email icon Email this citation


The Regulation of Political Conflict: A Game Theoretic Analysis

Ali Carkoglu

Bogazici University,
Department of Political Science and International Relations

International Studies Association
March 1998

Abstract

Conflict and cooperation constitute the basis of much interest in political analysis. Nevertheless, political theory abounds in disagreement as to which predominates, which constitutes the norm and as to the reasons for their occurrence. Game theoretical models have been applied to these problems with some success. The proposed article aims to contribute to our understanding of the phenomena of creating cooperation, and thus regulating a situation of conflict between two actors by a third party. A third party will be introduced to a prisoners' dilemma game and conditions for its cooperation creating regulatory strategies will be derived. The impact of informational and institutional structures on the prospects for cooperative outcomes will be discussed. Applications to governments' policies in regulating domestic conflict and to a hegemon's problem of maintaining order in the international arena will be shortly discussed.

Conflict and cooperation, as they are used in daily conversations, are vague terms that are associated with a wide array of different concepts. Although an extensive body of literature deals with conflict and cooperation in political interactions the same degree of vagueness often permeates these analyses too. In many instances lack of cooperation among the relevant actors is seen as the root cause of conflict and too often, either implicit or explicit in the analyses, lack of conflict is seen as the driving force for the creation of cooperation among actors. Often times the policy minded researchers find themselves occupied in discussions of how conflict situations can be geared towards cooperative solutions where actors do cooperate towards the achievement of a common objective.

Situations of conflict that are particularly interesting for political analysis arise between conscious and rational actors that wish to "carry out acts which are mutually inconsistent" (Nicholson, 1992, 11). Consciousness of the actors derives from the fact that these actors are aware of their wants, needs or obligations but this does not necessarily imply rational acts. Rationality of actors arises in situations where their behavior is goal directed. Independent of the morality of these goals rational actors choose a set of "best" strategies to achieve predetermined set of ends (Morrow, 1994, 17). Simply put, as long as strategy choices and implementation of these choices of actors are consistent with their goals the actors are said to be rational.

Many times the situation arises where conscious and rational actors are in a conflict situation but their actions do not necessarily involve any commitment of their resources for a forceful application of their strategy. Yet in other situations a particular strategy may be followed by actors that will necessarily result in the imposition of a solution in their own favor, and thus imply a loss for or simply the destruction of the other side. Both are situations of conflict; one is a dormant and the other is an active conflict situation. However, both arise due to rational choices of conscious actors.

Quite often one quickly and wrongly assumes that the dormant conflict situation is a reflection of cooperation among the actors. Following Keohane (1984, 51-52) in international relations literature cooperation is said to occur when "actors adjust their behavior to the actual or anticipated preferences of others, through a process of policy coordination". 1 Implicit in this conception is the fact that all actors adjust their actions or policies so as to reduce their negative consequences for the other parties. This does not mean that their "best" option among other available actions is necessarily the one that occurs due to cooperation. In fact many times just the opposite is the case; the cooperative actors choose an action that is not the best among available options just to avoid inflicting negative consequences on others since doing so may result in similar acts by the other parties that are costly on itself.

This conception of cooperation implicitly assumes that actors have goal-oriented behavior thus imply that all relevant actors are rational in the most simply defined way. It also implies that cooperation involves gains or rewards to all parties involved compared to the case of mutual non-cooperation (Milner, 1997, 7). Simply put, cooperation not only requires goal-directed rational behavior, but also the presence of a situation where mutual gains are available only in case of coordinated acts of agents that may, for some at least, run counter to their individually defined or uncoordinated acts as function of their goal seeking behavior.

It is important to note here that there are many alternatives to cooperative behavior. Besides the obvious choice of non-action, in unilateral behavior agents do not take the impact of their own action on other agents. Such action need not always create negative results in terms of hurting other agents or in terms of decreasing the loosely defined welfare of a group of agents. The working of market mechanism where each individual acts on the sole basis of individual utility maximization is the obvious example for such a context. However, in yet many other situations that form the focus of my attention in this paper the situation is very much contrary to such contexts.

