Sink or Sync: How decision-making structures affect civil war negotiations Brian R. Urlacher University of North Dakota July 2007

Abstract Civil wars are difficult to resolve through negotiations. The prospects for negotiations, concessions, and the emergence of trust can be strangled by extremist elements in both the government and opposition. This project looks at the role of institutional structure in determining the ability of governments and opposition groups to create stable constituencies for peace in civil wars. To test this argument a computer simulation of an iterated prisoner’s dilemma is used. This project examines the difference between majoritarian and consensus institutional configurations. Both decision rules can produce peaceful outcomes; however, when groups using majoritarian and consensus decision rule interact, the prospects for peace collapse. The different decision rules can make it difficult to synchronize cooperative efforts, which undermines the ability of groups to overcome conflictual patterns. To illustrate the applicability of the simulation’s logic to real world cases, a plausibility probe of the Kashmir insurgency is undertaken.

Peace building is hard, and in civil wars it is especially hard. Conflict participants often face a crippling security dilemma in that no group wants to disarm while the other side holds onto its weapons. The prisoner’s dilemma is the common metaphor for this type of security problem (Jervis 1978; 1976), yet the prisoner’s dilemma metaphor is often accompanied by the assumption that the various sides are unitary actors that act rationally. This assumption, while a useful analytic simplification, quickly breaks down upon closer investigation.1 When we look closely at the actors involved in forming and carrying out agreements in civil wars, it becomes obvious that the unitary actor assumption does not fit. Both the opposition and government are collectives of individuals operating within formal and informal institutional frameworks. Furthermore, both the government and opposition forces must work to manage internal struggles between factions, typically moderates and hardliners. The strategies that states and opposition forces select are the product of internal struggles between factions (Siqueira 2005). The internal political battles that occur on both sides of civil wars are shaped not only by preferences but also by institutions. How groups make decisions is at least as important in explaining a given outcome as the degree to which conflict participants desire peace. This premise parallels new institutionalist thinking, which argues that “human actions, social contexts, and institutions work upon each other in complicated ways” (March and Olsen 1984, 734). There is a great deal of variation in the policy structures of states and perhaps even greater variation in the institutional configuration of rebel movements. One obvious difference is that governments typically have formal institutional mechanisms by which disputes between political factions are resolved and translated into policy; opposition groups, however, may lack comparable mechanisms for resolving internal disputes or disputes between different rebel groups. Thus, even when a state’s leaders are deeply divided, the state itself may appear to speak with a single voice. By contrast, opposition forces may be highly factious, 1 The rational actor model in its various forms has been under attack for decades. Two of the most prominent assaults on the model were launched by Allison (1971) and by Janis (1972).

1

and even within a single rebel organization, individual commanders may operate with a degree of autonomy that undermines the possibility of a unified political front.2 It is only on the very surface that the civil war security dilemma resembles a prisoner’s dilemma. The policy choices made by the government and the opposition are the result of internal dynamics rather than rational calculations made by a two unitary actors. Robert Putnam (1988) has illustrated how the institutional design can affect the leverage that states have in international negotiations. Similarly, Young and Urlacher (2007) have demonstrated that the mechanisms groups use to make decisions can play an important role in determining the prospects for peace. The Young-Urlacher study explored the interactions of democratic (majoritarian) and authoritarian systems but did not address the fractured decision structures that are more commonly found among opposition forces in civil wars. This then prompts the question, how do the decision making structures of governments and opposition groups shape the prospects for negotiated solutions to civil wars?

1

Conceptual Framework

1.1

Getting to the Table

Civil wars can end through military victory or negotiated settlement. Although Licklider (1993) has demonstrated that during the Cold War civil wars were more likely to end with a military victory than with a negotiated settlement, this does not appear to be the case in the post-Cold War period. During the 1990s almost 3/4ths of civil wars ended through a negotiated settlement (Bohrer and Hartzell 2005). While getting to the negotiation table is often the only way out of intractable civil wars, it can be exceedingly difficult. There is usually a great deal of mistrust between the various parties, and distrust of the other can become rooted in identity of conflict participants (Stein 2001). 2

There are examples of highly disciplined and unified opposition groups, such as the Maoist rebels in Nepal. Thus, it can not be assumed that opposition forces will always be fractured and divided.

2

In addition to the psychological barriers to negotiation, there are structural barriers that may make negotiations difficult. First, a negotiated settlement in a civil war commonly involves some degree of disarmament on the part of opposition forces while the government maintains its military capacity. This leaves the side that disarms in a precarious position (Walter 2002; Fearon 1998). While there are ways to help alleviate the anxiety that one-sided disarmament produces, these possibilities are difficult to see while fighting continues. Second, conflict participants often benefit politically from continued conflict. The ruling party or political faction may depend on the support of hardliners to remain in power. Moving to negotiate could leave a government electorally vulnerable or increase the likelihood of a coup or an assassination. Yitzhak Rabin’s assassination in response to the Oslo peace accords illustrates how personally costly peace can be for political leaders. Third, conflict participants often benefit financially from continued conflict. For the opposition, conflict may be a lucrative business. To sustain a resistance movement, groups must find avenues to finance their activities. This financing may be supplied by a foreign entity, be it a state or a sympathetic diaspora. Groups may also rely upon the extraction of high value resources such as gemstones or the export of illicit drugs (Fearon 2004) or the export of commodities in territory under their control (Collier et al. 2004). In short, when there is money to be made, rebels may actually prefer continued conflict over a generous political settlement. Because of these numerous barriers, getting to the table often requires a radical transformation of a conflict that reorders the incentives in such a way that conflict participants become willing to move forward on conflict resolution. This alignment of incentives is often referred to as “ripeness” (Zartman 2000, 228; Haass 1990, 6).

Identifying the conditions

that produce ripeness is, thus, an important avenue of research in conflict studies.3 3

Stephan Steadman (1991, 240) cautions that if the concept of ripeness is to be potentially useful it requires greater precision; otherwise, the concepts risks being tautological. Similarly, Kleiboer (1994, 115) argues that periods of ripeness are only identifiable in hindsight and thus not prescriptively useful. This critique has been taken an additional step further by O’kane (2007)? who argues that ripeness is at its core a subjective concept.

3

Geortz and Deihl (1997) have argued that to overcome entrenched patterns of conflict, a “political shock” may be necessary. A political shock can foster ripeness by altering the conflict dynamic, creating incentives for conflict resolution. This may involve a change in relative power or it may involve a change in leadership.4 While a political shock is an expedient way to break the equilibrium of a conflict and prompt a change, conflict participants can learn their way out of conflicts as well. Touval and Zartman (2001) have argued that conflict participants can find their own way to the negotiation table. Ripeness can come about through two internal processes. First, conflict may push parties toward a dangerous point of no return. As conflict participants hurtle toward a disastrous precipice, they may determine that the danger of continued conflict is too high and they alter course, hopefully in time to avert disaster. Second, conflict participants may become trapped in cycles of violence. Both sides may be able to inflict pain on the other but are unable to achieve a military victory. This “hurting stalemate” may create sufficient dissatisfaction with continued conflict on both sides that peace becomes possible (Zartman and Aurik 1991; Zartman 1985). Both the precipice and hurting stalemate logics involve conflict participants changing course and opting for peace; however, the mechanisms that underlie the two ideas are quite different. The precipice argument hinges on the ability of conflict participants to peer deep into the future and see an impending danger. By contrast, the hurting stalemate argument hinges on the ability of conflict participants to feel pain and recognize that conflict is not making progress. It is important to stress that the hurting stalemate does not require agents to look into a hazy and uncertain future and make calculations about future probabilities and costs. Rather, in a hurting stalemate agents only need to recognize that past military offensives were unable to produce victory and that the situation is in the present unbearable. The hurting stalemate logic is essentially a backward looking logic. While these two logics 4

For a cataloguing of various types of political shocks, see Karen Rasler’s (2000) “Shocks, Expectancy Revision, and the De-Escalation of Protracted Conflicts: The Israeli-Palestinian Case.” Rasler also provides a much needed argument about what type of shocks are likely to alter conflict processes in a positive way.

