2010). MEND's shifting and loosely aligned factions enabled the Nigerian government to cut deals with leaders of sufficient strength to weaken the overall movement. While no leader was powerful enough to deter others from accepting lucrative security contracts from the government, those who accepted the government's deals were insufficiently powerful to stop others from continuing the campaign and engaging in spoiler violence. Furthermore, the movement's multiple factions and fluid factional boundaries incentivized leaders to cut deals with the government to avoid losing out later as their relative power declined. Consistent with our expectations, we observe the government negotiating with more powerful factions within a dynamic insurgency and with factions that stand to lose some of their influence within the movement. In response, however, we also observe rivals engaging in terrorist attacks to scuttle the deal. Conclusion We began with the question: Why do insurgent leaders resist bargaining, even when governments appear inclined to offer concessions to resolve conflicts? We argue that this resistance to negotiation may be motivated by internal politics within the insurgency. Since individuals join these movements for a variety of reasons, the goals of the leadership may not fully represent the goals of the individual members within the organization. Therefore, if an insurgent leader were to reach an agreement with the government, the leader may be tempted to horde all of the concessions for his followers while excluding members from competing factions. We argue that this incentive on the part of the leader creates a commitment problem within the violent movement and may lead to infighting following peace. Given this anticipation and the expectation that infighting may tempt the government to pursue a divide-and-conquer strategy, insurgent leaders may resist all efforts at negotiation. Although such leaders may appear behaviorally “irrational” or “extreme,” we argue that they are strategically dodging internal conflicts that may occur following negotiations. However, we demonstrate that leaders may be willing to strike deals and risk infighting when it appears that they will face political competitors within their movement. In these cases, leaders that believe their political power is waning may seek to lock in their control over the movement by making peace, which in turn may trigger an internal power struggle between the leadership and its rivals. This research highlights the importance of examining the internal politics of violent opposition movements. Like states, it is conceivable that attempts to retain power motivate many leaders of violent groups, particularly in cases where their politics are competitive. Table 1. Parameters Parameter  Interpretation  w  Gains made by insurgent movement (A, B) in conflict  p  Insurgent leader's faction A share of power  1-p  Set of other faction's B share of power  cA  Cost to A for infighting  cB  Cost to B for infighting  ε  Political cost to G for A's rejection  α  Political cost to G for failing to respond to terrorist attacks  φ  Shift in political influence  k  Successful terrorist attacks  σB  Probability B fails to sabotage agreement if excluded  Parameter  Interpretation  w  Gains made by insurgent movement (A, B) in conflict  p  Insurgent leader's faction A share of power  1-p  Set of other faction's B share of power  cA  Cost to A for infighting  cB  Cost to B for infighting  ε  Political cost to G for A's rejection  α  Political cost to G for failing to respond to terrorist attacks  φ  Shift in political influence  k  Successful terrorist attacks  σB  Probability B fails to sabotage agreement if excluded  View Large Footnotes 1 Insurgency refers to “a struggle for control and influence, generally from a position of relative weakness, outside existing state institutions” (Department of the Army 2014, 1-1). Following Moghadam, Berger, and Beliakova (2014), we conceive of insurgencies as being of sufficient size and power as to be able to control territory. Following Bakke et al., we define an insurgent movement as one that “appeals to a shared identity;” as Bakke et al. note, this does not presume that “movements necessarily possess ‘common purposes and solidarity’” (2012, 266–267). 2 In one study that appears to be the exception, Lyall (2009) finds that indiscriminate shelling from Russian forces assisted in reducing Chechen terrorist attacks. 3 We use faction and group to refer to distinct subunits of an insurgent movement that share some common interest as part of the movement and attempt to appeal to a common identity, but have different preferences regarding policy, leadership, or tactics. Cell refers to an operational unit within a group or faction (see, e.g., Bueno de Mesquita 2005; Department of the Army 2014). 4 Alternatively, Abrahms (2008, 101) argues that terrorists may resist negotiation not because of extremism, but because individuals in these groups are “social solidarity maximizers.” While this is not conceptually the same as extremism, we would behaviorally expect both types of individuals to resist any form of peace, even if it is Pareto-improving. Similarly, Ross (2006), Collier and Hoeffler (2004), and Weinstein (2007) argue that greed-driven rebels often fight for profit rather than political goals, giving these fighters little incentive to end conflicts. 5 For a discussion of fragmentation in rebel movements, see Bakke, Cunningham, and Seymour 2012; Christia 2012; Cunningham 2006; Cunningham 2011; Cunningham 2014; Gates 2002; Krause 2013; Rudloff and Findley 2016; Seymour, Bakke, and Cunningham 2016; Staniland 2012. 6 This negotiating challenge is apparent in many contexts. Putnam, for example, observed that rather than states being unitary actors as they are often modelled, they are instead comprised of actors with differing preferences, noting that “the unitary actor assumption is often radically misleading” (1988, 433). Indeed, this is also true with regard to insurgencies, as has been observed previously in the literature addressing cleavages within such movements (see in particular Journal of Conflict Resolution 56, no. 1, a special devoted to fragmentation of nonstate actors). 7 This strategy of using terrorism as provocation is discussed by several authors, including Bueno de Mesquita (2005); Bueno de Mesquita and Dickson (2007); de Figueiredo and Weingast (2001). 8 p > 0.5. 9 For simplicity, we analyze the two cases where φ = 1 and φ = 2. An alternative strategy is to allow φ to take on a continuous value. This strategy produces a cutoff value φ* that divides into cases where φ < φ* and cases where φ ≥ φ*. The behavior identified in the first case is the same as when φ = 1, and the behavior in the latter case is the same as when φ = 2. The results are robust to the alternative specification. 10 G receives a payoff of –w–ε if its attempt at negotiation fails. This indicates that, if G is aware that A will reject, G prefers to play ∼negotiate. 11 To maintain tractability, assume that, in cases where there are N factions, if one faction rejects the deal, the entire set in B rejects the deal as well. 12 We make this assumption both to ease analytic tractability and to focus on the cases of interest. The assumption stems from the disarmament and demobilization process, which the United Nations counsels should occur early in peace processes. Even where local autonomy is granted, the maintenance of order depends upon disarming potential armed opponents to the leadership. 13 This sequence is similar to Fearon (1998), except that the division of w is not endogenous. An earlier version of the model that allowed A to endogenously divide w with B yields similar results. Please see Part IV in the appendix for this discussion. 14 We assume that attempts are not individually costly and that there are no diminishing returns of additional attacks. Therefore, B attempts as many attacks as it can, but only some attacks are successful. 15 This number is arbitrary, but is reasonable given the current literature. Theoretically, the maximum number of attacks could be any number. 16 This feature allows the insurgent leader to face some uncertainty regarding how powerful their rivals are in terms of marshalling attacks to oppose negotiations. 17 See formal appendix for proofs. 18 For empirical discussions, see Findley (2008) and Cunningham (2011). 19 As noted earlier, this results from the assumption that rival B forfeits the ability to fight independently of A when they acquiesce to the deal A has cut with the government. As a rational B should never anticipate that a rational A would honor any commitment to share the concessions (assuming that A and B have different preferences on at least some issues including leadership of the movement, meaning that A's utility would be declining in the amount of resources shared with B), we exclude this node in our analysis of the game for greater tractability. The model's results are robust to the inclusion of this node. 20 As noted above, attacks may target the civilians or leader A's faction. We refer to these attacks as infighting within the movement because, regardless of the target of the attack, the purpose is to scuttle A's deal with the government. 21 Similarly, Greenhill and Major argue that the relative power balance between “competing factions on the ground and those implementing the peace” is the most critical determinant of whether spoiler violence will occur (2006/2007, 9). 22 p→1 – (cB/w) 23 See appendix for proofs. 24 Formally, if p ≥ 1 – (cB/w2), {G: Negotiate; A: Accept; B: Accept} is an equilibrium pathway. 25 If 1 – (cB /w2) > p > 1 – (cA/w)1/φ and φ = 1, A plays reject. 26 If 1 – (cA/w)1/φ > p and φ = 2, A plays accept if σB > σB** and reject otherwise. 27 While this does not substitute for a systematic empirical test, the cases illustrate the model's causal mechanisms. 28 While Ansaru is often considered a less extreme faction of Boko Haram, it has allied with al Qaida in the Islamic Maghreb and the Islamic State (ISIS). The faction is also known for kidnapping and killing Western civilians in Nigeria Chothia 2013; Elbagir et al. 2015; Ajibola 2015. 29 Walker (2012) notes that the name Boko Haram is commonly applied to the group as a pejorative. The group's name for itself is Jama'a Ahl as-Sunna Li-da'wa wa-al Jihad. 30 Though the ICG's (2014) report indicates that, in 2002, the Yusuf and his movement received the bulk of a $3 million budget provided by Bin Laden for Nigerian Islamist groups. 31 Zenn's account and the ICG differ on whether Barnawi was still alive in 2013, with Zenn writing as if he were. The ICG report cites an interview with a Boko Haram member in August 2013, claiming that al-Barnawi was killed in an August 2012 security forces raid of his hideout, which resulted from his cooperation with Shekau in attempting to take out a faction leader that both suspected of being a government mole. 32 In 2011, an individual claiming to represent the group told BBC that the group would disarm if certain conditions were met. A Boko Haram spokesman quickly denied the deal and claimed “internal divisions were eliminated” (quoted in Walker 2012, 11). Similarly, Yusuf's brother-in-law was assassinated after meeting with former President Olusegun Obasanjo in an apparent effort to discuss peace terms. In January 2012, a faction claiming to be more moderate declared on television that it was ready to negotiate; these individuals were quickly beheaded. In October 2014, the Nigerian government announced it had reached a ceasefire agreement with a previously unknown Boko Haram leader. There was no public confirmation through Boko Haram's usual channels. By the following day, suspected Boko Haram terrorists had carried out five attacks. 