In the well known prisoner's dilemma (PD) situation the two actors face a situation where their individual goal-oriented behavior always leads them to a non-cooperative action despite the fact that coordinated or cooperative action would improve upon their situation of mutual defection. The dilemma situation continues even in case the two agents could communicate to coordinate their actions in a cooperative way because neither one of the players have any incentive to keep their promises.

PD games are used to illustrate many different social situations such as the problem of collective action, environmental degradation or the problem of the commons, problematic nature of creating social reform. 2 Recently, Cohen (1994) applied the classic PD situation to gain some insight into the process of democratic breakdown and occurrence of military coups. Below I attempt to show that with some modification of the PD framework that makes the military third party to two civilian actors an active strategic player in the game of military coup one can create a more generic game where regulation of political conflict as reflected by the PD situation can be analyzed. A series of generic game situations will be discussed where, not only insights can be gained about policy oriented questions on the regulation of PD based conflict, but also on the nature of dynamic interaction in immature democracies between the military and the civilians.

1. Prisoner's Dilemma and the Military Coup Game

Excellent surveys of the literature on military regimes can be found in Collier (1979), Linz (1979), Lutwak (1979), Nordlinger (1977), Pastor (1989), Pinkey (1990), Stepan (1988). Instead of attempting a thorough review of previous analyses I will instead try to formulize a critique of their underlying methodology as a basis for the ensuing analysis.

Three interrelated explanatory frameworks of coup occurrence can be distinguished from the literature. First concentrates on broadly defined environmental factors due to political and economic development, social mobilization or modernization (e.g. Deutch, 1961; Huntington, 1968; Olson, 1963; Needler, 1966; Germani and Silvert, 1967; Jackman, 1976). In the second, inner dynamics of the military establishment interacting with socio-economic environment is emphasized and a certain `habituation' phenomenon is asserted to explain the greater tendency in some countries to experience military interventions (Alexander, 1958; Zolberg, 1968; Londregan and Poole, 1990; O'Keane, 1983). In the third approach, violent political conflict and differences in consequent reactions of political systems to it, forms the basis of explanation for the occurrence of military regimes (Eckstein, 1964-1965; Gurr, 1967-1968; Lieuwen, 1962; Needler, 1963; Pye, 1962).

The literature on military regimes concentrates on identifying common structural characteristics of coup stricken countries. These characteristics are then asserted to determine the political system's tendency to have military regimes. Most of the works on military coups are historical accounts of singular country experiences. Comparative studies of country experiences try to establish some kind of general pattern from among the preceding environmental factors of countries that experienced coups. Only recently these general patterns have been tested empirically (Jackman, 1978; O'Kane, 1981, 1983, 1987; Londregan and Poole, 1990; Carkoglu, 1994).

The problem with these kind empirical evaluations is simply the fact that no satisfactory account of the dynamics of interaction between key players can be given. Empirical generalizations do not exhaust possibilities of different reactions of key players to each other and necessarily give only a partial account of the actual phenomena, or implicitly assert an unsatisfactory determinism to the phenomena of coups.

Below I first briefly evaluate Cohen's (1994) analysis and build a simple model as a consequence of its criticism. In ensuing sections I develop several models of increasing complexity that not only allow strategic behavior by the civilians but also the military. By this line of analysis I aim to move beyond deterministic or probabilistic analyses that neglect the decisions of key players and provide a more insightful explanation of the coup phenomenon.