4

are not mutually exclusive, it is worth drawing out this distinction because the backward looking hurting stalemate logic mirrors the logic used in the computer simulation developed for this project.5

1.2

Decision Making Structures in Civil War Negotiations

A potentially useful way to think about civil war negotiations is through the lens of twolevel games. Robert Putnam (1988) initially developed the two-level games framework to understand international negotiations in which negotiations between state leaders (Level 1) were subject to domestic approval (Level 2). A similar dynamic may be at work in civil war negotiations. In civil wars, agreements forged between the leaders of a state and the leaders of rebel groups often require some form of political ratification. This is most obviously true for the leaders of the state apparatus, who may find negotiated settlements to civil wars blocked by independent legislatures or powerful factions within the government and society. For example, in 2004 the moderate Sri Lankan Prime Minister Ranil Wickremesinghe, who had managed to negotiate a durable ceasefire with the Liberation Tigers of Tamil Elam (LTTE), found his base of power wiped away. Sri Lanka’s hardline president Chandrika Bandaranaike Kumaratunga accused the Prime Minister of giving too much ground to the Tamil rebels and disbanded parliament. Subsequent elections swung the balance of power in the direction of the Sinhalese nationalists.6 The two-level game dynamic, however, may also apply to rebel groups. Rebels often aspire to supplant the existing state and, not surprisingly, rebel organizations also must deal with the political challenges of policy formulation and policy execution.7 Rebel groups face serious communication challenges and as a result are typically forced to decentralize their 5

This project uses a behavioral approach in which agents learn from past experiences and adjust their behavior accordingly. Other prominent studies of the prisoner’s dilemma (Axelrod 1984, 1997) rely upon a rationalist logic that stresses the shadow of the future, which is a forward looking logic similar to the precipice argument. This parallel is drawn out in greater detail in the next section. 6 See Amy Waldman’s April 15, 2004 article “Governing Party Lacks Majority In Sri Lanka” in the New York Times. 7 For a discussion of rebel group organization and discipline, see Jeremy Weinstein’s (2007) Inside Rebellion.

5

organizational structures, relying upon multiple local commanders to conduct operations. Findley and Edwards (2007, 588) argue that rebel groups are forced to balance the benefit in coordination gained from greater centralization against the risks of creating organizational vulnerability that the government can exploit by target the group’s leadership. Thinking about civil war negotiations as a two-level game is useful because it refocuses attention on how the institutional and political configurations of the government and the opposition can affect the ability for an agreement to emerge, take hold, and endure. How civil war participants are organized has important ramifications for how they make decisions. For this study I argue that many of the organizational and institutional configurations found in civil wars can be represented by majoritarian or consensus decision rules. These two decision rules are pure archetypes and are unlikely to map perfectly onto real world situations. Nor should these two archetypes be considered a comprehensive list, yet working through the logic of these two decision rules may help shed light on more complex policy making structures. The Unitary Actor Decision Rule This project began with the observation that decisions to fight or negotiate in civil wars are not made by unitary actors but by groups of actors operating through decision rules. Although this paper does not investigate the unitary decision rule, there may be cases in which the unitary decision rule is appropriate. The unitary actor assumption may be institutionally sound for a highly authoritarian regime8 Also, it has been argued that in times of crisis decision making concentrates at the top (Boyer 2000; James and Oneal 1991). While the unitary actor assumption may be more accurate in certain situations, in most instances policy selection is complicated process involving numerous actors. The Majoritarian Decision Rule The majoritarian decision would be an appropriate heuristic for at least three situations. 8

An attempt is made by Young and Urlacher (2007) to model authoritarian regimes, by weighting the preference of an authoritarian leader and the preferences of other key actors. The “leader” selects a strategy but incorporates the preferences of other agents into its own decision.

6

First, the majoritarian decision rule may operate in an informal way in small groups. Advisory councils to government leaders or a coalition of leaders representing a rebel movement may select strategies through an informal voting scheme. The applicability of the majority decision rule to small group interactions is largely dependent on the norms of the group.9 Second, a majoritarian decision rule parallels scenarios in which decisions are made in certain parliamentary situations. This is certainly the case for minority governments that govern by building ad hoc coalitions to respond to each problem. It is also appropriate for supra-majority coalitions in which the defection of more extreme parties does not result in the government collapsing.10 Third, the majoritarian decision rule may be appropriate during periods of democratic elections in which the decision to negotiate or take a more aggressive stance features prominently. In the case of a presidential election the “moderate” canidate and the “hardline” candidate reflect policy orientations, which are then fixed for a president’s term. A president’s position is of course free to evolve time; however, unless this evolution is the result of electoral pressures, it would not be appropriate to attribute the shift to a majoritarian decision rule. The Consensus Decision Rule The second decision rule examined in this study is consensus. Although few policies are explicitly crafted using a consensus decisions rule, the consensus rule may be an appropriate heuristic for civil war negotiations in a variety of situations. Below I outline three institutional or organizational configurations commonly found in civil wars that in practice resemble a consensus decision rule. First, institutional checks and balances on the side of the government may create multiple 9

There is qualifier to the small group voting argument: if the group has fallen victim to group think then the majoritarian decision rule may not be appropriate even if decisions are formally voted on. In these situations the pressure to achieve consensus is likely to override the outward decision mechanism. 10 The majoritarian decision rule is most appropriate for these two scenarios because the main coalition partner, which likely controls the agenda, does not face a potential veto by other coalition members. A prime minister’s primary concern in either of these situations is finding the required votes and not holding a governing coalition together.

7

veto points in the policy process so that any attempt to build a coalition in favor of peace must be agreed to by multiple political bodies. There is a broad literature on the impact of veto points in state policy formation, largely driven by the work of George Tsebelis (2002; 2000; 1995).

The veto players framework has also been applied to civil wars by David

E. Cunningham (2006), who found that the more rebel factions that have a veto over the negotiating process, the longer conflicts last. In general, the potential for substantial policy change is reduced by additional veto players, particularly when veto players take extreme positions (Tsebelis 2000, 464).11 Second, the consensus decision rule may reflect a political reality in which a leader’s political base is highly fractured. Attempts to cooperate and negotiate may bring about the defection of key elements of a formerly unified coalition, effectively ending a peace building effort. Tsebelis (1995, 306) has designated actors with the ability to block policy through defection from a coalition as “partisan veto players.” He developed this in the context of parliamentary governments; however, the challenge of holding hardline factions in check as peace moves forward confronts both government and rebel coalitions. It is not uncommon in peacebuilding for spoilers to emerge from formerly unified fronts as factions see opportunities for gain or in reaction against a particular concession. A third scenario in which the consensus decision rule is a useful heuristic is when the opposition is represented by multiple independent organizations. This is certainly a common in civil wars.12 In some cases, these multiple rebel groups may band together to form a united political front, but they may also operate independently with only minimal attempts to coordinate strategy and tactics. For example, during the 1990s, the Israeli government 11

Tsebelis points out an exception to the general rule that additional veto players narrow the policy space. When the preferences of an additional veto player are subsumed within the policy preferences of the other veto players, there is no practical effect on the ability to make policy. Cunningham (2006) incorporates this as an important part of his argument, stressing the divergent preferences of the various actors. 12 For a concise discussion on the reasons behind the emergence of multiple rebel groups, see David E. Cunningham’s (2006) article “Veto Players and Civil War Duration.” Cunningham (2006, 878) argues that multiple factions may be the product of splintering or the emergence of new groups that are “dissatisfied with the current policy (or the government’s position) and also with the policy preference of the other original groups in the conflict.”