33 While, in this case, it is control of a natural resource that produces incentives for rivals to emerge, it is important to note that such natural resources are not necessary to produce such a scenario. For example, viable rivals might emerge as the result of state sponsorship of rebels or of questionable leadership decisions coupled with effective counterinsurgency by the state, as in the case of the Algerian Islamic Group (GIA), which lost control of its movement after using massive indiscriminate violence, which resulted in the defection of many of its followers to a splinter group, the Salafist Group for Call and Combat (GSPC). Supplementary Information Supplementary information is available at the Journal of Global Security Studies data archive. References Abrahms Max. 2008. “What Terrorists Really Want: Terrorist Motives and Counterinsurgency.” International Security  32 ( 4): 78– 105. Google Scholar CrossRef Search ADS   Acemoglu Daron, Robinson James A.. 2009. Economic Origins of Dictatorship and Democracy . Cambridge: Cambridge University Press. Ajibola Iyabobola O. 2015. “Nigeria’s Amnesty Program: The Role of Empowerment in Achieving Peace and Development in Post-Conflict Niger Delta.” SAGE Open July/September: 1-11. Akinola Adeoye O. 2011. “Niger Delta Crisis: The Nexus between Militants’ Insurgency and Security in West Africa.” African Security  4 ( 1): 65– 80. Google Scholar CrossRef Search ADS   Amadi Luke, Imoh-Itah Imoh, Obomanu Edmund F.. 2016. “Oil: Niger Delta and the Making of Militia Capitalists: A Theoretical Assessment.” International Journal of Social Science and Humanities Research  4 ( 1): 172– 84. Bakke Kristin M., Cunningham Kathleen Gallagher, Seymour Lee J. M.. 2012. “A Plague of Initials: Fragmentation, Cohesion, and Infighting in Civil Wars.” Perspectives on Politics  10 ( 2): 265– 84. Google Scholar CrossRef Search ADS   Bapat Navin A., Bond Kanisha. 2012. “Alliances between Militant Groups.” British Journal of Political Science  42 ( 4): 793– 824. Google Scholar CrossRef Search ADS   BBC. 2014. “Nigeria's Boko Haram Leader Abubakar Shekau in Profile,” May 9. http://www.bbc.com/news/world-africa-18020349. Berman Eli. 2009. Radical, Religious, and Violent: The New Economics of Terrorism . Cambridge: MIT Press. Bey Matthew, Tack Sim. “ The Rise of a New Nigerian Militant Group.” Stratfor Global Intelligence, last modified February 21, 2013, accessed on July 20, 2017. http://www.stratfor.com/weekly/rise-new-nigerian-militant-group#axzz37DJvltbY. Bueno de Mesquita Ethan. 2005. “Conciliation, Counterterrorism, and Patterns of Terrorist Violence.” International Organization  59 ( 1): 145– 76. “Boyloaf accuses Ijaw leaders of inciting militants not to accept amnesty.” 2009. Vanguard, July 24, 2009. https://www.vanguardngr.com/2009/07/boyloaf-accuses-ijaw-leaders-of-inciting-militants-not-to-accept-amnesty/. “Boyloaf accuses Ijaw leaders of inciting militants not to accept amnesty.”. 2007. “Politics and the Suboptimal Provision of Counterterror.” International Organization  61 ( 1): 9– 36. Bueno de Mesquita Ethan, Dickson Eric. 2007. “The Propaganda of the Deed: Terrorism, Counterterrorism, and Mobilization.” American Journal of Political Science  51 ( 2): 364– 81. Google Scholar CrossRef Search ADS   Campbell John. “Nigeria: Why are Boko Haram Fighters Successful?” Africa in Transition (blog).  Council on Foreign Relations, March 4, 2014. http://blogs.cfr.org/campbell/2014/03/04/nigeria-why-are-boko-haram-fighters-successful/ Canada: Immigration and Refugee Board of Canada. 2011. Nigeria: Government Amnesty Program for Niger Delta Militants, Particularly with Respect to the Movement for the Emancipation of the Niger Delta (MEND) (2009–August 2011) . NGA103789.FE. http://www.refworld.org/docid/50740a192.html.Chai. Chai Sun-Ki 1993. An Organizational Economics Theory of Antigovernment Violence. Comparative Politics  26 ( 1): 99– 110. Google Scholar CrossRef Search ADS   Chothia Farouk 2013. “Profile: Who are Nigeria’s Ansaru Islamists?” BBC News, March 11, 2013. http://www.bbc.com/news/world-africa-21510767. Christia Fotini. 2012. Alliance Formation in Civil Wars . Cambridge: Cambridge University Press. Google Scholar CrossRef Search ADS   Clarke Colin P., Paul Christopher. 2014. “From Stalemate to Settlement: Lessons for Afghanistan from Historical Insurgencies that have been Resolved through Negotiations.” Rand National Defense Research Institute, accessed July 20, 2017. http://www.rand.org/content/dam/rand/pubs/research_reports/RR400/RR469/RAND_RR469.pdf. Collier Paul, Hoeffler Anke. 2004. “Greed and Grievance in Civil War.” Oxford Economic Papers  56 ( 4): 563– 95. Google Scholar CrossRef Search ADS   Courson Elias. 2011. “MEND: Political Marginalization, Repression, and Petro-Insurgency in the Niger Delta.” African Security  4 ( 1): 20– 43. Google Scholar CrossRef Search ADS   “Cracks in militants’ camp widens.” 2009. Vanguard, August 10, 2009. https://www.vanguardngr.com/2009/08/crack-in-militants-camp-widens/. Cunningham David. 2006. “Veto Players and Civil War Duration.” American Journal of Political Science  50 ( 4): 875– 92. Google Scholar CrossRef Search ADS   Cunningham Kathleen Gallagher. 2011. “Divide and Conquer or Divide and Concede: How States Respond to Internally Divided Separatists.” American Political Science Review  105 ( 2): 275– 97. Google Scholar CrossRef Search ADS   Cunningham Kathleen Gallagher. 2014. Inside the Politics of Self-determination . New York: Oxford University Press. Google Scholar CrossRef Search ADS   Cunningham Kathleen Gallagher, Bakke Kristin M., Seymour Lee J. M.. 2012. “Shirts Today, Skins Tomorrow: Dual Contests and the Effects of Fragmentation in Self-Determination Disputes.” Journal of Conflict Resolution  56 ( 1): 67– 93. Google Scholar CrossRef Search ADS   de Figueiredo Rui J. P. Jr., Weingast Barry R.. “Vicious Cycles: Endogenous Political Extremism and Political Violence.” Institute of Governmental Studies Working Paper no. 2001–9, 2001. Department of the Army. 2014. Insurgencies and Countering Insurgencies  (FM 3–24/MCWP3-33.5). Washington, DC: Department of the Army. “Explosions hit Nigeria oil amnesty talks.” 2010. BBC News, March 15, 2010. http://news.bbc.co.uk/2/hi/africa/8567876.stm. Elbagir Nima, Paul Cruickshank, Mohammed Tawfeeq 2015. “Boko Haram purportedly pledges allegiance to ISIS.” CNN, March 9, 2015. http://www.cnn.com/2015/03/07/africa/nigeria-boko-haram-isis/index.html. Fearon James. 1998. “Bargaining, Enforcement, and International Cooperation.” International Organization  52 ( 2): 269– 305. Google Scholar CrossRef Search ADS   Findley Michael G. 2008. “Agents and Conflict: Adaptation and the Dynamics of War.” Complexity  14 ( 1): 22– 35. Google Scholar CrossRef Search ADS   Fjelde Hanne, Nilsson Desiree. 2012. “Rebels Against Rebels: Explaining Violence between Rebel Groups.” Journal of Conflict Resolution  56 ( 4): 604– 28. Google Scholar CrossRef Search ADS   Gates Scott. 2002. “Recruitment and Allegiance: The Microfoundations of Rebellion.” Journal of Conflict Resolution  46 ( 1): 111– 30. Google Scholar CrossRef Search ADS   Greenhill Kelly, Major Solomon. 2006/2007. “The Perils of Profiling: Civil War Spoilers and the Collapse of Intrastate Peace Accords.” International Security  31 ( 3): 7– 40. Google Scholar CrossRef Search ADS   Humphreys Macartan, Weinstein Jeremy M.. 2008. “Who Fights? The Determinants of Participation in Civil War.” American Journal of Political Science  52 ( 2): 436– 55. Google Scholar CrossRef Search ADS   International Crisis Group (ICG). 2014. Curbing Violence in Nigeria (II): The Boko Haram Insurgency , accessed July 20, 2017. https://www.crisisgroup.org/africa/west-africa/nigeria/curbing-violence-nigeria-ii-boko-haram-insurgency. Kalyvas Stathis N. 2006. The Logic of Violence in Civil War . New York: Cambridge University Press. Google Scholar CrossRef Search ADS   Kalyvas Stathis N.. 2000. “Commitment Problems in Emerging Democracies: The Case of Religious Parties.” Comparative Politics  32 ( 4): 379– 99. Google Scholar CrossRef Search ADS   Kocher Matthew A., Pepinsky Thomas B., Kalyvas Stathis N.. 2011. “Aerial Bombing and Counterinsurgency in the Vietnam War.” American Journal of Political Science  55 ( 2): 201– 18. Google Scholar CrossRef Search ADS   Krause Peter. 2013. “The Structure of Success: How the Internal Distribution of Power Drives Armed Group Behavior and National Movement Effectiveness.” International Security  38 ( 3): 72– 117. Google Scholar CrossRef Search ADS   Kydd Andrew, Walter Barbara F.. 2006. “The Strategies of Terrorism.” International Security  31 ( 1): 49– 80. Google Scholar CrossRef Search ADS   Lake David. 2002. “Rational Extremism: Understanding Terrorism in the Twenty-First Century.” Dialogue IO  1 ( 1): 15– 29. Google Scholar CrossRef Search ADS   Lyall Jason. 2009. “Does Indiscriminate Violence Incite Insurgent Attacks.” Journal of Conflict Resolution  53 ( 3): 331– 62. Google Scholar CrossRef Search ADS   Moghadam Assaf, Berger Ronit, Beliakova Polina. 2014. “Say Terrorist: Think Insurgent: Labeling and Analyzing Contemporary Terrorist Actors.” Perspectives on Terrorism  8 ( 5): 2– 17. Nygård Håvard, Weintraub Michael. 2015. “Bargaining Between Rebel Groups and the Outside Option of Violence.” Terrorism and Political Violence  27 ( 3): 557– 80. Google Scholar CrossRef Search ADS   Pearlman Wendy. 2008/2009. “Spoiling Inside and Out: Internal Political Contestation and the Middle East Peace Process.” International Security  33 ( 3): 79– 109. Google Scholar CrossRef Search ADS   Peterson Roger. 2001. Resistance and Rebellion: Lessons from Eastern Europe . New York: Cambridge University Press. Google Scholar CrossRef Search ADS   Powell Robert. 2006. “War as a Commitment Problem.” International Organization  60 ( 1): 169– 203. Google Scholar CrossRef Search ADS   Putnam Robert. 1988. “Diplomacy and Domestic Politics: The Logic of Two-Level Games.” International Organization  42 ( 3): 427– 60. Google Scholar CrossRef Search ADS   Ross Michael. 2006. “A Closer Look at Oil, Diamonds, and Civil War.” Annual Review of Political Science  9 ( 1): 265– 300. Google Scholar CrossRef Search ADS   Rudloff Peter J., Findley Michael G.. 2016. “The Downstream Effects of Combatant Fragmentation on Civil War Resolution.” Journal of Peace Research  53 ( 1): 19– 32. Google Scholar CrossRef Search ADS   Sergie Mohammed Aly, Johnson Toni. “Backgrounder: Boko Haram.” Council on Foreign Relations, last modified 2014, accessed July 20, 2017. http://www.cfr.org/nigeria/boko-haram/p25739. Seymour Lee, Bakke Kristin, Cunningham Kathleen. 2016. “E Pluribus Unum, ex Uno Plures: Competition, Violence, and Fragmentation in Ethnopolitical Movements.” Journal of Peace Research  53 ( 1): 3– 18. Google Scholar CrossRef Search ADS   Staniland Paul. 2012. “Between a Rock and a Hard Place: Insurgent Fratricide, Ethnic Defection, and the Rise of Pro-State Paramilitaries.” Journal of Conflict Resolution  56 ( 1): 16– 40. Google Scholar CrossRef Search ADS   Stedman Stephen John. 1997. “Spoiler Problems in Peace Processes.” International Security  22 ( 2): 5– 53. Google Scholar CrossRef Search ADS   Tattersall Nick 2010. “Q+A-The implications of bomb attacks in Nigeria’s oil delta.” Reuters, March 15, 2010. https://www.reuters.com/article/nigeria-delta-idAFLDE62E1A520100315. Waldman M. “Special Report: Dangerous Liaisons with the Afghan Taliban: The Feasibility and Risks of Negotiations.” United States Institute for Peace , last modified 2010, accessed July 20, 2017. http://www.usip.org/sites/default/files/SR%20256%20-%20Dangerous%20Liaisons%20with%20the%20Afghan%20Taliban.pdf. Walker Andrew. 2012. “Special Report: What is Boko Haram?” United States Institute for Peace . Accessed July 20, 2017. http://www.usip.org/publications/what-boko-haram. Weinstein Jeremy M. 2007. Inside Rebellion: The Politics of Insurgent Violence . Cambridge: Cambridge University Press. Wood Elisabeth. 2003. Insurgent Collective Action and Civil War in El Salvador . Cambridge: Cambridge University Press. Google Scholar CrossRef Search ADS   Wood Graeme. 2015. “What ISIS Really Wants.” Atlantic , March 2015. http://www.theatlantic.com/magazine/archive/2015/03/what-isis-really-wants/384980/. Zenn Jacob. 2013. “Cooperation or Competition: Boko Haram and Ansaru After the Mali Intervention.” CTC Sentinel  6 ( 3): 1– 4. Published by Oxford University Press on behalf of The International Studies Association 2018. This work is written by (a) US Government employee(s) and is in the public domain in the US. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Global Security Studies Oxford University Press

Bargaining with Insurgencies in the Shadow of Infighting

Loading next page...