1.1 The Coup Game: A First Attempt 3

The game is played in a country with a multi-party competitive democracy. The economic conditions of the country reflect a severe and continuous crisis situation. The government sector is large, mobilized based on political incentives of patronage distributing politicians and thus forms the largest center of inefficiency in the economy. As an outcome of rapid urbanization, economic development that deteriorates income distribution and cultural change the electorate is very fractionalized. The electoral support is spread across many political parties. As a consequence parties resort to ideological polarization with hopes of differentiating themselves from their competitors and solidifying their electoral support facing high levels of volatility. Due to insufficient electoral support for any one party, even after intense patronage distribution, the political system eventually becomes immobilized. No single party can establish a majority government and legislate according to its own mandate.

Ideologically the party system is decomposed into moderate left and right wing parties as well as hostile camps of extreme left and right wing parties. Extreme right and left parties both have direct connections to legal and illegal organizations engaged in violent activities to pressure any ruling government that threatens their extreme policy stands. Moderate right and left wing parties need to maintain good relations with their own extremists, in order first to have an ideological harmony and credibility. Secondly, these extremists provide a mobilizing electoral force that serve as a credible deterrence against extremists of the other camp.

The immediate consequence of political immobilization is an intense struggle to capture state mechanisms to be used against the opposing camp. Eventually rule of law is lost and the political frictions are carried into streets in the form of violent clashes between the opposing groups. The game is assumed to be played right at this stage. All parties know that a violent climax of political conflict is near. No one party can be certain that as a result of this show down it will remain as the winner and the costs of continuing the violent struggle are considerable. All parties are also aware of the fact that their electoral basis is slowly shifting towards a primary preference for stability and order as opposed to remaining solidly behind the above outlined party struggle. Accordingly, besides the civilian parties, a new player, the military rises in prominence. The civilians are aware of the possibility of military intervention due to either a history of military interventions or the historical weight of the military in the country's political arena. More importantly, the apparent inability of the civilian parties to control, and regulate, political conflict as well as responding to social and economic demands of large segments of the electorate with anything other than simple patronage distribution, leads to serious deterioration of trust in civilian solutions and undermines their support in favor of a non-democratic solution to rising political conflict. Fully aware of its rising popularity and trust amongst the electorate the military shapes its decision to intervene or remain in the side lines allowing the civilian competition run its full course.

Due to electoral fractionalization neither moderates of the left or the right-wing parties are in a position to form a ruling majority coalition with the extremists of their own camp. Any ruling majority coalition has to be formed by cooperating with the moderates of the other side. Cohen (1994) builds a model on these same principles by asserting that the moderates of the left and the right are engaged in a PD game. In such a context both sides' dominant strategy becomes defection rather than cooperation with the other moderate party. Cohen follows the usual PD situation and asserts that each moderate party's pay-off from the game is lowest when they choose to cooperate while their counterpart defects, and highest when they defect while their opponent cooperates. The mutual defection pay-off is nevertheless lower for both sides than the pay-off from mutual cooperation but higher than the unilateral cooperation pay-off. Cohen's (1994, 67) model for the occurrence of coups as a result of simultaneous strategic action by the moderates of the political system is reproduced in Figure 1 below:

The model above has no strategic action by the military. The game is played between the moderates who are assumed to be under the electoral influence of the extremists. The influence of the extremists is implicit in the payoff structure given by the ranking T>R>P>S that lead to the PD outcome: defection is the dominant strategy for both players. Under mutual defection condition the moderates coalesce with their extremists against the opposing ideological camp, no agreement is reached. Political polarization rises and Cohen seems to argue that military intervention is inescapable. The military's decision to intervene or to stay out is not explicitly included in the model. The impact of the military action upon the payoffs of the political parties is totally ignored. However, perhaps the most striking characteristic of coup stricken countries is the predominance of the military as an important political player to be reckoned with. As I will illustrate below, the neglect of military participation in political decisions runs not only against empirical evidence but it also critically undermines our full understanding of the dynamics that lead to democratic breakdown.