8

conducted most of its negotiations with the leadership of Fatah. Fatah, however, was in a weak position to negotiate a peace agreement on behalf of the Palestinian people. Fatah could only rhetorically speak on the behalf of other militant groups and certainly did not have the level of operational control over the activities of groups like Hamas that would be necessary to effectively deliver on a negotiated settlement.13 If a rebel group cannot deliver a settlement, there is little incentive for the government to make concessions. In the face of multiple rebel groups, the process of coordinating a political settlement is all the more difficult14 In addition to coordination problems there are practical problems associated with trying to build peace when the opposition is represented by multiple independent groups. The government may find itself trying to buy off multiple spoilers.15 On the flip side, rebel groups face the temptation to hold out as long as possible because the group that settles last can reap disproportionably large rewards (Cunningham 2006, 880).. The above sections have attempted to detail how decision rules can affect the prospects for a negotiating a settlement to civil wars. Division within the government or the opposition can certainly threaten conflict resolution; however, not all breakdowns in peace building are attributable to internal divisions. Zartman (1985, 274) cautions that a unified leadership committed to continued conflict creates doldrums that can stall any peace effort. This draws attention back to the issue of preferences. Peace can only move forward when conflict participants are ready to attempt conflict resolution, but the ability of conflict participants to capitalize on the re-alignment of preferences in favor of settlement depends on the structures by which they make decisions. Thus, a more coherent theory of conflict resolution would recognize the importance of 13

In addition to the Palestinian cause being represented by multiple groups, Fatah’s military structure also suffered from inadequate control over local commanders, compounding the challenge of peace building. Evidence of this weak central control can be seen in recent attempts to restructure Fatah’s command and control over its militias. This restructuring is discussed in the April 1, 2007 New York Times article “Palestinian Faction Aims to Unite its Militias.” 14 It has, certainly, been demonstrated that coordination becomes increasingly difficult as the number of actors increases (Olson 1965).. This observation is also made by (Bloomfield et. al. 1998, 69-70). in a practitioners handbook on civil war negotiations. 15 For more on the role of spoilers in the peace building process, see Stedman (1997).

9

both the individual willingness of conflict participants to pursue conflict resolution and the constraints created by the institutional dynamics that conflict participants must work through to actually pursue peace. In the next section I develop a framework for better understanding how institutional configurations affect two-level negotiations.

2

Research Design

2.1

The Simulation

In this project I use a computer simulation of the prisoner’s dilemma to investigate the impact of group decision making structure on the prospects for achieving stable cooperative agreements. Computer simulations of the prisoner’s dilemma have been used to advance political science research, most famously by Robert Axelrod (1997; 1984).

These studies

found that rational agents can escape the the prisoner’s dilemma’s Defect-Defect trap when the game is iterated. An alternative approach to the prisoner’s dilemma, developed in sociology by Michael W. Macy (1991a; 1991b; 1989), relies not on the rationality of agents but on their ability to adjust behavior in response to the success or failure of past strategies. Macy’s behavioral model has been adapted to the study of political science by several scholars (Kanazawa 1998; Young and Urlacher 2007). The study by Young and Urlacher (2007) pioneered a new approach in the study of group decision making and the prisoner’s dilemma, focusing on the relative effectiveness of majoritarian and authoritarian decision rules in achieving cooperative outcomes. The Young-Urlacher study found that majoritarian mechanisms were generally more efficient in overcoming the challenges inherent in the prisoner’s dilemma. This project attempts to build on this earlier work by incorporating a consensus decision rule that is more applicable in civil wars where opposition forces may be highly fractured and where each faction may have an effective veto over the process. In an attempt to make the simulation as transparent as possible, this section discusses the nuances of the simulation decision rules and learning mechanism. While this detailed 10

discussion of the simulation mechanics is important for purposes of replication, the simulation was originally conceived of as a simple six step process. Step 1: Individual agents select strategy (Cooperate or Defect) through a random process based on each agent’s cooperation tendency. Step 2: Agents vote for their selected strategy in the selection of a group strategy. Votes are aggregated according to the group’s decision rule. Step 3: Groups play their selected strategies in a two-player prisoner’s dilemma. Step 4: Payoffs from the prisoner’s dilemma are assigned. Step 5: Agent cooperation tendencies are adjusted to reflect group performance. Step 6: The next iteration begins at Step 1. Each agent in the simulation possesses a unique cooperation tendency that can vary from 0 to 1. This cooperation tendency reflects an agent’s propensity to cooperate. Higher cooperation tendencies correspond to higher probabilities of cooperation. To simulate the conflictual environment of a civil war, agents begin the simulations with cooperation tendencies set at 0. In Step 1 each agent selects a strategy of cooperation or defection. This is done by comparing an agent’s cooperation tendency to a randomly generated number drawn from a uniform distribution ranging from 0 to 1. If the agent’s cooperation tendency is greater than or equal to the random number, then the agent will choose to cooperate. If the Cooperation tendency is less than the randomly generated number, the agent will opt for defection. Once agents have selected their preferred individual strategies, these preferences are aggregated into a group decision using either a majoritarian or consensus mechanism (Step 2). For the majoritarian mechanism at least 3 of the 5 agents must opt for cooperation for the group to cooperate. For the consensus rule, all 5 agents must vote for cooperation. If one agent opts for defection, then the group as a whole will defect. This sets defection as the status quo. While this may appear a normative bias toward conflict, it can be justified in two ways. First, the dominant strategy for the Prisoner’s Dilemma is defection, so the game

11

by its very nature is conflictual and encourages defection. Second, the real world situation that this simulation seeks to emulate is peace building in civil wars. In civil wars conflict and mistrust are deeply entrenched, and failure to build peace implies a continuation of violent conflict. Once groups have selected strategies, these strategies are played in a standard two person prisoners dilemma game16 (Step 3) and payoffs are assigned to agents according to the outcome of the game (Step 4)(see Figure 1). In step 5, the cooperation tendency for agents is adjusted in response to the payoff to the group from step 4. The learning process is based on Macy’s (1991a) model, which is presented in equation 1.

Player 1

Cooperate Def ect

Player 2 Cooperate Def ect 1, 1 −2, 2 2, −2 −1, −1

Figure 1: Payoffs in the Classic Prisoner’s Dilemma. The learning mechanism for this simulation is a modification of the classic Bush-Mosteller (1955) learning mechanism, which reinforces or erodes the cooperation tendency of agents in response to success or failure in the game. Decisions to cooperate that lead to successful outcomes reenforce the cooperation tendency while cooperative decisions that result in unsatisfactory outcomes decrease the cooperation tendency of agents. Macy (1989) modified the Bush-Mosteller mechanism in several conceptually important ways (see Equation 1). Equation 1 elaborates the change to the cooperation tendency (p) of agent j from iteration i. The first bracketed section of the equation reflects the change produced from a decision to cooperate (C = 1). The second bracketed section of the equation reflects the change that results from a decision to defect (C = 0). A critical factor in the learning mechanism is the parameter Oi,j , which sets the learning rate. Macy’s first adaptation of the Bush-Mosteller learning mechanism was to make the value of Oi,j proportional to the magnitude of the 16

PD: (T > R > P > S and 2R > T + S) where T is the payoff for player 1 from a DC outcome, R is the payoff from CC outcome, P is the payoff for a DD outcome, and S is the payoff for CD.

12

payoffs (in this simulation Oi,j = .1).17 Payoffs from CD or DC outcomes have twice the impact on agents as payoffs from CC or DD outcomes because the greater the payoff, the more that the cooperation tendency of an agent is altered. Also, because Oi,j is proportional to the payoffs, Oi,j can take on both positive and negative values, which affects the direction of learning. Positive values reenforce previous behavior. Negative values prompt agents to move away from unsuccessful strategies. In addition to allowing Oi,j to vary, Macy also adapted the Bush-Mosteller learning mechanism to incorporate an element of diminishing returns into the learning process. This is done by including

1 |Oi,j |

as an exponent in the equation. The effect of this element is that

the learning rate decays asymptotically as an agent’s cooperation tendency approaches either 0 or 1.