 
/lp/ou_press/bargaining-with-insurgencies-in-the-shadow-of-infighting-oVgbApm7Ec
Publisher
Oxford University Press
Copyright
Published by Oxford University Press on behalf of The International Studies Association 2018.
ISSN
2057-3170
eISSN
2057-3189
D.O.I.
10.1093/jogss/ogx025
Publisher site
See Article on Publisher Site

Abstract

Abstract Despite the long standing “no concessions” argument, scientific studies now suggest that governments can benefit from negotiating with militant insurgencies. However, despite government efforts, the leaders of insurgent movements often appear fanatical and unwilling to negotiate. This behavior presents a puzzle: If the leaders of insurgencies mobilize to create political change, and a government offers concessions, why do insurgent leaders refuse to negotiate? Using a game-theoretic model, we argue that insurgent leaders may rationally reject negotiation due to an internal commitment problem. Specifically, when leaders cannot credibly share the benefits of peace with their rivals, insurgent leaders may reject offers over fear of an internal conflict, which could leave the entire group vulnerable to government exploitation. However, the model demonstrates that insurgent leaders should negotiate if power in the insurgency is shifting in favor of their rivals, as it could help them maintain control of the movement. We illustrate these hypotheses using evidence from the Nigerian state's conflict with the Movement for the Emancipation of the Niger Delta (MEND) organization and Boko Haram. infighting, negotiation, insurgency For many years, policymakers argued that states should never negotiate with insurgencies due to the inability of these movements to form credible commitments. Negotiation would allow insurgencies to gain from armed conflict, which in turn would encourage these movements to continue using violence to blackmail governments in the future. Recent studies, however, indicate that, while negotiation is risky, there are conditions in which governments can form credible agreements with insurgencies to facilitate conflict termination and transition rebel fighters into normal politics.1 Negotiating further allows governments to avoid the costly use of force, which appears often to be counterproductive when used indiscriminately (Kalyvas 2006; Kocher, Pepinsky, and Kalyvas 2011; Weinstein 2007).2 Yet, despite this growing consensus, several insurgent leaders appear apocalyptic and seem to view violence as an end into itself (e.g., Wood 2015). These leaders show little interest in government efforts to negotiate peacefully. This puzzling behavior raises the question: Why are some insurgent leaders unwilling to negotiate, even when governments are open to making concessions? Using a game-theoretic model, we argue that negotiation creates an internal commitment problem for insurgencies, in which leaders cannot credibly refrain from monopolizing the peace dividend and exploiting their rivals in the movement. This commitment problem raises the risk of infighting, leading strategic insurgent leaders to reject negotiation outright. Strategic intransigence is often interpreted as extremism. However, despite the risk of infighting following peace, we demonstrate that insurgent leaders may accept negotiation if they face growing challenges from their rivals for control over the movement. Empirically, the model predicts that leaders will reject negotiation when their power is at relative parity with their rivals, but the rivals’ power is not increasing. On the other hand, insurgent leaders will negotiate if power in the insurgency is shifting in favor of their rivals, thereby increasing the probability of infighting. Intra-Insurgency Politics and the Problem of Negotiation Initial studies focused on whether states should attempt to negotiate with violent nonstate actors, given that aggressive policies appeared to be counterproductive. Many of these studies concluded that the presence of extremist factions3 in nonstate groups posed a key barrier to negotiation (Kydd and Walter 2006; Pearlman 2008/2009; Stedman 1997).4 These “spoilers” typically responded to negotiation by resorting to terrorism, or attacks against noncombatants, to prevent both the state and the moderates from fulfilling their obligations. This intransigence makes reaching credible deals with multifaction insurgencies very difficult. However, a key empirical problem is how to identify extremists prior to observing the outcome of negotiations. Formal studies suggest there are strategic incentives for all militant fighters to signal extremism to gain bargaining power (Lake 2002). These incentives may hinder efforts to separate moderates from extremists ex ante. For example, consider the Quetta Shura in Afghanistan. The foremost Taliban group embraces violence, signals unwillingness to compromise, and appears content to continue fighting the United States. Yet, privately, some Quetta Shura fighters, including some believed to be in more senior leadership positions, have signaled some willingness to negotiate. It is therefore difficult to definitively classify these factions as moderates versus extremists prior to observing their response to negotiation. If we can only identify factions as extremist following negotiation, it becomes tautological to use the presence of extremists to predict the failure of negotiation. Defining extremists as those who gain rents from conflict or have strong tastes for violence constitutes a post hoc explanation since we cannot observe whether extremists exist ex ante. Descriptively, violence often increases following negotiation, and those acting as spoilers appear vehemently opposed to peace. What causes some to engage in extremist opposition to negotiation? Let us first consider the makeup of insurgencies. These movements may be comprised of a diverse set of individuals with a variety of motivations for engaging in violence (Pearlman, 2008/2009; Cunningham, Bakke, and Seymour 2012; Humphreys and Weinstein 2008; Peterson 2001; Wood 2003). Some may be true believers in a cause and may share the motivations of the leadership; others may be motivated to join by economic incentives, thrill-seeking, boredom, revenge, or issues tangential to the group's stated goals (Chai 1993). For example, Waldman (2010) demonstrates that the fighters and commanders within the Quetta Shura express widely different motivations for engaging in violence. The fighters disagree on how to reform the Afghan state, lack a common definition of Shari'a law, and express inconsistent conceptualizations of victory. While some fighters favor a fundamentalist state, others appear to prefer a limited central state with greater local autonomy. These numerous divergences indicate that victory as perceived by some may not be victory for others, and may even be a loss. Those seeking a strict interpretation of Islam enforced by violence are likely to challenge individuals seeking to limit the ability of the state to intervene in local issues. Similarly, success for Afghan nationalist fighters may conflict with success as defined by Pakistani-linked elements of the group, who may prefer to keep the government weak. Despite competing preferences, the fighters are united in their fight against the United States and the Afghan government. Other insurgencies similarly consist of multiple factions with different preferences and agendas.5 This diversity creates a substantial challenge for the leadership, which must unify the factions to maximize the insurgency's collective power.6 The variety of motivations also creates the potential for competition and rivalries, which may create sizeable challenges for negotiation (Bapat and Bond 2012; Christia 2012; Cunningham 2006; Fjelde and Nilsson 2012; Nygård and Weintraub 2015). If the insurgency is fractionalized, negotiations must appease both the leadership and the multitudes of fighters in order to achieve full disarmament. Ideally, the leadership would reach some agreement regarding distribution of the benefits to their followers after a settlement is reached. Once a leader receives concessions from the government, they must convince their rivals that they will share the benefits of the peace equitably. Unfortunately, given that the leader must first appease the elites that best help maintain their power, such a promise may not be credible (Acemoglu and Robinson 2009; Fearon 1998; Kalyvas 2000; Powell 2006). Instead, the leader may distribute a disproportionate share of the benefits to their own followers. Moreover, the leadership may engage in actions that will weaken its rivals’ political influence in the future. The leader might fund multiple infrastructure projects for their followers while providing minimal investment in their rivals’ zones of control. Over time, the wealth accumulated by the leader relative to their rivals may translate into greater political power, thereby leaving rival leaders vulnerable to further marginalization. This demonstrates the internal commitment problem produced by state negotiation: leaders cannot credibly refrain from marginalizing rivals following a settlement. This suggests that the leader's rivals should oppose settlements and strive to prevent implementation. We might therefore expect rivals to escalate incendiary terrorist attacks against the government's population or punish the leadership and its followers.7 Strategically, the purpose of these attacks may be to weaken the leadership and create public outrage, thereby making it politically impossible for governments to honor commitments under the negotiated settlement. This behavior fits the common description of extremist violence following peace negotiations. The internal commitment problem offers a rationalist explanation for extremism, in which violence is motivated by rivals’ fear of marginalization at the hands of the leader. However, a key problem with infighting is that it weakens the movement collectively, thereby leaving the entire insurgency vulnerable to “divide and conquer” tactics (Cunningham 2011, 276). Infighting destroys collective resources and turns the factions’ focus from fighting the government to fighting each other. Further, infighting leaves the leader's faction vulnerable to exploitation. If the government strikes a deal that reflects the movement's power at time t, but the movement weakens due to infighting at time t + 1, the government may use the infighting opportunistically to conquer the entire movement. Consider the Sudanese government's (GoS) negotiations with the Darfuri insurgency in 2006. The Sudan Liberation Army (SLA) consisted of two groups, one from the Fur group led by Abdel Wahid, and the other from the Zaghawa group, led by Minni Minawi. In early 2006, the two wings of the SLA and the Justice and Equality Movement (JEM) met in Abuja to discuss a possible peace agreement. Minawi agreed to disarm his group and was appointed to the fourth highest position in the government in return. The excluded groups branded Minawi and embarked on a campaign of violence to prevent implementation of the deal. Responding to the violence, Sudanese President Bashir announced in September that the GoS would no longer comply with United Nations Security Council (UNSC) Resolution 1706, authorizing United Nations (UN) peacekeeping in Darfur. Without consulting Minawi, Bashir announced that the GoS would deploy to Darfur to suppress the outlaws. In this case, negotiations led to infighting, which enabled the GoS to divide and conquer the insurgency. Given that infighting leaves the insurgency vulnerable to exploitation, leaders may fear and reject negotiations, giving the appearance of “extremism” and “irrationality.” However, we do empirically observe that several insurgent leaders do agree to negotiate, despite the risk of infighting (Clarke and Paul 