1.2 Alternative Models: Electoral Incentives and Strategic Military Action

As a first attempt to remedy the defects of the above model the situation is modeled as a three player two stage extensive form game between two parties L and R and the military M, illustrated in Figure 2. 4 For simplicity, all actors are assumed to be homogenous groups. The civilians move first and simultaneously choose whether to cooperate or not. Here cooperation means recognition of the urgent problems the country is facing. Underlying this recognition is an acceptance of certain policy positions that may run against the party's immediate policy mandates and thus certain policy moderation as a remedy to the rising political conflict. Defection means further polarization in the system. The defecting party resorts to a political alliance with ideological extremists of their own side of the spectrum. Such an alliance between the moderates and extremists is assumed to have some electoral benefits but increasing attempts to suppress the competitors if necessary by using violence overshadows these. The military moves after observing their decisions, to either intervene or to let the civilian exchange continue.

All payoffs in the model are stylized. The payoffs for political parties are assumed to be measurable in terms of electoral losses and rewards. For the political parties the net electoral gain is the basis of their decision to cooperate or defect. For the military the payoff involves rewards and costs for intervention. The net reward forms the basis of the decision to intervene. The military is assumed to compare the net rewards for intervention with the popularity it has under status quo.

The military faces three distinct situations. The first one is the case of mutual cooperation where the two political parties have openly moved towards cooperating with each other to resolve the crisis situation prevailing in the country. The second one is the unilateral cooperation or unilateral defection case where either one of the political parties choose to defect while the other cooperates. The third distinct situation is the case of mutual defection where both parties choose to defect. While mutual cooperation and mutual defection cases are non-problematic in terms of the ease of attributing responsibility to the parties in question, the unilateral defection case is problematic. I assume for simplicity that the responsibility of defection can be attributed with certainty to any one party. In other words, the military knows whether L or R had defected. Accordingly at the second stage of the game the military has four distinct information sets.

Under the mutual cooperation condition both parties enjoy an electoral reward (r C L and r C R ) from some section of their constituency as well as being punished (e C L and e C R ) for cooperating with the other party by yet some other groups within the supporters of the party. As a result, I focus on net electoral benefit. The net electoral gain (r C i - e C i ) C,C for i=L, R need not be equal to each other, and it could be both negative or positive. The military will incur costs (C C ) as well as a reward (W C ) for carrying out a coup. However, the net reward of the military for carrying out a coup (W C - C C ) C,C when mutual cooperation prevails is expected to be negative and lowest among all possible cases.

When the military carries out a coup, a certain level of oppression will have to be exercised over the political parties in order to suppress a civilian challenge to the new regime and thus to control the populace. This oppression is assumed to be measurable in electoral loss terms. Political parties accordingly incur a loss of net electoral support due to military oppression (Op C i C,C ) i=L,R in case when a coup occurs. If the military chooses not to intervene, the political parties will continue to enjoy their net electoral gains without having to bear the costs associated with the oppression of the military when a coup was carried out. Clearly, when the military observes without ambiguity that the representative civilians do cooperate it infers that strong net electoral support has accrued to L and R at the expense of other extremist parties. Accordingly, it will be very hard for the military to effectively control and regulate masses according to their wishes. Accordingly, when and if mutual cooperation occurs, the military does not intervene at all since the payoff for intervention is lower, in fact negative, for this case whereas staying out under status quo will continue to bring a positive support from the populace. This status quo popularity of the military (Pop sq ) CC under mutual cooperation is assumed to be positive.

The option to intervene is assumed to be open to the military under all possible situations of cooperation and defection combinations of the civilians. Under the unilateral defection condition the military finds itself faced with either cooperation by the L but defection by the R, or cooperation by the R and defection by the L. At M2 and M3 the military clearly observes the side which cooperated and the side which defected. The reaction of the M to the parties is difficult to model under unilateral defection. It all depends on whether the parties agree to disagree with each other but form a strong resistance to military regime or disagree with each other but remain open to cooperation with the M to eradicate or weaken their competitors.