( |O1

pi+1,j = pi,j + [Oi,j (1 − pi,j

i,j |

)

( |O1

)Ci,j ] − [Oi,j (1 − pi,j

i,j |

)

)(1 − Ci,j )]

(1)

Working through the logic of this learning mechanism and the payoffs available to agents reveals that agents can be pushed toward cooperation in two ways. First, a CooperateCooperate outcome results in mutual satisfaction with the decision to cooperate. Thus, the cooperation tendencies of agents are nudged upward, making cooperation more likely in the future. Second, a Defect-Defect outcome results in mutual dissatisfaction with the decision to defect. Thus, the cooperation tendencies of agents are again pushed upwards. Defect-Defect outcomes make agents more willing to cooperate because defection fails to achieve a satisfactory payoff. Just as in the hurting stalemate logic, this dissatisfaction with a Defect-Defect outcome may eventually prompt one or both groups to attempt cooperation. An offer of cooperation, however, is a risky strategy and leaves the cooperating group open 17

To establish the robustness of this simulation’s findings, other values for Oi,j were also tested, including .05 and .2. It is important to examine different values for Oi,j because if this value is set too low, cooperation is exceedingly difficult. Setting Oi,j at .1 allows for groups to lock into cooperation in only 8 symmetric moves. Doubling Oi,j lowers the bar for achieving cooperation while setting Oi,j at .05 makes it much more difficult for groups to lock into cooperative equilibriums (Macy 1989, 212).

13

to exploitation. If the other group meets cooperation with defection, the result is that both groups conclude that cooperation does not work. Agents in the cooperating group will lower their cooperation tendencies because their cooperative efforts were taken advantage of. Agents in the defecting group reap benefits from exploiting the other group and will seek to do so in the future, increasing the likelihood of defection. This dynamic may result in groups becoming trapped in an uncooperative equilibrium. The Defect-Defect outcome is unworkable, but unilateral attempts to cooperate are even worse. The only hope that groups have to escape from the uncooperative equilibrium is to coordinate strategies. Defection must be met with defection until both groups are ready to simultaneously attempt cooperation. If cooperation is met with cooperation, then the willingness of agents to attempt cooperation in the future is increased. This corresponds to confidence building measures that are so important in creating a foundation of trust in negotiations. The result of synchronized choices is that cooperation becomes increasingly stable as cooperation tendencies of agents approach 1. It cannot be emphasized enough that without synchronization there is no way out of the uncooperative equilibrium. To evaluate the impact of majoritarian and consensus decision rules on the prospects for achieving synchronization and by extension cooperation, several computer simulations were conducted. These simulations all involved two groups of five agents. Groups were assigned either majoritarian-majoritarian, consensus-consensus or majoritarian-consensus decision rules. For each combination the simulation was run 200 times, resulting in a total of 600 sets of interactions between groups. Each set involved two groups engaging in 21 iterations of the prisoner’s dilemma.18 The interaction histories for each group were then analyzed for the total number of cooperative decisions and to identify if a group was able to establish a stable pattern of cooperative 18

From a rationalist perspective, fixing the number of iterations threatens the stability of cooperation. Because the rationality assumption posits agents that are forward looking and utility maximizing, agents that know a game is ending cease to be motivated by the shadow of the future and play the final iteration as if it were a one-shot prisoner’s dilemma. From a behavioral perspective, agents are backward looking. They pursue the strategies that have been shown in the past to be effective. Thus, fixing the number of iterations does not cause any particular problem.

14

behavior by the end of 10 and 21 iterations.19 The robustness of these findings was evaluated by modifying the number of iterations, the starting points, and learning rates. The results of these additional simulations, which conformed to expectations, are not presented in here, but are available along with the simulation code on [the author’s personal webpage].

3

Findings of the Simulation

The goal of this analysis is to distinguish between the degrees of cooperation achieved by groups with different decision rules. To do this, the results of the simulations are analyzed in several different ways. One approach is to calculate the mean number of cooperative decisions. Cooperative choices by groups were coded as 1 and decisions to defect were coded as 0. Thus the mean value corresponds to the proportion of interactions in which a group chooses to cooperate. The second approach was to code a group as having achieved a stable constituency in favor of cooperation. If a group selected cooperative strategies in each of the last five iterations, then the group was coded as having achieved a stable cooperative equilibrium. T-tests are used to assess differences in cooperative behavior.

3.1

Comparison of Decision Rules

The simplest way to compare the different levels of cooperation resulting from variations in group decision rules is to look at the proportion of cooperative decisions. It should be noted that a decision by a group to cooperate is different than a cooperative outcome in the prisoner’s dilemma. A group can opt to play a cooperative strategy in the game, but for the game to produce a cooperative outcome, both groups must choose to cooperate. For this reason, the numbers of decisions to cooperate does not in itself reflect more cooperative interactions because groups may find themselves being exploited for their attempts at co19

Twenty-one iterations may seem an odd number of iterations for a simulation. For the first iteration all agent cooperation tendencies are set to 0, which rules out the possibility of cooperation. Thus in the 21 total iterations, there are only 20 actual opportunities for agents to select a cooperative strategy.

15

Table 1: Mean Levels of Cooperation and Decision Rules (First 10 Iterations) Paired Combinations Mean Majoritarian .1888 Consensus .0695

Mean Diff. .119

Std. Error P-value .008 .000

Majoritarian Mixed

.1888 .1005

.088

.007

.000

Consensus Mixed

.0695 .1005

-.031

.006

.000

N = 400 for each group.

operation. Yet, because agents learn from past interactions, it is unlikely that a group that was victimized for trying to cooperate will continue to cooperate indefinably. The only way for groups to sustain repeated cooperative decisions is to build a cooperative equilibrium. For the first 10 iterations, the interaction of two groups using majoritarian decision rules resulted in cooperative choices by groups in 18.88% of the time. As rare as cooperation is in these early stages for Majoritarian groups, the interactions of two consensus groups resulted in markedly fewer cooperative decisions. Consensus groups opt for cooperation just under 7% of the time when facing other consensus groups. This is a statistically significant difference compared to majoritarian-majoritarian groups as well as the mixed majoritarian-consensus groups (see Table 1). The groups in the mixed simulations selected cooperation 10.05% of the time. A closer look reveals that the bulk of these decisions to cooperate are attributable to the majoritarian groups in the mixed simulations.20 The initial pattern emerging from the first 10 iterations is consistent with the arguments from the veto-players literature. The consensus decision rule makes it difficult to deviate from the status quo. This is seen in the differences between majoritarian-majoritarian and consensus-consensus simulations and within the majoritarian-consensus simulations. When 20

Within the majoritarian groups cooperation occurred 19.1% of the time. For consensus groups 10% of interactions involved cooperation. This is comparable to the patterns seen in the consensus-consensus and majoritarian-majoritarian simulations.

16

Table 2: Mean Levels of Cooperation and Decision Rules (All 21 Iterations) Paired Combinations Mean Majoritarian .3324 Consensus .3061

Mean Diff. .0263

Std. Error P-value .014 .056

Majoritarian Mixed

.3324 .1560

.1764

.011

.000

Consensus Mixed N = 400 for each group.

.3061 .1560

.1501

.011

.000

the full 21 iterations are taken into account; however, the story becomes more complex. When the full 21 iterations are considered the differences between consensus-consensus and majoritarian-majoritarian decision rules evaporate. The interaction of two groups using majoritarian decision rules resulted in cooperative choices by groups in 33.24% of the time. Surprisingly, the interactions of two consensus groups resulted in comparable levels of cooperative decisions. Consensus groups opt for cooperation just over 30% of the time when facing other consensus groups. This difference is just short of statistical significance, and is substantively negligible (see Table 2). The improvement of the consensus-consensus decision rule is all the more surprising given how infrequent cooperation was in the first 10 iterations. When the full 21 iterations are examined, the least cooperative combination was the mixed combination in which one group used a majoritarian decisions rule and the other used a consensus decision rule. Groups in the mixed simulations offered to cooperate a mere 15.6% of the time, the only group to not increase its level of cooperation from the first 11 iterations. This difference is statistically significant from the simulations that used either strictly Majoritarian or only Consensus decision rules. While the average number of cooperative decisions is a useful starting point for comparing the ability of groups with different decision rules to achieve cooperative outcomes, it is also important to look at where groups end up. Because of the random process behind