2014). The next step is to identify the specific conditions under which insurgent leaders refuse or are deterred from making peace and those under which leaders are willing to risk negotiation. To do so, let us turn to the formal model. Model Figure 1 presents a stylized model of a conflict between a government G and an insurgent movement M, consisting of a leadership A and their rivals B, who oppose G but maintain a different agenda than A. Assume both A and B initially agree to maintain cooperation, directing their violence at G. The payoffs to G, A, and B, respectively, for the status quo in which conflict occurs and there is no negotiation are equal to –w; pφw; (1 – pφ)w. Table 1 provides a summary of the model’s parameters. We represent gains made by the movement as w ∈ (0, 1). Since these gains come at G's expense, we represent its loss from conflict as –w. Since w represents joint gains by A and B, we assume the gains are divided based on the balance of power within the insurgency. The model represents the power of leader A relative to B as pφ ∈ (0, 1) and B's power relative to A as 1 – pφ. Although we assume A is more powerful than B at the start of the game (P > 1 – p),8 we assume two types of insurgencies: a static insurgency in which A is not facing a growing challenge from their rival B, and a competitive movement in which B is gaining strength and influence. To represent this distinction, we assume that φ = 1 in the former type, whereas φ = 2 in the latter. This indicates that A's power relative to B does not change in the former type since P1 = p, but A's power is declining in the latter type since P2 < p.9 Figure 1. View largeDownload slide Intrafactional conflict game with payoffs for G, A, and B, respectively. Figure 1. View largeDownload slide Intrafactional conflict game with payoffs for G, A, and B, respectively. Moves G begins the game by either continuing to fight or opening negotiations. If G negotiates, it offers a set of concessions w, which represents a level of concessions equal to what the insurgency would gain from continuing the conflict. Since A is in the leadership position, they decide whether to accept or reject G's offer. If leader A rejects the deal, the game ends in conflict, and A and B receive payoffs equal to (pφw, [1 – pφ]w), respectively.10 Notice that A's payoff is greater if the political situation in the movement is static (φ = 1) than if a leader is emerging in B to challenge their control (φ = 2). However, if leader A accepts G's offer, they take possession of the concessions G makes and are given the responsibility of distributing the concessions to the larger group. Let us assume that G receives a peace dividend if A agrees to negotiate, which is normalized to 1. If leader A accepts G's offer, the rival B next decides whether to accept the agreement or resort to terrorism to subvert it.11 If the rivals accept, we assume that they forfeit the ability to fight independently of leader A's faction.12 Leader A can then distribute the benefits of peace consistent with the balance of power in the insurgency or use their leadership position to seize control of the rivals’ gains.13 Despite any ex ante agreements, B will have no ability ex post to force leader A to share concessions; therefore, A will be unable to use promises of future concessions to sway B. Alternatively, the rivals may turn to terrorism in an effort to subvert the settlement before it consolidates. The attacks may be aimed selectively at civilians that support leader A or may be indiscriminate and target the government's civilians. Regardless, the purpose is to provoke the government into abrogating the agreement. Assume next that the direct costs of this terrorist violence are borne by both factions and that each one pays some cost c(A, B) ∈ [0, 1]. The provocative nature of the attacks also creates political pressure on G to retaliate and abandon the settlement. Assume that the political cost to G for exercising restraint and refusing to retaliate is represented by α ∈ [0, 1], which is an increasing function of the number of successful terrorist attacks B generates. Assume the number of successful terrorist attacks, k, is drawn from a Poisson distribution for which λ represents the mean number of successful attacks and is a function of the relative capabilities of the factions and the overall success of the movement.14 Formally, assume λB = 10(1 – pφ)w, which assumes that the maximum number of successful attacks is k = 10.15 This indicates that the expected number of successful terrorist attacks is a function of the rivals’ capability and the overall movement's success in fighting the government. The rivals produce more successful attacks if they are in a better military position vis-à-vis G (w→1) and if they are more powerful relative to A. Neither the government, leader A, nor rival B are certain how many successful attacks will materialize, but all are aware of the distribution from which the number is drawn.16 The use of N in Figure 1 denotes that this specific process—rather than a rival's strategic choice—determines the number of successful attacks. Figure 2 maps a hypothetical distribution of successful attacks initiated by B if G negotiates with A. When p = 0.4, A is relatively weaker than B. This means if G negotiates with leader A, it can expect that B can respond with a greater number of successful terrorist attacks. The mean number of successful attacks in this case is 4.2, indicating G should expect around four attacks. By contrast, if A is relatively stronger and p = 0.8, the mean number of attacks that G should expect in response to negotiations falls to 1.4. All else equal, excluding stronger factions from negotiation is likely to produce more terrorism, relative to excluding weaker factions. Figure 2. View largeDownload slide Distribution of number of successful terrorist attacks by B as a function of the balance of power in the movement (p). Note: Squares when p = 0.4, indicating A is relatively weaker than B; circles when p = 0.8, indicating A is relatively stronger. Figure 2. View largeDownload slide Distribution of number of successful terrorist attacks by B as a function of the balance of power in the movement (p). Note: Squares when p = 0.4, indicating A is relatively weaker than B; circles when p = 0.8, indicating A is relatively stronger. Assume G's payoff for refusing to retaliate and holding to a negotiated settlement is equal to 1 – w – kα. G receives the benefit of peace, minus the value of its concession, and minus the political cost –kα for failing to respond to the attacks, where k ≥ 0 represents the total number of successful attacks initiated by the rivals. After observing successful attacks, G may uphold the agreement, as the British upheld the Good Friday Agreement following the Omagh bombing in 1998. Doing so requires G to pay the political penalty α, which is increasing in the number of successful attacks. Alternatively, G can avoid paying the political price by abrogating its commitment to the collaborators. Abrogating the agreement at this stage is also advantageous due to the insurgency's infighting, which weakens its ability to fight and survive G's repression. We represent the power of the divided insurgency to resist G as w2. This indicates that G receives a payoff of –w2 if it resorts back to violence and a payoff of 1 – w – kα if it upholds the settlement. Solution Because the game is characterized by complete yet imperfect information, we solve it using the subgame perfect solution concept (SPE).17 This section characterizes the main insights behind the solution, leaving most of the formal details to the appendix. We begin at the game's final move (i.e., the government's decision to uphold or abrogate the settlement). Since negotiated settlements produce a peace dividend, the government strictly prefers to reach a settlement. However, if the rivals resort to terrorism, the government may be unable to fulfill its bargain for two reasons. First, the attacks may make it politically impossible for the government to refrain from violence (Bueno de Mesquita 2007). Second, infighting increases G's temptation to renege in favor of repressing the entire insurgency. Since G's payoff is decreasing in w, its payoff for conflict increases from –w against a unified movement to –w2 against a divided movement. Therefore, infighting may benefit the government, in that it fractures the movement and leaves it vulnerable to divide and conquer tactics.18 Formally, G abrogates peace in response to terrorist violence if –w2 > 1 – w – kα, which simplifies to   \begin{equation}{\rm{k}} > ( {1-{\rm{ w}} + {{\rm{w}}^2}} )/\alpha \end{equation} (1) We define k* = (1 – w + w2)/α. This indicates that the rival fails to stage enough successful attacks to sabotage the settlement if k ≤ k* and generates enough attacks to destroy the settlement if k > k*. If the government negotiates, the probability the peace deal holds is   \begin{equation}{\sigma _{\rm{B}}} = \Gamma ([ {{{\rm{k}}^*} + 1} ],{\lambda _{\rm{B}}})/[ {{{\rm{k}}^*}} ]!\end{equation} (2) The Rivals Rival B recognizes that accepting the deal will result in marginalization at the hands of A.19 Alternatively, resorting to terrorism has several potential drawbacks. First, terrorism is costly. Second, there is no guarantee B can stage enough successful attacks to scuttle the deal.20 Third, infighting weakens the entire insurgency, leading to fewer gains from conflict. B turns to terrorism and plays reject if   \begin{equation} {\sigma _{\rm{B}}} < ([1 - {{\rm{p}}^\varphi }]{{\rm{w}}^2} - {{\rm{c}}_{\rm{B}}})/([1 - {{\rm{p}}^\varphi }]{{\rm{w}}^2})\end{equation} (3) We define the critical probability that B fails to generate enough attacks to sabotage peace as σB* = ([1 – pφ]w2-c B)/([1 – pφ]w2). As σB decreases, B is increasingly likely to turn to terrorism. Comparative statics indicate that B is less likely to resort to terrorism if cB→1 and more likely to do so if cB→0. The cost of violence is often interpreted as a measure of resolve, or willingness to engage in conflict. In this case, cB may represent the animosity between the leader A and rival B. Empirically, if the leaders’ preferences are highly divergent, B is more likely to respond violently to a settlement between G and A. Condition 3 demonstrates that B responds with terrorism if the conflict is going relatively well for the movement (w→1). Infighting is therefore less likely when the larger insurgent movement is weak and insignificant. If these rivals were to succeed in sabotaging peace, which they are less likely to do given the lower value of w, infighting would effectively undermine the ability of these groups to combat the government. Moreover, governments are less likely to uphold peace deals with weaker movements, given that these deals are less costly to make. The balance of power between the leader and their rivals also influences B's decision to resort to violence.21 If leader A commands a significantly larger fighting force than B, and controls most of the insurgency's military power (p→1), B is likely to be deterred from terrorism. In such cases, attacks are costly, but unlikely to undermine the settlement. B is better off accepting the distasteful deal while sparing the cost of staging attacks. Formally, B is deterred and accepts G's deal with A if p ≥ 1 – (cB/w). However, if B maintains a larger share of military power, it is possible for B to resort to terrorism following negotiations. Leaders Leader A anticipates whether B will resort to terrorism. If A is preponderant in the movement (p→1), B will acquiesce, and A strictly prefers to accept G's offer to negotiate, since w > pφw. However, if A's forces are not powerful enough to deter B from rejecting the deal, making peace is risky. Alternatively, if A refuses to negotiate, they