Of course another complication is about the M's willingness to join forces with one of the parties to concentrate politically disabling oppression upon their competitors. According to one line of thinking the M is seen open to cooperation with a civilian group. Few military regimes last undisturbed by competitors. In the long run competing factions or civilian regimes often replace them. Accordingly, on the part of the military, the willingness to join forces with one of the parties seems a preferable option in both the short and the long run. That will only increase the M's ability to survive any challenge that may arise against them. However, the other line of argument sees critical risks in politicizing the military. Especially, the professional structure of the military will deteriorate in such a coalition of forces between the military and one of the civilian parties. The corrupt patronage structure of the civilians will transform the military into yet another untrustworthy political player. Such a development clearly brings lower payoffs especially to the elite groups in the military, which accordingly opt for no alliance with the civilians. On the other hand, if such deterioration can be avoided there seems no reason why the rewards of intervention should be shared with another party. Accordingly, the military is to remain distant to any alliance with the civilians.

It is also doubtful that any political party would survive such cooperation with the military and remain credible facing the electorate after the regime is transferred back to civilians. At least the efforts to remain credible may not worth the rewards of cooperation with the military. Accordingly, I will assume that, in general, no matter how stubborn the civilian parties might be in accepting cooperative terms with their civilian competitors, if they are to remain credible facing the electorate they would not cooperate with the military in order to weaken or destroy their competitors.

However, since civilian cooperation is required to resolve rising political conflict in the nation a unilateral defection context will facilitate the job of the military to establish their rule upon the populace. The defecting party will join forces with their extremists to defeat their competitor. The cooperating party will loose the support of their extreme factions and they will be reduced to a strictly minority political force. In other words, the payoff structure under unilateral defection is given by the following inequalities: (r D L - e D L ) D,C > (r C R -e C R ) D,C and (r D R - e D R ) C,D > (r C L - e C L ) C,D . The defecting political party will find ample opportunity to follow a policy stand exclusively shaped by their own ideological position. However, this policy package needs to be differentiated from the military stands if it so happens that the M decides to intervene and follow a similar policy package. Then the defector will be perceived by the electorate to be cooperating with the M. The defector dominant party will strongly resist to any move that can destroy their electoral basis and will thus resist the M rule openly. The cooperating party will at this stage be deserted by their extremists and possibly face action against them by their now dominant competitors. For long term electoral support they will not seek military's help to survive in the short-term. However, they will not be a credible threat to the M either. In short, the total political power of the civilians will be lowered thus enabling the M to establish their regime with lower levels of oppression.

Accordingly, I assume that a defector can be easily punished by the M for not cooperating and placing the party priorities ahead of bringing a halt to rising political conflict. Since defection can be diagnosed unambiguously, the punishment by the M is going to be unequal for the two parties. The oppression applied to the defector will accordingly be higher than the one applied to the cooperating side OpD L D,C > OpD R D,C (or similarly it will be OpD R C,D > OpD L C,D ). 5

In other words, controlling the defection of one side by oppression is not more difficult than it is the case when the other side defects. As a consequence of this assumption the payoff for the military will be independent of the defecting side being the R or the L; i.e.(W D1 -C D1 ) C,D = (W D1 -C D1 ) D,C . Intervention option under unilateral defection brings a positive support from the populace higher than the status quo popularity of the military under unilateral defection or simply (W D1 - C D1 ) C,D = (W D1 - C D1 ) D,C > (Pop sq ) C,D = (Pop sq ) D,C . This last inequality reflect a critical observation. Any failure by the political parties to cooperate will create such an atmosphere of political manoeuvring by the defector, now the dominant party, that will not be tolerable by a non-trivial section of the population at large. Merging with its extremist the defector is assumed to pursue a policy of exterminating the cooperating side. Such a political context of civilian political feud is not without political consequences. I assume that such a political feud will help increase the support for the M. Accordingly, the payoff for the M is expected to be greater than the status quo pay-off (Pop sq ) C,D = (Pop sq ) D,C . The payoff of the defecting party is assumed to be larger than the cooperating party which obtains the lowest possible payoff in the game by unilaterally cooperating.