17

Table 3: Number of Groups Achieving Cooperative Equilibrium

Group Size Majoritarian Consensus Mixed

Number of Groups 134 175 7

95% Confidence Interval Lower Bound Upper Bound 115.49 152.52 155.84 194.56 -.05 12.36

strategy selection, even agents with very low cooperation tendencies will opt for cooperation occasionally. A second approach focuses more specifically on the ability of groups to lock into cooperative outcomes. Table 3 compares the number of groups that locked into cooperation (defined by 5 consecutive cooperative decisions in the final iterations of the simulation). The ability of groups to achieve cooperative equilibrium does not clearly mirror the proportion of cooperative decisions across decision rules. Of the 400 groups in the 200 majoritarian-majoritarian simulations, 134 managed to lock in a cooperative strategy. The consensus-consensus groups performed somewhat better with 175 groups locking into cooperation. By comparison the mixed decision rule simulations performed quite poorly. Only a tiny fraction of groups (7 out of 400) in the mixed simulations involving one majoritarian group and one consensus group were able to end the simulation with a stable cooperative pattern. The differences between these three outcomes are statistically significant. The simulations has produced a somewhat surprising set of findings. First, contrary to initial expectations consensus-consensus combinations are quite effective at achieving stable cooperative equilibriums. Consensus-consensus combinations start out slow but quickly make up ground, achieving rates of cooperation comparable to what is seen in the majoritarianmajoritarian simulation. Second, it is fair to say that the interaction of different types of decision rules (majoritarian and consensus) are disastrous for efforts at achieving cooperation. Clearly, the issue is not that majoritarian or consensus decision rules are bad for cooperation. The decision rules perform quite well when both groups use the same decision mechanism. 18

The problem appears to stem from the use of different mechanisms for aggregating group preferences. In the next section I reflect on what might account for this as well as try to explain the surprisingly strong performance of consensus-consensus combinations.

3.2

Explaining the Differences

To explain why the consensus-consensus and majoritarian-majoritarian simulations were generally successful in building cooperation and why the majoritarian-consensus simulations failed to achieve cooperative outcomes 98% of the time, it is necessary to dig deeper into how these mechanisms operate. A key part of this explanation is how decision rules translate agent preferences into group strategies. Figure 2 illustrates the probability of a group with five agents choosing to cooperate given a particular mean value for agent cooperation tendencies. The majoritarian and consensus decision rules translate agent preferences in very different ways. This is important because, as the earlier discussion stressed, the secret to escaping the Defect-Defect trap is to coordinate strategies so that both groups grow dissatisfied with defection at the same rate and move in tandem toward a more cooperative world. Once groups become willing to attempt cooperation both must simultaneously switch over to cooperation to sustain and nurture the drift toward the cooperative equilibrium. Failure to stay synchronized can cause the willingness of agents to cooperate to plummet. The majoritarian-majoritarian combination is effective at keeping groups synchronized. When the average cooperation tendency of a group is less than .5, majoritarian groups are more inclined to choose defection than the individual cooperation tendencies of the agents would suggest. Similarly, when the average cooperation tendency of majoritarian groups is greater than .5, groups are more likely to choose cooperation than would be expected given the average value of cooperation tendencies (see Figure 1). Essentially, majoritarian groups are more predictable at the extremes. This is important for the prospects of achieving stable cooperation because when majoritarian groups are in an uncooperative world, they reliably choose to defect, but after a few rounds of pain from a Defect-Defect outcome, majoritarian 19

Figure 2: Probability of Cooperation Across Cooperation Tendencies

20

Table 4: Mean Levels of Cooperation and Decision Rules (First 10 Iterations) Paired Combinations Mean Majoritarian .375 Consensus .139 Majoritarian Mixed

Mean Diff. .236

.375 .221

.154

Std. Error P-value .015 .000 .013

.000

N = 400 for each group.

agents become more amenable to cooperation. The majoritarian mechanism enables the group strategy to quickly shift in response to the change in preferences of agents. It may take several additional iterations to build a stable coalition in favor of cooperation, but essentially majoritarian-majoritarian groups achieve peace by learning from the past and changing course. The mechanism underlying the consensus decision rule operates somewhat differently. Because the consensus rule sets a very high bar for selecting cooperation, defection is almost a given until the average agent cooperation tendencies begins to approach the maximum value of 1 (see Figure 2). Consensus groups hold to the defect course long after the individual members are willing to consider cooperation as a viable strategy. Even if all members of a consensus group are inclined to cooperate 90% of the time, the probability of the group actually choosing to cooperate would be just under 60%. When two consensus groups interact with each other, they engage in merciless defection. This unrelenting defection continues until all parties are utterly dissatisfied with defection (cooperation tendencies approaching 100%). Unfortunately, the consensus path to peace requires several additional rounds of punishment from defection to achieve a cooperative equilibrium that is needed for the majoritarian path to peace. Paradoxically, consensus groups achieve peace by learning but not adapting accordingly. Or more precisely, they are structurally blocked from adapting until defection has been totally discredited as a strategy for all agents.

21

The majoritarian path to peace (learning and adapting)and the consensus path to peace (learning but not adapting) do not work well when paired together. After a few iterations with Defect-Defect outcomes, all agents start to become dissatisfied with the pain that defection brings. Majoritarian groups are able to translate this dissatisfaction into tentative offers to cooperate. Consensus groups, in spite of having agents that are equally dissatisfied with defection, hold almost unflinchingly to the defection course. Majoritarian groups will make offers to cooperate which are almost certainly met by defection. The result is that agents in the majoritarian group feel that they have been exploited and are less likely to cooperate in future iterations. The agents in the consensus group also become less likely to cooperate in future iterations because their steadfast commitment to defection has finally paid off. The decision rules translate agent preferences into policy choices in different ways, which can seriously undermine the ability of groups to synchronize their strategies.

4

Applying the Logic

This project has focused up until this point on working out the logic of different decision rules in the prisoner’s dilemma game. With the exception of a few anecdotes, there have not been any concrete links between the simulation logic and real world civil war negotiations. In this next section, I attempt to show the parallels between the logic of two-level social trap and a real world case.

4.1

Case Selection and Methodology

A useful case for exploring majoritarian-consensus situations should meet the following four criteria. 1) The government should continue to function throughout the civil war. 2) The government should be democratic (preferably parliamentary system with a minority government rather than presidential system). 3)The opposition should be highly fractured. 4) The various factions should hold heterogenous demands. When the first two of these cases

22

are considered, the pool of potential cases would include Great Britain, Spain, India, and a handful of others. When the second set of conditions are imposed, India’s struggle against the Kashmir insurgency stands out as a promising case to evaluate the claims put forward in this study, not only because it meets all four criteria, but also because over the course of the 18 year insurgency, India has been governed by both minority governments and coalition governments. Thus in a single case study it is possible to evaluate the the simulation’s logic for both majoritarian-consensus and consensus-consensus scenarios. The one potential drawback to focusing on the Kashmir insurgency is the challenge of disentangling the conflict from its larger context of India-Pakistan relations. The waxing and waning of violence within Kashmir is to some extent a reflection of relations between India and Pakistan. Third party involvement is certainly common in civil wars, but it does introduce an additional factor that needs to be considered.21 The following case study of the Kashmir insurgency is intended as a plausibility probe. Obviously, a single case cannot offer conclusive evidence in favor of an argument. That having been acknowledged, a single case study can falsify an argument, particularly when the case is a “most likely” case (Rogowski 1995). If the following case parallels the patterns seen in the simulation, then this opens the door for a more systematic study. If the following case is incompatible with the logic of the simulation that should be sufficient to reject the simulations applicability to civil war cases.