receive a payoff of pφw for continuing the conflict. A therefore accepts if σB(w) + (1 – σB)(pw2) – cA ≥ pφw. Simplifying,   \begin{equation} {\sigma _{\rm{B}}} \ge ({{\rm{c}}_{\rm{A}}} + {\rm{w}} [{{\rm{p}}^\varphi } - {\rm{pw}}])/( {{\rm{w}} - {\rm{p}}{{\rm{w}}^2}} )\end{equation} (5) We define A's critical probability that a settlement holds as σB** = (cA + w[pφ – pw])/(w – pw2). A accepts G's offer if σB ≥ σB** and rejects it otherwise. We use comparative statics to examine what key factors affect leader A's decision to accept. Here, the critical variable is pφ, or the power A's forces will control if A refuses to negotiate. If power within the movement is static (φ = 1) and A's position is stable, the right-hand side is larger, and condition 5 is more difficult to fulfill. If A's power is declining (φ = 2), condition 5 is easier to fulfill. Figure 3 plots the critical probability σB** as a function of pφ. The solid line represents the case in which A's power is static, while the dashed line represents the case in which A's political influence is diminishing relative to B's. A is less likely to accept G's offer if their leadership will not be challenged, but more likely to do so if they face a political threat from a competitor in B. However, since A is presently more powerful within the movement than B,22 A has little reason to risk an internal conflict if B does not appear politically threatening. Alternatively, if A's power is diminishing relative to B's (φ = 2) and they reject negotiations, their gains from conflict are diminished since B is surpassing them in power. However, if A accepts, they may seize control of all gains from conflict. If they face a viable competitor and their power is in decline, leader A has the greatest incentive to strike a deal immediately. Figure 3. View largeDownload slide Threshold for insurgent leader A to accept as a function of its power relative to B (pφ). Figure 3. View largeDownload slide Threshold for insurgent leader A to accept as a function of its power relative to B (pφ). Equilibrium23 Figure 4 maps the equilibria as a function of the balance of power within the movement (p). The game has a unique SPE that varies according to different configurations of parameter values. Let us first consider the behavior of the players when A is preponderant. If leader A is dominant within the movement, they reject government offers so long as the movement is static, but accepts if power is fluid. On the other hand, leader A will accept G's offer if B is deterred from engaging in terrorism, regardless of whether power within the movement is static or fluid. Equilibrium 1: Governments and insurgent leaders can credibly negotiate without facing attacks from rival insurgents if the leader is sufficiently dominant over all other rivals within the insurgent movement.24 Figure 4. View largeDownload slide Equilibrium behavior as a function of p in static and dynamic cases. Figure 4. View largeDownload slide Equilibrium behavior as a function of p in static and dynamic cases. In the case where leader A is relatively weaker, B will resist any deal with violence. Therefore, if the movement is static and A's future gains will resemble current gains, A is deterred by B's threat and rejects if A is relatively weak. A is unwilling to risk conflict with B given that their payoff for continuing to fight together remains relatively higher. Interestingly, this suggests that if A does not have secure control over the movement, but does not believe that they will be challenged politically in the absence of peace efforts, they are likely to resist any efforts at peace. It follows then that insurgent leaders who do not face growing political competition and have moderate control over their movement should be the most intransigent in negotiation and may appear extremist or irrational by rejecting any efforts at peace. The model suggests this refusal to negotiate is not motivated by irrationality, but rather by a fear of infighting that will result from negotiation. This may explain the behavior of various terrorist leaders, such as the late Mullah Omar and Abubakr Shekau, who face little political competition but wield only moderate control over their movements. These leaders may fear that negotiation will cause an internal conflict that will leave their entire movement vulnerable, a perception that deters them from negotiation. However, if these leaders do face political competition, they may risk infighting in order to hold onto their power. Facing internal challenges, these leaders choose to bandwagon against their competitors by striking deals with the government. Equilibrium 2: Governments and insurgent leaders cannot reach negotiated settlements if no shifts in power are occurring and the leader is near parity with their rivals in the movement.25 Equilibrium 3: Governments and insurgent leaders reach negotiated settlements if power is shifting away from the leader and the leader is near parity with their rivals in the movement.26 Implications The model produces three hypotheses. First, leaders who maintain a preponderance of power within their movements are free to negotiate without fear of infighting. These leaders can reach Pareto efficient deals with governments and should seek to do so as peace is preferable to continued conflict. H1:Politically powerful leaders are more likely to negotiate. Alternatively, if leader A is only moderately powerful relative to their potential competitors, their rival B escalates violence in response to any attempt by leader A to negotiate. Since the rivals are not weak, their campaign is more likely to provoke G into abrogating the agreement, thereby allowing G to fight a weakened, fractured movement. Given this possibility, A may be deterred from striking an efficient deal with G. If A is only moderately powerful, but the movement is static and lacks an alternative to the leader, A will refuse to negotiate with G. A has no incentive to risk infighting when their political control is uncontested, yet vulnerable to prospective challengers. However, if leader A faces competition from B, inaction risks a future loss of control over the movement. The model therefore predicts that, while moderately powerful but politically unchallenged leaders will adopt hardline stances, those facing competitors may strike deals to lock in control over their movements. H2:Moderately powerful leaders are less likely to negotiate if their rivals are not increasing in power. H3:Moderately powerful leaders are more likely to negotiate if their rivals are increasing in power. The Case of Nigeria To illustrate, we examine Nigeria's conflicts with two prominent insurgencies, Boko Haram and the Movement for the Emancipation of the Niger Delta (MEND).27 These cases allow us to hold constant both the target government's capacity and time period, as both groups began in the early to mid-2000s. Also, foreign state influence is not a factor in either conflict. In both movements, factions have used violence in similar ways. We therefore contend there are no moderate factions in either movement.28 Government attempts at negotiation, however, have met different levels of success. Both Boko Haram and MEND have geographic influence. Boko Haram's presence extends to many northeastern states in Nigeria, including Borno, Yobe, Katsina, and Bauchi. MEND's core influence is in the South, largely throughout the Niger Delta. The two groups maintain a fairly diffuse power structure, and power is not equally distributed among their factions. Within Boko Haram, Abubakar Shekau leads a thirty-member Shura council that directs multiple scattered cells, many of which never interact (Walker 2012, 8). The International Crisis Group (ICG) reports that, of Boko Haram's six factions, Shekau's is the largest and, at the time of the report, had been responsible for the majority of a recent spate of attacks (ICG 2014). MEND similarly consists of multiple factions with no real hierarchical structure, though some factions appear more powerful. In contrast to Boko Haram, the organization does not have a single readily identifiable leader (Courson 2011, 31). During the heavy fighting that occurred between 2006 and 2009, major leaders in MEND included Henry Okah, Tom Polo in the Delta State, and General Boyloaf in the Bayelsa state (Amadi, Imoh-Itah, and Obomanu 2016, 176). These three leaders appeared to be the most significant in the organization. However, MEND also consisted of many smaller factions, such as those led by Ateke Tom, Egberi Papa, and Asari Dokubo (Amadi, Imoh-Itah, and Obomanu 2016, 177). Power in MEND was neither equitably distributed nor concentrated in any one leader. Instead, power fluctuated between factions as fighters’ loyalties shifted. Both Boko Haram and MEND fall into the “moderately powerful” category of armed organizations, where some factions are somewhat more powerful than others, but none is preponderant. The two groups adhere to very different ideologies. While Boko Haram is Salafi Jihadi, the bulk of MEND's activity revolves around resisting the operations of Western oil companies (Courson 2011, 33). The group's individual factions have a range of diverse objectives and ample access to economic resources through oil bunkering, ransom, protection money, and gun running (Akinola 2011, 76). According to Weinstein's (2007) logic, this suggests that, relative to Boko Haram, MEND is more likely to recruit and attract opportunistic insurgents who, having been initially attracted by the promise of personal enrichment, are more inclined to shift their loyalties to a rival faction or otherwise challenge leadership when doing so promises material gain. Not only does Boko Haram's relative lack of resources limit opportunism, but Berman (2009) argues that radical religious organizations are better equipped to screen out opportunists and build loyalty to the group and its leadership among their cadre or recruits. According to Berman, groups like Boko Haram should see greater cohesion and membership stability because they require members to make very costly initial sacrifices that screen out opportunists likely to defect to other factions, while including those who are loyal to the group and its leadership, and because members become more dependent on the group as they are separated from the outside world both physically and through adopting the extreme lifestyle dictated by the group's religious prohibitions (2009, 81). Thus, if we assume that these recruitment mechanisms remain fairly constant across factions, we should expect fewer rapid power shifts within Boko Haram, since its members are less likely to challenge leadership, defect to opposing factions, create their own opposition forces, or seek material rewards. We should expect therefore to see more stable factions and few leadership challenges within Boko Haram and, conversely, more shifts among the MEND factions where material incentives to switch exist, loyalty to faction leaders is expected to be lower and opportunism is expected to be higher. Indeed, leadership stability should be much higher in Boko Haram, compared to MEND. Boko Haram: No Viable Political Competition By most accounts, Boko Haram is considered a cohesive organization that operates through a cellular structure, wherein Shekau passes instructions via trusted leaders (BBC 2014). The organization's roots are in the Yusufiyya, a movement founded in 2002 by Mohammed Yusuf that is often called the Nigerian Taliban for its aim to purify Islam in the country (ICG 2014, 9). The Yusufiyya took on a state-like role, engaging in the care of refugees and religious policing. Around 2007, it began to conduct bombings and targeted assassinations, and Yusuf was eventually captured and killed in 2009. At that time, Shekau took control of the movement, and it