Under the mutual defection case the military faces a situation where both the L and the R refuses to cooperate with each other and thus chooses to coalesce with their allies at their own extreme end of the ideological spectrum. These choices are made with the expectation of higher net electoral support to be attracted from the extremist groups. However, these electoral returns now will have to be downsized considerably by the M's policies in case it chooses to intervene. The M in this case faces two groups of parties. Both the L and the R join forces with the extremists on their end of the ideological spectrum to compete with the other if necessary by force. The military if it decides to intervene will have to apply more oppression compared to unilateral defection case. The net reward of an intervention to the M is given by (W D2 -C D2 ) D,D > (Pop sq ) D,D .

In evaluating the pay-off structure of the M, I assume that (Pop sq ) C,C > (W D2 - C D2 ) D,D > (W D1 - C D1 ) C,D = (W D1 - C D1 ) D,C > (Pop sq ) C,D = (Pop sq ) D,C > 0 > (W c - C c ) C,C . Across different stages of the game the highest payoff for the M occurs when mutual cooperation arises from the civilians' exchange and the M does not intervene. In other words, among all possibilities the M is expected to choose staying out when mutual cooperation occurs. The M's second highest payoff is expected to arise due to its intervention when both L and R defects since the chaotic and violent competition between the political parties increase the rewards more than the costs of intervention. Similarly, I assume that the net reward for intervention under mutual cooperation case is to be negative since then the civilian side would have a compact undivided determination to resolve the country's problems through a civilian process. The cost of maintaining order and legitimacy under this case will simply be too high. The case of unilateral defection is open to different possibilities. For simplicity I assume that the pay-off in this case is positive and greater than the pay-off under status quo (Pop sq ) C,D = (Pop sq ) D,C . If this pay-off is different for the case of defection by the L than it is the case for defection by the R then the implication would be that the military is better off having one side defecting than the other side defecting. However, as I assumed before the ease of controlling the defectors are one and the same. If the defection of the L is leading to an intervention pay-off higher than the case under unilateral defection by the right; i.e. (W D1 - C D1 ) D,C > (W D1 - C D1 ) C,D , one is lead to believe that the military has a higher tendency for intervention when the defecting side is the L rather than the R. It should also be noted that due to costs of maintaining the military organisation under control and also keeping the level of satisfaction of the electorate at a sufficiently high level the payoff for the military to intervene is always less than no intervention under mutual cooperation. However, military benefits from intervention when and if the first stage game results in unilateral or mutual defection.

I will also assume for both the L and the R that when the party in question unilaterally defects it receives a higher pay-off than it are the case when both sides cooperates. Mutual cooperation in return brings a higher pay-off then mutual defection, which corresponds to a higher pay-off than cooperation when the other side defected. In short we have the usual prisoners dilemma pay-off structure: (r D L - e D L ) D,C > (r C L - e C L ) C,C > (r D L - e D L ) D,D > (r C L - e C L ) C,D and for the R we have (r D R - e D R ) D,C > (r C R - e C R ) C,C > (r D R - e D R ) D,D > (r C R - e C R ) C,D .

1.2.1 Analysis

Applying backward induction to the last stage moves of the military we see that the game can be reduced back to a version of two-person simultaneous move game. In this reduction I assume that the military decides to stay out when mutual cooperation exists, but decides to enter under all other three options since expected pay-off for doing so is strictly larger than the status quo popularity of the M. The strategic form of the resulting game is given in Figure 3 below:

In order to have defection as the dominant strategy for L we see above that the following conditions need to be satisfied:

((r c R - e c R ) c,c ) < ((r D L - e D L ) D,C ) - Op DL D,C
(r c L - e c L ) C,D - Op DL C,D < ((r D L - e D L ) D,D ) - Op DL D,D

or simply we have

Op DL D,C < ((r D L - e D L ) D,C ) - ((r c L - e c L ) C,C )       (1)

The above inequality states that the oppression by M to L when a military regime is established should be less than the differences in net electoral reward for L when it defected alone and when mutual cooperation existed. Since unilateral defection brings the highest net reward for L the right-hand side of the above inequality is to be positive. However, for dominance of defection the oppression applied upon the unilaterally defecting left should be smaller than the difference on the right-hand side of the inequality.