4.2

The Kashmir Insurgency: Decision Rules and Civil War

The conflict in Kashmir is a legacy of the complex process of Indian partition following the collapse of British rule. Kashmir is India’s only majority Muslim province22 and has been a 21

The impact of third parties and negotiated settlements is difficult to disentangle. It has been argued that third party support for one side in a conflict decreases the likelihood of a negotiated settlement by increasing the prospect of victory (Mason et al 1999). Patrick Regan (2002), who makes a similar argument has found little or no systematic impact on conflict duration from biased intervention. Furthermore, there has been some research to suggest that biased third parties may be able to extract concessions form the factions that they support, making settlement more feasible (Svensson 2007) 22 The state of Jammu and Kashmir is as a whole predominantly Muslim; however, the muslim population is not distributed evenly across the three main regions within Jammu and kashmir. The Kashmir valley is

23

source of enduring tension between India and Pakistan, including three wars in 1947, 1965, and 1999. Both countries argue that they have a legitimate claim to Kashmir. Pakistan has grounded its claim on the principle that partition was decided on the basis of religious distribution. Kashmir with its majority Muslim population ought to have gone to Pakistan. Yet, Jammu and Kashmir were a princely state within the British Raj. Consequently, the decision on Kashmir’s fate lay not with the British but with Maharaja Hari Singh. Under attack from forces crossing the Pakistan boarder, Singh appealed to India for aid and legally agreed to accession to India. This sparked a larger war between India and Pakistan and Kashmir was divided as a practical matter along what came to be known as the Line of Control (LOC). Initially The Indian government granted Jammu and Kashmir a great deal of autonomy and economic assistance. This autonomy was slowly whittled away and the politics of Jammu and Kashmir became increasingly turbulent. By the end of the 1980s there was widespread dissatisfaction with the political process. Although it appears that the uprising in Kashmir began domestically, Pakistan quickly moved to assist the insurgency, seeing it as a low cost way to weaken India (Ganguly 2001, 92). A number of militant groups received assistance from Pakistan, and the flow of militants from Pakistan into Kashmir has been a point of contention between the two countries. The outbreak of the Kashmir insurgency was initially met with brutal force. Jammu and Kashmir were brought under the direct control of the central government. The Indian government carried out an aggressive campaign involving as many as 500,000 troops and a variety of irregular pro-Indian militias. The Jammu and Kashmir Distributed Areas Act passed 1990 gave the Indian security forces almost unlimited latitude to crush the Kashmir insurgency (Ganguly 1997). The result was tens of thousand dead, and several hundred thousand displaced due to a systematic pattern of human rights abuses by the Indian military predominantly Muslim. By contrast Jammu at the start of the conflict had a slight Hindu majority. The Hindu population in Jammu has swelled as a result of Hindu displacement from the Kashmir valley. Lastly, the area of Ladakh has a plurality Buddhist population.

24

and pro-Indian militias not to mention abuses by Kashmir insurgents (Human Rights Watch 1999). The aggressiveness of the Indian response to the Kashmir insurgency stems in part from the way in which Kashmir touches on the identity of India and its antagonism with Pakistan. Yet, at least some portion of the hardline response can be trace to parliamentary politics. In 1990 the Janata Dal coalition led by V.P. Singh was in a precarious position and relied heavily on the Hindu nationalist Bharatiya Janata Party (BJP) party for its continued survival (Sridharan 2003; Ganguly 2001, 93).

While a softer hand might have prevented

the further alienation of the Kashmir population from India, it was not good politics. During this first phase of violence virtually no substantive negotiations between the two sides took place. This is not surprising given the leverage the BJP had over the ruling coalition. In 1991 the Janata Dal coalition was replaced and the Congress party formed a minority government. The Congress minority government was less beholden to parties such as the BJP, which continued to advocate for a hardline stance. During this period, India made offers to negotiate with militants at regular intervals23 and attempted to take advantage of offers made by militant factions to negotiate.24 Two factors scuttled negotiations in this early period. First the Indian government put forward the condition that all negotiations take place within the framework of the Indian constitution. This was an unacceptable starting point for the pro-succession groups. Yet, India’s inflexibility on conditions for talks was matched by the insurgents. The leading political-militant front, the All Party Hurriyat Conference (APHC or ‘Hurriyat’), which represented several dozen secessionist groups, held 23

India offered to negotiate with Kashmir militants, albeit within the framework of the Indian constitution on a number of occasions. Offers to negotiate were made in 1991 1993, 1994, and 1997. Each of these offers was rejected. See the December 2nd, 1991 Agence France Press story “Kashmir militants reject peace talks: minister” for a description of the 1991 offer. The 1993 offer is documented in the Uppsala Conflict Database case description for the Kashmir conflict. The 1994 offer is described in the April 12, 1994 story “Indian govn’t ready to hold talks on Kashir: minister” put out by the Xinhua News Agency. Lastly, according to the Associated Press’s July 28 1997, article ”Indian prime minister modifies offer to kashmiri militants” India made a short lived attempted to move India toward unconditional talks with Kashmir militants. 24 In On several occasions militant groups in Kashmir expressed a willingness to negotiate with India. This occurred in 1996, when a handful of militia leaders proposed talks. See the February 9th 1996 Inter Press Service article “India-Kashmir: militant Peace Offer May Lack Key Support” for a brief description of the tensions between militia groups prompted by this move.

25

fast to its demand that Pakistan be included in any substantive talks. The demands laid down on both sides offered little room for negotiations to occur. The second barrier to negotiations in this early period stems from the internal struggles between hardline and moderate factions. Efforts by moderates on both sides to move away from these conditions for negotiation were challenged by hardliners and where possible blocked. In July of 1997, Indian Prime Minister Gujral made an offer for unconditional talks with Kashmir militants. The BJP verbally lashed out at the Prime Minister for his acquiescence and ”attitude of total surrender” The Prime Minister quickly clarified his offer, arguing that India could only negotiate with those who laid down their weapons.25 A similar scenario, involving more aggressive tactics, occurred on the insurgent side in 1996. Four prominent militant leaders broke with the Hurriyat position and offered to negotiate with India without Pakistan’s participation. Other militant groups denounced the leaders for their willingness to “sellout” to India, and the Hurriyat Conference organized a general strike to disrupt the negotiations. Following elections in 1996, the Janata Dal coalition again returned to power. It forged majority government known as the United Front and drew support from a number of parties outside of the coalition government, including the Congress party. Although the United Front government was unstable, it was not beholden to hardline parties for its continued survival as it had been from 1988 to 1991. In 1998, the BJP managed to find a winning strategy moving toward the center and building a broad but unstable and short-lived coalition(Behera 2002 360). Not surprisingly, this unstable coalition did not make a strong push to restart the peace process in Kashmir. In 1999 the BJP managed to score a major electoral victory and built a robust supramajority coalition that could survive the defection of multiple coalition partners (Sridharan 2003, 141). Following the 1999 Kargil War, India was looking for progress on the Kashmir conflict. 25

Gujral’s offer is detailed in the Associated Press’ July 28th 1997 article “Indian prime minister modifies offer to Kashmiri militants” The reaction of the BJP is covered by the Queensland, Australia news organization Courier Mail in a July 29th 1997 story “Gujral in Kashmir backflip.”

26

This resulted in several tentative overtures to the Hurriyat Conference, including the release of several Hurriyat leaders. Seeing the shift, extremist militants threatened to strike at the leadership of the Hurriyat Conference if it were to move toward dialogue with India (Samii 2006, 69). Vajpayee also faced pressure from the hardliners within his own coalition, but the coalition was broad enough that it was not an imminent threat to the government. In July 2000, the militant group Hizb-ul-Mujahideen unexpectedly offer cease-fire. Rather than bringing a reprieve in the violence, other groups escalated violence in the face of Hizb-ulMujahideens move toward peace. Hizb-ul-Mujahideen also struggled with internal battles between hardliners and moderates that brought the cease-fire to an end within weeks. The short lived cease-fire did not undermine Indias determination to find a way forward in Kashmir, and in November India offered its own cease-fire. This unprecedented step almost overshadows the fact that Vajpayee also managed to work around India’s earlier requirement that all talks with militants be within the framework of the Indian constitution (Behera 2002, 360-361).