became known as Boko Haram (ICG 2014).29 As members who were exiled following Yusuf's death returned to Nigeria in 2010, they raided the homes of those who had cooperated with the police, and killed families who refused to leave the confiscated dwellings of Boko Haram agents. Boko Haram also began to supplement its cash flow through robbery. However, rather than shifting its focus toward personal enrichment of leaders or members, Boko Haram goes to great lengths to free captured operatives, provides for widows, and pays and supplies active fighters more consistently than the Nigerian armed forces (Campbell 2014). Although Boko Haram supplements its income through crime, the group seems to have no significant external funding sources (Campbell 2014).30 While Boko Haram eventually gained resources through theft, its limited initial resource access made it more dependent on member sacrifices. Evidence suggests that early members of the Yusufiyya and Boko Haram were anything but opportunistic and were instead highly loyal to their leaders. Walker writes that a senior member explained “more than anything else, it was what Yusuf revealed to him about the Quran that convinced him to throw his lot in with the group, give up his job, and bring his family to live in the mosque” (2012, 9). The combination of limited initial resources, radical religious motivations, heavy initial sacrifices, and isolation from outside influences is expected to limit the level of opportunism among group members and produce a more cohesive group with little shirking. Indeed, Boko Haram's ability to carry off sophisticated attacks, such as a 2010 jailbreak freeing more than seven hundred inmates, speaks to the commitment of its members (Sergie and Johnson 2014). Following Yusuf's death, internal divisions formed between nationalist elements under Shekau and factions more focused on transnational activities that also maintained ties to regional groups. A splinter group by the name of Ansaru emerged following a Boko Haram attack that killed nearly two hundred Nigerians in a Muslim-majority area in January 2012. Its formation was in part a reaction to Boko Haram's practice of killing other Muslims, particularly defectors, and also the result of the growing nationalist-transnational divide within the organization (Bey and Tack 2013; Zenn 2013). While Ansaru initially distanced itself from Shekau and his practice of killing other Muslims, a subsequent Ansaru leader, al-Barnawi, began to cooperate with Shekau (Zenn 2013; ICG 2014). Ansaru and Boko Haram also coordinated joint kidnapping and ransom efforts, with Shekau providing the manpower (Zenn 2013; ICG 2014, 22).31 A report from the ICG also goes on to say that, while Shekau “nominally controls fighters led by rival leader Mamman Nur ... most commanders and foot soldiers are more loyal to Nur ... but [Nur] chooses to maintain the status quo because of Shekau's ruthlessness” (ICG 2014, 22). These dynamics highlight Shekau's vulnerability to potential infighting. Furthermore, “both al-Barnawi and Nur lack sufficient grassroots networks in Borno, where Boko Haram carries out more than 80 percent of its attacks,” for them to be able to lead the group were Shekau to be killed or captured (Zenn 2013, 29). Although there are no direct challenges to Shekau's position of power, his loyalist faction does not maintain absolute control over the organization. Our model predicts that the loose structure of the organization, combined with the absence of challengers, should cause Shekau to adopt a hardline, intransigent view toward negotiation. This appears correct, as Shekau often seems to lack a clear set of demands and appears more interested in continuing violence, rather than negotiating. Citing an interview with an intelligence officer, the ICG reports that Shekau orders the killing of members who propose negotiations, thereby “silencing other prodialogue individuals” (ICG 2014, 21). Though Shekau makes every effort to appear powerful, his behavior suggests that he maintains less control over the organization than many analysts believe. This is evident through both the presence of multiple factions within Boko Haram and a 2012 video in which Shekau embraces al Qaeda. The latter in particular is indicative of a shift away from the Shekau faction's previously exclusive focus on domestic dynamics and toward one of a more regional nature embraced by rival factions (ICG 2014, 28). This shift may have been intended to prevent Ansaru, who aligned with al Qaeda in the Islamic Maghreb (AQIM), from gaining ground. These shifts in organizational focus indicate that if Shekau were to accept an offer of amnesty or settlement from the Nigerian state, Boko Haram may become vulnerable to infighting. Given these risks, Shekau is more likely to prefer avoiding a potentially divisive settlement. Absent outright challenges to his power, Shekau appears free to continue his terrorist campaign while avoiding risky settlements with the Nigerian state. The unwillingness of Boko Haram to come to the negotiating table, as well as the group's practice of cracking down on those who reach out to the state without the leadership's authorization to do so, is consistent with our model's characterization of a moderately powerful leadership. These circumstances suggest that, while various factions may not totally embrace a group leader's control, there are no viable challengers that threaten the overall internal balance of power.32 MEND: A Competitive Political Environment Unlike Boko Haram, which is characterized by the presence of weak factions, more powerful factions within MEND that fear their own future decline should be more likely to seek opportunities to negotiate with the state, at the expense of weaker factions. MEND coalesced in 2005 from an assortment of opposition groups in the Niger Delta with the stated goals of ensuring the return of a larger share of the proceeds from the region's oil wealth to its people and the withdrawal of government troops from the region (Courson 2011). MEND factions were united in their opposition to oil companies, but divided in their individual objectives. The lucrative nature of MEND's activities attracts both ideological and greed-motivated fighters (Akinola 2011, 76). The combination of low entry barriers, great resource wealth, and numerous autonomous factions results in dynamic power shifts among factions. MEND has identifiable faction leaders, but no central command structure. Formed from rival groups, MEND maintains many factions operating with a high degree of autonomy, which cooperate when it is mutually beneficial to do so (Courson 2011). In 2009, when faction leader Henry Okah faced capital charges, the Nigerian government found itself facing a large violent movement with no faction sufficiently powerful to deter others from cooperating with the government. Taking advantage of this opportune moment in the internal structure of the armed group, President Yar'Adua initiated negotiations with Bayelsa state faction leaders and announced an amnesty program. On August 7, Bayelsa's leader Boyloaf became one of the first to accept amnesty, citing the government's release of Okah and the offer of development programs as justification. Boyloaf's acceptance prompted other MEND leaders to announce his replacement. According to a Vanguard report from July 24, 2009, Boyloaf charged Ijaw youth leaders in the area with attempting to persuade terrorist leaders that the government offer was insincere, and the news outlet reported on August 10, 2009, that eleven MEND leaders in Delta State had disowned Boyloaf (Boyloaf accuses Ijaw 2009; Cracks in militants’ camp 2009). By year's end, many other major leaders had negotiated amnesty agreements. These agreements offered large payouts and protection contracts to critical leaders including Ateke Tom (leader of the Icelanders gang and the Niger Delta Vigilante Movement in Rivers State), Tompolo (leader of a Delta State faction), and Boyloaf. Each was paid to secure the oil lines they had previously attacked (Amadi, Imoh-Itah, and Obomanu 2016, 178). Job training and smaller payments, often to be distributed by commanders, were offered to lower-ranking fighters. However, these concessions did little to satisfy MEND's initial demands and were thus rejected by some factions. While a few top commanders like Tompolo and Boyloaf became increasingly wealthy and powerful, many in the region, including former terrorists, did not benefit. Why were the leaders willing to negotiate despite the risk of infighting? Theoretically, the model predicts that moderately powerful leaderships will negotiate if they face political challengers. Since the power of MEND commanders was solely based on their control of territory and wealth associated with the oil trade, the fluidity of power guaranteed that other commanders would crop up to challenge the old guard. Therefore, in the MEND case, the presence of oil inhibits the sort of leadership stability that we argue leads to intransigence in other cases.33 This is because oil creates material incentives for opportunistic individuals to join or form militant groups to which they do not have strong loyalties. Opportunistic rebels in turn are more likely than those that feel personal loyalty to the group and its leaders to defect from one group and join or form a rival if doing so promises material gains. Even as the amnesty concluded, the old guard faced challenges from individuals who claimed that the deal did nothing to stop oil pollution or address income inequality in the South. Several of these individuals claimed that they would go back to the creeks and resume the campaign of violence (Ajibola 2015). Had Boyloaf, Tompolo, and Tom continued with insurgency, there was no guarantee that they would not have been replaced by the myriad other “generals” within the MEND organization. Therefore, each of these leaders strategically decided to negotiate before losing all of their power to potential competitors. While the payments and job training programs calmed the violence temporarily and included an estimated twenty-six thousand former militants as of 2011 (Canada 2011), the amnesty did little to address underlying issues of the insurgency, and many former militants remained unsatisfied. Consistent with the model's predictions, the excluded militants who did not benefit from the program responded by escalating terrorist violence, breaking MEND's 2009 truce with attacks specifically targeting amnesty discussions in the Delta State. A March 15, 2010, BBC report quotes a Delta State spokesperson as responding with the following: “I think the intention is obvious, just to scuttle the talks and make it seem as if Warri in Delta State is not safe” (qtd. in Explosions hit Nigeria 2010). A Reuters report from the same day notes that, if MEND had lived up to its threats of “renewed attacks on oil installations in the coming days,” this could have derailed the amnesty and Nigeria's economy (Tattersall 2010). MEND's shifting and loosely aligned factions enabled the Nigerian government to cut deals with leaders of sufficient strength to weaken the overall movement. While no leader was powerful enough to deter others from accepting lucrative security contracts from the government, those who accepted the government's deals were insufficiently powerful to stop others from continuing the campaign and engaging in spoiler violence. Furthermore, the movement's multiple factions and fluid factional boundaries incentivized leaders to cut deals with the government to avoid losing out later as their relative power declined. Consistent with our expectations, we observe the government negotiating with more powerful factions within a dynamic insurgency and with factions that stand to lose some of their influence within the movement. In response, however, we also observe rivals engaging in terrorist attacks to scuttle the deal. Conclusion We began with the question: Why do insurgent leaders resist bargaining, even when governments appear inclined to offer concessions to resolve conflicts? We argue that this resistance to negotiation may be motivated by internal politics within the insurgency. Since individuals join these movements for a variety of reasons, the goals of the leadership may not fully represent the goals of the individual members within the organization. Therefore, if an insurgent leader were to reach an agreement with the government, the leader may be tempted to horde all of the concessions for his followers while excluding members from competing factions. We argue that this incentive on the part of the leader creates a commitment problem within the violent movement and may lead to infighting following peace. Given this anticipation and the expectation that infighting may tempt the government to pursue a divide-and-conquer strategy, insurgent leaders may resist all efforts at negotiation. Although such leaders may appear behaviorally “irrational” or “extreme,” we argue that they are strategically dodging internal conflicts that may occur following negotiations. However, we demonstrate that leaders may be willing to strike deals and risk infighting when it appears that they will face political competitors within their movement. In these cases, leaders that believe their political power is waning may seek to lock in their control over the movement by making peace, which in turn may trigger an internal power struggle between the leadership and its rivals. This research highlights the importance of examining the internal politics of violent opposition movements. Like states, it is conceivable that attempts to retain power motivate many leaders of violent groups, particularly in cases where their politics are competitive. Table 1. Parameters Parameter  Interpretation  w  Gains made by insurgent movement (A, B) in conflict  p  Insurgent leader's faction A share of power  1-p  Set of other faction's B share of power  cA  Cost to A for infighting  cB  Cost to B for infighting  ε  Political cost to G for A's rejection  α  Political cost to G for failing to respond to terrorist attacks  φ  Shift in political influence  k  Successful terrorist attacks  σB  Probability B fails to sabotage agreement if excluded  Parameter  Interpretation  w  Gains made by insurgent movement (A, B) in conflict  p  Insurgent leader's faction A share of power  1-p  Set of other faction's B share of power  cA  Cost to A for infighting  cB  Cost to B for infighting  ε  Political cost to G for A's rejection  α  Political cost to G for failing to respond to terrorist attacks  φ  Shift in political influence  k  Successful terrorist attacks  σB  Probability B fails to sabotage agreement if excluded  View Large Footnotes 1 Insurgency refers to “a struggle for control and influence, generally from a position of relative weakness, outside existing state institutions” (Department of the Army 2014, 1-1). Following Moghadam, Berger, and Beliakova (2014), we conceive of insurgencies as being of sufficient size and power as to be able to control territory. Following Bakke et al., we define an insurgent movement as one that “appeals to a shared identity;” as Bakke et al. note, this does not presume that “movements necessarily possess ‘common purposes and solidarity’” (2012, 266–267). 2 In one study that appears to be the exception, Lyall (2009) finds that indiscriminate shelling from Russian forces assisted in reducing Chechen terrorist attacks. 3 We use faction and group to refer to distinct subunits of an insurgent movement that share some common interest as part of the movement and attempt to appeal to a common identity, but have different preferences regarding policy, leadership, or tactics. Cell refers to an operational unit within a group or faction (see, e.g., Bueno de Mesquita 2005; Department of the Army 2014). 4 Alternatively, Abrahms (2008, 101) argues that terrorists may resist negotiation not because of extremism, but because individuals in these groups are “social solidarity maximizers.” While this is not conceptually the same as extremism, we would behaviorally expect both types of individuals to resist any form of peace, even if it is Pareto-improving. Similarly, Ross (2006), Collier and Hoeffler (2004), and Weinstein (2007) argue that greed-driven rebels often fight for profit rather than political goals, giving these fighters little incentive to end conflicts. 5 For a discussion of fragmentation in rebel movements, see Bakke, Cunningham, and Seymour 2012; Christia 2012; Cunningham 2006; Cunningham 2011; Cunningham 2014; Gates 2002; Krause 2013; Rudloff and Findley 2016; Seymour, Bakke, and Cunningham 2016; Staniland 2012. 6 This negotiating challenge is apparent in many contexts. Putnam, for example, observed that rather than states being unitary actors as they are often modelled, they are instead comprised of actors with differing preferences, noting that “the unitary actor assumption is often radically misleading” (1988, 433). Indeed, this is also true with regard to insurgencies, as has been observed previously in the literature addressing cleavages within such movements (see in particular Journal of Conflict Resolution 56, no. 1, a special devoted to fragmentation of nonstate actors). 7 This strategy of using terrorism as provocation is discussed by several authors, including Bueno de Mesquita (2005); Bueno de Mesquita and Dickson (2007); de Figueiredo and Weingast (2001). 8 p > 0.5. 9 For simplicity, we analyze the two cases where φ = 1 and φ = 2. An alternative strategy is to allow φ to take on a continuous value. This strategy produces a cutoff value φ* that divides into cases where φ < φ* and cases where φ ≥ φ*. The behavior identified in the first case is the same as when φ = 1, and the behavior in the latter case is the same as when φ = 2. The results are robust to the alternative specification. 10 G receives a payoff of –w–ε if its attempt at negotiation fails. This indicates that, if G is aware that A will reject, G prefers to play ∼negotiate. 11 To maintain tractability, assume that, in cases where there are N factions, if one faction rejects the deal, the entire set in B rejects the deal as well. 12 We make this assumption both to ease analytic tractability and to focus on the cases of interest. The assumption stems from the disarmament and demobilization process, which the United Nations counsels should occur early in peace processes. Even where local autonomy is granted, the maintenance of order depends upon disarming potential armed opponents to the leadership. 13 This sequence is similar to Fearon (1998), except that the division of w is not endogenous. An earlier version of the model that allowed A to endogenously divide w with B yields similar results. Please see Part IV in the appendix for this discussion. 14 We assume that attempts are not individually costly and that there are no diminishing returns of additional attacks. Therefore, B attempts as many attacks as it can, but only some attacks are successful. 15 This number is arbitrary, but is reasonable given the current literature. Theoretically, the maximum number of attacks could be any number. 16 This feature allows the insurgent leader to face some uncertainty regarding how powerful their rivals are in terms of marshalling attacks to oppose negotiations. 17 See formal appendix for proofs. 18 For empirical discussions, see Findley (2008) and Cunningham (2011). 19 As noted earlier, this results from the assumption that rival B forfeits the ability to fight independently of A when they acquiesce to the deal A has cut with the government. As a rational B should never anticipate that a rational A would honor any commitment to share the concessions (assuming that A and B have different preferences on at least some issues including leadership of the movement, meaning that A's utility would be declining in the amount of resources shared with B), we exclude this node in our analysis of the game for greater tractability. The model's results are robust to the inclusion of this node. 20 As noted above, attacks may target the civilians or leader A's faction. We refer to these attacks as infighting within the movement because, regardless of the target of the attack, the purpose is to scuttle A's deal with the government. 21 Similarly, Greenhill and Major argue that the relative power balance between “competing factions on the ground and those implementing the peace” is the most critical determinant of whether spoiler violence will occur (2006/2007, 9). 22 p→1 – (cB/w) 23 See appendix for proofs. 24 Formally, if p ≥ 1 – (cB/w2), {G: Negotiate; A: Accept; B: Accept} is an equilibrium pathway. 25 If 1 – (cB /w2) > p > 1 – (cA/w)1/φ and φ = 1, A plays reject. 26 If 1 – (cA/w)1/φ > p and φ = 2, A plays accept if σB > σB** and reject otherwise. 27 While this does not substitute for a systematic empirical test, the cases illustrate the model's causal mechanisms. 28 While Ansaru is often considered a less extreme faction of Boko Haram, it has allied with al Qaida in the Islamic Maghreb and the Islamic State (ISIS). The faction is also known for kidnapping and killing Western civilians in Nigeria Chothia 2013; Elbagir et al. 2015; Ajibola 2015. 29 Walker (2012) notes that the name Boko Haram is commonly applied to the group as a pejorative. The group's name for itself is Jama'a Ahl as-Sunna Li-da'wa wa-al Jihad. 30 Though the ICG's (2014) report indicates that, in 2002, the Yusuf and his movement received the bulk of a $3 million budget provided by Bin Laden for Nigerian Islamist groups. 31 Zenn's account and the ICG differ on whether Barnawi was still alive in 2013, with Zenn writing as if he were. The ICG report cites an interview with a Boko Haram member in August 2013, claiming that al-Barnawi was killed in an August 2012 security forces raid of his hideout, which resulted from his cooperation with Shekau in attempting to take out a faction leader that both suspected of being a government mole. 32 In 2011, an individual claiming to represent the group told BBC that the group would disarm if certain conditions were met. A Boko Haram spokesman quickly denied the deal and claimed “internal divisions were eliminated” (quoted in Walker 2012, 11). Similarly, Yusuf's brother-in-law was assassinated after meeting with former President Olusegun Obasanjo in an apparent effort to discuss peace terms. In January 2012, a faction claiming to be more moderate declared on television that it was ready to negotiate; these individuals were quickly beheaded. In October 2014, the Nigerian government announced it had reached a ceasefire agreement with a previously unknown Boko Haram leader. There was no public confirmation through Boko Haram's usual channels. By the following day, suspected Boko Haram terrorists had carried out five attacks. 