While the cease-fire was an important opening for peace, the militant groups

continued to oppose any progress on talks, and the Hurriyat Conference was deeply divided over how to respond to the Indian overture. Violence continued throughout the cease-fire and provided the opening for militant groups to advance into urban areas. In May 2001 India ended its cease-fire and changed its strategy yet again. India attempted to implement a two prong strategy: dialogue with Pakistan over Kashmir and military confrontation with the militants in Kashmir. By 2002, key leaders of the Hurriyat Conference were again moving toward dialogue with India. This move was preempted by the assassination of the leading moderate voice within the Hurriyat Conference, Abdul Gani Lone. The assassination shifted the balance within Hurriyat to the Hardliners. Faced with a hardening Hurriyat Conference, India attempted to neutralize the leading militant voice within the Hurriyat, Syed Ali Shah Geelani, arresting him in June 2002. This temporarily worked, and the Hurriyat Conference took a more open position on dialogue with India and were less critical of local elections in 2002. Geelani,

27

once released from prison, quickly set about reversing this conciliatory drift by the Hurriyat Conference. The increasing divide between hardliners and moderates was not limited to the Hurriyat Conference. The Hizb-ul-Mujahideen militia, which had earlier taken tentative steps towards talks with India was weakened by the assassination of Abdul Majid Dar in March 2003. The split in the ranks of the militants created new opportunities and new challenges. Free of hardline constraints the moderate faction of Hurriyat was able to move forward on talks with India without the condition of Pakistani participation.26 Yet, the split in the Hurriyat Conference also moved the moderate Hurriyat faction further from the militants, decreasing the group’s ability to actually deliver a settlement should negotiations succeed. The return to power of the Congress party in November 2004, brought a temporary disruption of the BJP’s talks with Hurriyat leaders. The new Indian Prime Minister Manmohan Singh, however, the new Congress government, which was built around a robust supra-majority coalition, took steps that the previous government had been unwilling to take, allowing leaders of Hurriyat to travel to Pakistan to meet with militant groups on the other side of the Line of Control. This concession sparked a shrieks of protest by hardliners on both sides. The split in the Hurriyat Conference between moderates and hardliners deepened further as the leading voice in the hardline faction Syed Ali Shah Geelani refused to travel with the Hurriyat delegation even though the right to travel to Pakistan had been a long standing demand of the Hurriyat leadership. Congress also found itself under fire from former Prime Minister Vajpayee for the concession to Hurriyat.27 The Hurriyat Conference met with the Indian Prime Minister in late 2005, but this was not leveraged into a full scale peace process. Throughout 2006 Singh shifted his peace building efforts to a series of round tables intended to bring together the leaders from across Kashmiri society. The Hurriyat 26

See “Peace by Piece” in the February 2nd, 2004 issue of India Today. The open letter criticizing the Hurriyat visit sent by A.B. Vajpayee to Prime Minister Singh on June 15th 2005, is available on line at http://www.bjp.org/Press/june 1605.htm. Also, for a discussion of the BJP’s electoral calculations during this period, see Jonah Blank’s (2003, 186-188) article “Kashmir: All tactics, no strategy.” in India Review. 27

28

Conference boycotted both round tables as did the leadership of most of the militant factions.

4.3

Evaluating the Kashmir Case

The logic of the simulation predicted that India would be able to move to the table and to make concessions with much greater ease than would the highly fractured Kashmir independence movement. This was particularly true when the India was ruled by a moderate minority government or a supra-majority government that could withstand the defection of a few parties. Both sides wrestled with divisions between hardliners and moderates; however, the success of hardliners in blocking concessions by moderates was not the same for the Indian government and the Kahsmiri militants. India was able to compromise on two contentious points: demanding that the Indian constitution be the only framework for talks and allowing the Hurriyat leadership to travel to Pakistan. In both instances, the hardline BJP denounced the moves, but the BJP was not able to prevent these concessions from moving forward. The Kashmir independence movement, by contrast, was unwilling to move toward peace. Although the Jammu and Kashmir Liberation Front (JKLF) set aside its violent tactics in 1994, other militant groups did not. Hurriyat, which clearly had broad based public support throughout much of the 1990s was unable to make concessions. Efforts by Hurriyat leaders to work with India to find a viable solution were rejected by militant groups and eventually fractured the Hurriyat coalition. The willingness of the various parties to negotiate and make concessions is not sufficient to explain the intractability of the Kashmir conflict. Parties on both side seek a peaceful solution and parties on both sides are determined to not move a single inch. When the Indian government operates under a majoritarian decision rule with no relevant veto players, moderates are able to push forward with peace initiatives. On the Kashmiri side it is the fractured nature of the Kashmir independency movement, reflecting a consensus decision rule, that enables hardliners to scuttle peace efforts. The basic narrative of the Kashmir 29

conflict seems to mirror the simulation logic. The case study, however, highlights several areas in which the simulation does not accurately capture the nuances of real world interactions. This is not necessarily a problem, as models are by definitions simplifications; however, it is worth noting what the model does not anticipate. First, the simulation does not capture the tendency of extremist militant groups to target moderates. Negotiations are undermined not only by the continuation of violence, but also because would be peacemakers are themselves targeted. Second, the simulation has no way for groups to identify and support potential peace partners in the other group. Indian leaders were certainly able to to distinguish between foreign militant groups, local militant groups, and the more moderate Hurriyat political front. India regularly looked for ways to strengthen the moderates and build a foundation for a settlement. While the failure of this approach underscores the challenge of a divide and settle strategy, it is certainly a reasonable approach when facing an opponent operating under a consensus decision rule.

5

Conclusions

This project set out to investigate how the decision structures of governments and rebel groups might affect the prospects for peace. The results of the computer simulation helped uncover several useful insights into how decision structures affect the ability of conflict participants to overcome conflict. The first major finding is that the timing of ripeness be driven in large part by institutions and the internal political divisions of the government and the opposition. The majoritarian and consensus decision rules seem to suggest different pathways to ripeness. With a majoritarian system, building a coalition in favor of peace is comparatively easy once it becomes clear that continued conflict is not working. Agents recognize quickly that defection does not work, and their group is able to change strategies, testing cooperation and potentially learning that it is a viable approach.

30

By contrast, the consensus decision rule makes it very difficult for conflict participants to build the unified position in favor of peace. Consequently, defection remains the norm long after conflict participants have learned that conflict is not a viable path. The consensus rule extends conflict until all participants are dissatisfied with continued violence. It may take a long time for all conflict participants to loose their stomach for violence, but once this happens, cooperation is quickly locked in as the new equilibrium. I argued earlier that there are two pathways to peace. Majoritarian groups learn and quickly adapt. Consensus groups learn but are much slower to adapt. Both pathways are consistent with the hurting stalemate logic; however, they offer different predictions about when the hurting stalemate will produce moments of ripeness and the level of effort needed to realize peace. The majoritarian decision rule suggests that less violence is needed before conflict participants will be ready to attempt negotiations, but this pro-peace coalition may be very fragile. A greater emphasis on confidence building may be needed to build broadbased support for a durable peace. The consensus decision rule suggests that violence may be protracted, but once the hurting stalemate produces ripeness, a durable settlement should follow quickly. Again, this is consistent with the hurting stalemate logic, but it adds an additional level of refinement, linking it to institutions and internal political dynamics. It has been argued that little can be done to move conflict participants toward a durable peaceful settlement, and the best strategy available to the international community is to hold back let civil wars work themselves out (Luttwak 1999; 2001). The findings of this simulation acknowledge the wisdom of this wait and see advice for a subset of conflicts and rejects it for other conflicts. Coercive tactics such as sanctions, embargoes or intervention may raise the cost of continued conflict by some small degree for one or both parties. This increase in costs may only matter if the target of the sanctions is able to change course in response to increased costs. Majoritarian decision rules allow for the dissatisfaction of agents to be translated into actual policy changes. Consensus rules effectively block policy change even if there is an increase in the costs of continued conflict.

31

To get conflict participants using consensus decision rules to the negotiation table may require the imposition of more extreme costs. This should serve as a warning to those advocating for coercive intervention. For a third party to force conflict participants to the table may require inflicting very heavily casualties over an extended period of time. For this reason, military planners and policy makers should think carefully about the political and organizational structure of an opponent before attempting to use force to compel peace. A hurting stalemate may provide the energy for moments of ripeness, but decision rules conflict participants use to make decisions help to determine if ripeness actually emerges.

Bibliography Allison, Graham T. 1971. Essence of Decision: explaining the Cuban Missile Crisis. Boston, MA: Little Brown. Axelrod, Robert. 1984. The Evolution of Cooperation. New York, NY: Basic Books. Axelrod, Robert. 1997. The Complexity of Cooperation. Princeton, NJ: Princeton University Press. Behera, Navnita Chadha. 2002. “Kashmir: A testing ground.” South Asia: Journal of South Asian Studies 25(3):343–364. Blank, Jonah. 2003. “Kashmir: All tactics, no strategy.” India Review 2(3):181–202. Bloomfield, David, Charles Nupen & Peter Harris. 1998. Negotiation Processes. In Democracy and Deep-Rooted Conflict: Options for Negotiators, ed. Peter Harris & Ben Reilly. Stockholm: Institute for Democracy and Electoral Assistance. Bohrer, Robert E. II & Caroline A. Hartzell. 2005. After the Shooting Stops: Civil War Settlements and the Postwar Environment. In International Studies Association Annual Meeting. Honolulu, HI: . 32

Boyer, Mark A. 2000. “Issue Definition and Two-Level Negotiations: An Application to the American Foreign Policy Process.” Diplomacy & Statecraft 11(2):185–212. Bush, R. R. & F. Mosteller. 1955. Stochastic Models for Learning. New York: John Wiley and Sons Inc. Collier, Paul, Anke Hoeffler & Mns Sderbom. 2004. “On the Duration of Civil War.” Journal of Peace Research 41(3):253–273. Cunningham, David E. 2006. “Veto Players and Civil War Duration.” American Journal of Political Science 50(4):875–892. Fearon, James D. 1998. Commitment Problems and the Spread of Ethnic conflict. In The International Spread of Ethnic Conflict, ed. David A Lake & Donald Rothchild. Princeton, NJ: Princeton University Press pp. 107–126. Fearon, James D. 2004. “Why Do Some Civil Wars Last So Much Longer Than Others?” Journal of Peace Research 41(3):275–301. Findley, Michael G. & Scott Edwards. 2007. “Accounting for the Unaccounted: Weak-Actor Social Structure in Asymmetric Wars.” International Studies Quarterly 51:583–588. Ganguly, Sumit. 1997. The Crisis in Kashmir: Portents of War, Hopes of Peace. Cambridge: Cambridge University Press. Ganguly, Sumit. 2001. Conflict Unending: india-Pakistan Tensions since 1947. New York, NY: Columbia University Press. Goertz, Gary & Paul F. Diehl. 1997. “The Initiation and Termination of Enduring Rivalries: The Impact of Political Shocks.” American Journal of Political Science 39(1):30–52. Haass, Richard N. 1990. Conflicts Unending. New Haven, CT: Yale University Press. India: Behind The Kashmir Conflict. 1999. Technical report Human Rights Watch. 33

James, Patrick & John Oneal. 1991. “The Influence of Domestic and International Politics on the President’s Use of force.” Journal of Conflict Resolution 35(2):307–332. Janis, Irving L. 1972. Victims of groupthink: a psychological study of foreign-policy decisions and fiascoes. Boston, MA: Houghton Mifflin. Jervis, Robert. 1976. Perception and Misperception in International Politics. Princeton, NJ: Princeton University Press. Jervis, Robert. 1978. “Cooperation Under the Security Dilemma.” World Politics 30(2):167– 214. Kanazawa, S. 1998. “A Possible Solution to the Paradox of Voter Turnout.” The Journal of Politics 60:974–995. Kleiboer, Marieke. 1994. “Ripeness of Conflict: A Fruitful Notion?” Journal of Peace Research 31(1):109–116. Licklider, Roy. 1993. Stopping the Killing: How Civil Wars End. New York, NY: New York University Press. Luttwak, Edward N. 1999. “Give War a Chance.” Foreign Affairs 78:36–44. Luttwak, Edward N. 2001. The Curse of Inconclusive Intervention. In Turbulent Peace: The Challenges of Managing International Conflict, ed. Chester A. Crocker, Fen Osler Hampson & Pamela Aall. Washington, DC: United States Institute for Peace pp. 265– 272. Macy, Michael W. 1989. “Walking out of Social Traps: A Stochastic Learning Model for the Prisoner’s Dilemma.” Rationality and Society 1(2):197–219. Macy, Michael W. 1991a. “Chains of Cooperation: Threshold Effects in Collective Action.” American Sociological Review 56(6):730–747. 34

Macy, Michael W. 1991b. “Learning to Cooperate: Stochastic and tacit Collusion in Social Exchange.” American Journal of Political Science 97:808–843. March, James C. & Johan P. Olsen. 1984. “The New Institutionalism: Organizational Factors in Political Life.” American Political Science Review. 78(3):734–749. Mason, T. David, Joseph P. Weingarten & Patrick J. Fett. 1999. “Win, Lose, or Draw: Predicting the Outcome of Civil Wars.” Political Research Quarterly 52(2):239–268. Olson, Mancur. 1965. The Logic of Collective Action; public goods and the theory of groups. Cambridge, MA: Harvard University Press. Putnam, Robert D. 1988. “Diplomacy and Domestic Politics: The Logic of Two-Level Games.” International Organizations 42(3):427–460. Rasler, Karen. 2000. “Shocks, Expectancy Revision, and the De-Escalation of Protracted Conflicts: The Israeli-Palestinian Case.” Journal of Peace Research 37(6):699–720. Regan, Patrick M. 2002. “Third-Party Interventions and the Duration of Intrastate Conflicts.” Journal of Conflict Resolution 46(1):55–73. Rogowski, Ronald. 1995. “The Role of Theory and Anomaly in Social-Scientific Inference.” American Political Science Review 89(2):467–470. Samii, Cyrus. 2006. “Seizing the Moment in Kashmir.” SAIS Review 26(1):65–78. Siqueira, Kevin. 2005. “Political and Militant Wings within Dissident Movements and Organizations.” Journal of Conflict Resolution 49(2):218–236. Sridharan, Eswaran. 2003. “Coalitions and Party Strategies in India’s Parliamentary Federation.” The Journal of Federalism 33(4):135–152. Stedman, Stephen J. 1991. Peacemaking in Civil War: International Mediation in Zimbabwe, 1974-1980. Boulder, CO: Lynne Rienner. 35

Stedman, Stephen J. 1997. “Spoiler Problems in Peace Processes.” International Security 22(2):5–53. Stein, Janice Gross. 2001. Image, Identity and the Resolution of Violent Conflict. In Turbulent Peace: the Challenges of Managing International Conflict, ed. Charles A. Crocker, Fen Osler Hampson & Pamela Aall. Washington, DC: United States Institute of Peace pp. 189–208. Svensson, Isak. 2007. “Bargaining, Bias and Peace Brokers: How Rebels Commit to Peace.” Journal of Peace Research 44(2):177–194. Touval, Saadia & I. William Zartman. 2001. International Mediation in the Post-Cold War Era. In Turbulent Peace, ed. Charles A. Crocker, Fen Osler Hampson & Pamela Aall. Washinton, D.C.: United States Institute of Peace. Tsebelis, George. 1995. “Decision Making in Political Systems: Veto Players in Presidentialism, Parliamentarism, Multicameralism and Multipartyism.” British Journal of Political Science 25(3):289–325. Tsebelis, George. 2000. “Veto Players and Institutional Analysis.” Governance: An International Journal of Policy and Administration 13(4):441–474. Tsebelis, George. 2002. Veto Players: How Political Institutions Work. Princeton, NJ: Princeton University Press. Walter, Barbara F. 2002. Committing to Peace: Successful Settlements of Civil War. Princeton, NJ: Princeton University Press. Young, Joseph K. & Brian R. Urlacher. 2007. “Cantankerous Cooperation: Democracies, Authoritarian Regimes, and the Prisoner’s Dilemma.” International Interactions 33(1):51– 73.

36

Zartman, I. William. 1985. Ripe for Resolution: Conflict and Intervention in Africa. New York: Oxford University Press. Zartman, I. William. 2000. Ripeness: The Hurting Stalemate and Beyond.”. In International Conflict Resolution After the Cold War, ed. Paul C. Stern & Daniel Druckman. Washington, DC: National Academy Press. Zartman, I. William & Johannes Aurik. 1991. Power Strategies in De-Escalation. In Timing the De-Escalation of International Conflicts, ed. Louis Kriesberg & Stuart Thorson. Syracuse, NY: Syracuse University Press pp. 152–181.

37