33 While, in this case, it is control of a natural resource that produces incentives for rivals to emerge, it is important to note that such natural resources are not necessary to produce such a scenario. For example, viable rivals might emerge as the result of state sponsorship of rebels or of questionable leadership decisions coupled with effective counterinsurgency by the state, as in the case of the Algerian Islamic Group (GIA), which lost control of its movement after using massive indiscriminate violence, which resulted in the defection of many of its followers to a splinter group, the Salafist Group for Call and Combat (GSPC). Supplementary Information Supplementary information is available at the Journal of Global Security Studies data archive. References Abrahms Max. 2008. “What Terrorists Really Want: Terrorist Motives and Counterinsurgency.” International Security  32 ( 4): 78– 105. Google Scholar CrossRef Search ADS   Acemoglu Daron, Robinson James A.. 2009. Economic Origins of Dictatorship and Democracy . Cambridge: Cambridge University Press. Ajibola Iyabobola O. 2015. “Nigeria’s Amnesty Program: The Role of Empowerment in Achieving Peace and Development in Post-Conflict Niger Delta.” SAGE Open July/September: 1-11. Akinola Adeoye O. 2011. “Niger Delta Crisis: The Nexus between Militants’ Insurgency and Security in West Africa.” African Security  4 ( 1): 65– 80. Google Scholar CrossRef Search ADS   Amadi Luke, Imoh-Itah Imoh, Obomanu Edmund F.. 2016. “Oil: Niger Delta and the Making of Militia Capitalists: A Theoretical Assessment.” International Journal of Social Science and Humanities Research  4 ( 1): 172– 84. Bakke Kristin M., Cunningham Kathleen Gallagher, Seymour Lee J. M.. 2012. “A Plague of Initials: Fragmentation, Cohesion, and Infighting in Civil Wars.” Perspectives on Politics  10 ( 2): 265– 84. Google Scholar CrossRef Search ADS   Bapat Navin A., Bond Kanisha. 2012. “Alliances between Militant Groups.” British Journal of Political Science  42 ( 4): 793– 824. Google Scholar CrossRef Search ADS   BBC. 2014. “Nigeria's Boko Haram Leader Abubakar Shekau in Profile,” May 9. http://www.bbc.com/news/world-africa-18020349. Berman Eli. 2009. Radical, Religious, and Violent: The New Economics of Terrorism . Cambridge: MIT Press. Bey Matthew, Tack Sim. “ The Rise of a New Nigerian Militant Group.” Stratfor Global Intelligence, last modified February 21, 2013, accessed on July 20, 2017. http://www.stratfor.com/weekly/rise-new-nigerian-militant-group#axzz37DJvltbY. Bueno de Mesquita Ethan. 2005. “Conciliation, Counterterrorism, and Patterns of Terrorist Violence.” International Organization  59 ( 1): 145– 76. “Boyloaf accuses Ijaw leaders of inciting militants not to accept amnesty.” 2009. Vanguard, July 24, 2009. https://www.vanguardngr.com/2009/07/boyloaf-accuses-ijaw-leaders-of-inciting-militants-not-to-accept-amnesty/. “Boyloaf accuses Ijaw leaders of inciting militants not to accept amnesty.”. 2007. “Politics and the Suboptimal Provision of Counterterror.” International Organization  61 ( 1): 9– 36. Bueno de Mesquita Ethan, Dickson Eric. 2007. “The Propaganda of the Deed: Terrorism, Counterterrorism, and Mobilization.” American Journal of Political Science  51 ( 2): 364– 81. Google Scholar CrossRef Search ADS   Campbell John. “Nigeria: Why are Boko Haram Fighters Successful?” Africa in Transition (blog).  Council on Foreign Relations, March 4, 2014. http://blogs.cfr.org/campbell/2014/03/04/nigeria-why-are-boko-haram-fighters-successful/ Canada: Immigration and Refugee Board of Canada. 2011. Nigeria: Government Amnesty Program for Niger Delta Militants, Particularly with Respect to the Movement for the Emancipation of the Niger Delta (MEND) (2009–August 2011) . NGA103789.FE. http://www.refworld.org/docid/50740a192.html.Chai. Chai Sun-Ki 1993. An Organizational Economics Theory of Antigovernment Violence. Comparative Politics  26 ( 1): 99– 110. Google Scholar CrossRef Search ADS   Chothia Farouk 2013. “Profile: Who are Nigeria’s Ansaru Islamists?” BBC News, March 11, 2013. http://www.bbc.com/news/world-africa-21510767. Christia Fotini. 2012. Alliance Formation in Civil Wars . Cambridge: Cambridge University Press. Google Scholar CrossRef Search ADS   Clarke Colin P., Paul Christopher. 2014. “From Stalemate to Settlement: Lessons for Afghanistan from Historical Insurgencies that have been Resolved through Negotiations.” Rand National Defense Research Institute, accessed July 20, 2017. http://www.rand.org/content/dam/rand/pubs/research_reports/RR400/RR469/RAND_RR469.pdf. Collier Paul, Hoeffler Anke. 2004. “Greed and Grievance in Civil War.” Oxford Economic Papers  56 ( 4): 563– 95. Google Scholar CrossRef Search ADS   Courson Elias. 2011. “MEND: Political Marginalization, Repression, and Petro-Insurgency in the Niger Delta.” African Security  4 ( 1): 20– 43. Google Scholar CrossRef Search ADS   “Cracks in militants’ camp widens.” 2009. Vanguard, August 10, 2009. https://www.vanguardngr.com/2009/08/crack-in-militants-camp-widens/. Cunningham David. 2006. “Veto Players and Civil War Duration.” American Journal of Political Science  50 ( 4): 875– 92. Google Scholar CrossRef Search ADS   Cunningham Kathleen Gallagher. 2011. “Divide and Conquer or Divide and Concede: How States Respond to Internally Divided Separatists.” American Political Science Review  105 ( 2): 275– 97. Google Scholar CrossRef Search ADS   Cunningham Kathleen Gallagher. 2014. Inside the Politics of Self-determination . New York: Oxford University Press. Google Scholar CrossRef Search ADS   Cunningham Kathleen Gallagher, Bakke Kristin M., Seymour Lee J. M.. 2012. “Shirts Today, Skins Tomorrow: Dual Contests and the Effects of Fragmentation in Self-Determination Disputes.” Journal of Conflict Resolution  56 ( 1): 67– 93. Google Scholar CrossRef Search ADS   de Figueiredo Rui J. P. Jr., Weingast Barry R.. “Vicious Cycles: Endogenous Political Extremism and Political Violence.” Institute of Governmental Studies Working Paper no. 2001–9, 2001. Department of the Army. 2014. Insurgencies and Countering Insurgencies  (FM 3–24/MCWP3-33.5). Washington, DC: Department of the Army. “Explosions hit Nigeria oil amnesty talks.” 2010. BBC News, March 15, 2010. http://news.bbc.co.uk/2/hi/africa/8567876.stm. Elbagir Nima, Paul Cruickshank, Mohammed Tawfeeq 2015. “Boko Haram purportedly pledges allegiance to ISIS.” CNN, March 9, 2015. http://www.cnn.com/2015/03/07/africa/nigeria-boko-haram-isis/index.html. Fearon James. 1998. “Bargaining, Enforcement, and International Cooperation.” International Organization  52 ( 2): 269– 305. Google Scholar CrossRef Search ADS   Findley Michael G. 2008. “Agents and Conflict: Adaptation and the Dynamics of War.” Complexity  14 ( 1): 22– 35. Google Scholar CrossRef Search ADS   Fjelde Hanne, Nilsson Desiree. 2012. “Rebels Against Rebels: Explaining Violence between Rebel Groups.” Journal of Conflict Resolution  56 ( 4): 604– 28. Google Scholar CrossRef Search ADS   Gates Scott. 2002. “Recruitment and Allegiance: The Microfoundations of Rebellion.” Journal of Conflict Resolution  46 ( 1): 111– 30. Google Scholar CrossRef Search ADS   Greenhill Kelly, Major Solomon. 2006/2007. “The Perils of Profiling: Civil War Spoilers and the Collapse of Intrastate Peace Accords.” International Security  31 ( 3): 7– 40. Google Scholar CrossRef Search ADS   Humphreys Macartan, Weinstein Jeremy M.. 2008. “Who Fights? The Determinants of Participation in Civil War.” American Journal of Political Science  52 ( 2): 436– 55. Google Scholar CrossRef Search ADS   International Crisis Group (ICG). 2014. Curbing Violence in Nigeria (II): The Boko Haram Insurgency , accessed July 20, 2017. https://www.crisisgroup.org/africa/west-africa/nigeria/curbing-violence-nigeria-ii-boko-haram-insurgency. Kalyvas Stathis N. 2006. The Logic of Violence in Civil War . New York: Cambridge University Press. Google Scholar CrossRef Search ADS   Kalyvas Stathis N.. 2000. “Commitment Problems in Emerging Democracies: The Case of Religious Parties.” Comparative Politics  32 ( 4): 379– 99. Google Scholar CrossRef Search ADS   Kocher Matthew A., Pepinsky Thomas B., Kalyvas Stathis N.. 2011. “Aerial Bombing and Counterinsurgency in the Vietnam War.” American Journal of Political Science  55 ( 2): 201– 18. Google Scholar CrossRef Search ADS   Krause Peter. 2013. “The Structure of Success: How the Internal Distribution of Power Drives Armed Group Behavior and National Movement Effectiveness.” International Security  38 ( 3): 72– 117. Google Scholar CrossRef Search ADS   Kydd Andrew, Walter Barbara F.. 2006. “The Strategies of Terrorism.” International Security  31 ( 1): 49– 80. Google Scholar CrossRef Search ADS   Lake David. 2002. “Rational Extremism: Understanding Terrorism in the Twenty-First Century.” Dialogue IO  1 ( 1): 15– 29. Google Scholar CrossRef Search ADS   Lyall Jason. 2009. “Does Indiscriminate Violence Incite Insurgent Attacks.” Journal of Conflict Resolution  53 ( 3): 331– 62. Google Scholar CrossRef Search ADS   Moghadam Assaf, Berger Ronit, Beliakova Polina. 2014. “Say Terrorist: Think Insurgent: Labeling and Analyzing Contemporary Terrorist Actors.” Perspectives on Terrorism  8 ( 5): 2– 17. Nygård Håvard, Weintraub Michael. 2015. “Bargaining Between Rebel Groups and the Outside Option of Violence.” Terrorism and Political Violence  27 ( 3): 557– 80. Google Scholar CrossRef Search ADS   Pearlman Wendy. 2008/2009. “Spoiling Inside and Out: Internal Political Contestation and the Middle East Peace Process.” International Security  33 ( 3): 79– 109. Google Scholar CrossRef Search ADS   Peterson Roger. 2001. Resistance and Rebellion: Lessons from Eastern Europe . New York: Cambridge University Press. Google Scholar CrossRef Search ADS   Powell Robert. 2006. “War as a Commitment Problem.” International Organization  60 ( 1): 169– 203. Google Scholar CrossRef Search ADS   Putnam Robert. 1988. “Diplomacy and Domestic Politics: The Logic of Two-Level Games.” International Organization  42 ( 3): 427– 60. Google Scholar CrossRef Search ADS   Ross Michael. 2006. “A Closer Look at Oil, Diamonds, and Civil War.” Annual Review of Political Science  9 ( 1): 265– 300. Google Scholar CrossRef Search ADS   Rudloff Peter J., Findley Michael G.. 2016. “The Downstream Effects of Combatant Fragmentation on Civil War Resolution.” Journal of Peace Research  53 ( 1): 19– 32. Google Scholar CrossRef Search ADS   Sergie Mohammed Aly, Johnson Toni. “Backgrounder: Boko Haram.” Council on Foreign Relations, last modified 2014, accessed July 20, 2017. http://www.cfr.org/nigeria/boko-haram/p25739. Seymour Lee, Bakke Kristin, Cunningham Kathleen. 2016. “E Pluribus Unum, ex Uno Plures: Competition, Violence, and Fragmentation in Ethnopolitical Movements.” Journal of Peace Research  53 ( 1): 3– 18. Google Scholar CrossRef Search ADS   Staniland Paul. 2012. “Between a Rock and a Hard Place: Insurgent Fratricide, Ethnic Defection, and the Rise of Pro-State Paramilitaries.” Journal of Conflict Resolution  56 ( 1): 16– 40. Google Scholar CrossRef Search ADS   Stedman Stephen John. 1997. “Spoiler Problems in Peace Processes.” International Security  22 ( 2): 5– 53. Google Scholar CrossRef Search ADS   Tattersall Nick 2010. “Q+A-The implications of bomb attacks in Nigeria’s oil delta.” Reuters, March 15, 2010. https://www.reuters.com/article/nigeria-delta-idAFLDE62E1A520100315. Waldman M. “Special Report: Dangerous Liaisons with the Afghan Taliban: The Feasibility and Risks of Negotiations.” United States Institute for Peace , last modified 2010, accessed July 20, 2017. http://www.usip.org/sites/default/files/SR%20256%20-%20Dangerous%20Liaisons%20with%20the%20Afghan%20Taliban.pdf. Walker Andrew. 2012. “Special Report: What is Boko Haram?” United States Institute for Peace . Accessed July 20, 2017. http://www.usip.org/publications/what-boko-haram. Weinstein Jeremy M. 2007. Inside Rebellion: The Politics of Insurgent Violence . Cambridge: Cambridge University Press. Wood Elisabeth. 2003. Insurgent Collective Action and Civil War in El Salvador . Cambridge: Cambridge University Press. Google Scholar CrossRef Search ADS   Wood Graeme. 2015. “What ISIS Really Wants.” Atlantic , March 2015. http://www.theatlantic.com/magazine/archive/2015/03/what-isis-really-wants/384980/. Zenn Jacob. 2013. “Cooperation or Competition: Boko Haram and Ansaru After the Mali Intervention.” CTC Sentinel  6 ( 3): 1– 4. Published by Oxford University Press on behalf of The International Studies Association 2018. This work is written by (a) US Government employee(s) and is in the public domain in the US.

Journal

Journal of Global Security StudiesOxford University Press

Published: Jan 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off