The second condition that needs to be satisfied for L to choose defection as dominant strategy is

(r c L - e c L ) C,D - Op DL C,D < ((r D L - e D L ) D,D ) - Op DL D,D
Op DL C,D - Op DL D,D > (r c L - e c L ) C,D - (r D L -e D L ) D,D< /sup>
Op DL D,D - Op DL C,D < (r D L -e D L ) D,D< /sup> - (r c L - e c L ) C,D        (2)

In other words, the difference in net rewards between mutual defection and unilateral cooperation for L should be greater than the difference between M's oppression upon L for mutual defection and unilateral cooperation.

For R, we have the following similar conditions in order to have defection as the dominant strategy:

((r c R - e c R ) c,c ) < ((r c R - e c R ) C,D ) - Op DR C,D
((r c R - e c R ) D,C ) - Op DR D,C < ((r D R - e D R ) D,D ) - Op DR D,D

or simply

((r c R - e c R ) C,D ) - ((r c R - e c R ) C,C ) > Op DR C,D   &n bsp;    (3)

Op DR D,D - Op DR D,C < ((r D R - e D R ) D,D ) - ((r c R - e c R ) D,C )        (4)

Given the assumptions we made about the relative rankings of the payoffs for players the above conditions seem to be plausibly satisfied. In other words, given the structure of payoffs the game is expected to end in mutual defection which eventually leads to military intervention maximizing only M's payoffs and minimizing both political parties' payoffs. It should be noted here that if one ignores, as Cohen (1994) does, the oppression of the military on political parties the above inequalities are trivially satisfied by simply assuming the PD payoff structure. This trivial solution can be seen if one considers that as military oppression terms are equated to zero the inequalities in (1), (2), (3) and (4) become:

0 < ((r D L - e D L ) D,C ) - ((r c L - e c L ) C,C )    (1')
0 < ((r D L - e D L ) D,D ) - ((r c L - e c L ) C,D )      (2')
0 < ((r c R - e c R ) C,D ) - ((r c R - e c R ) C,C )     (3')
0 < ((r D R - e D R ) D,D ) - ((r c R - e c R ) D,C )    (4')

The above inequalities only require that unilateral defection brings a larger payoff than mutual cooperation and that mutual defection payoff is larger than unilateral cooperation. These conditions are simple restatements of the prisoners' dilemma conditions.

However, once the effects of military action on the political parties are taken into account the occurrence of coup can be avoided depending on some conditions to be discussed below. Since the above stated conditions explicitly relate the net electoral rewards of the parties to the only political conflict regulation mechanisms, e.g. oppression, we are now in a position to speculate about the likely outcome of the game under different assumptions about the relative magnitudes of the payoffs.

First of all when the above inequalities are reversed we find a mutually cooperative outcome. The reversal of these inequalities however simply means that if the oppression by the military on both parties is larger than their net rewards difference between unilateral defection and mutual cooperation, and at the same time the difference in oppression under mutual (D,D) and unilateral defection (D,C) is larger than the net rewards for the parties under the same conditions lead to mutual cooperation. In other words, the military can use oppression as an effective tool for creating cooperation between the civilians. In other words credible threats of the military to punish the civilians according to the above stated conditions then the prisoners' dilemma game collapses into a cooperative game.

If we allow changes in the magnitudes of net rewards mutual cooperation can also be obtained. For example, keeping the oppression levels by the M constant, we see that two conditions are necessary. The first one occurs when the net reward under mutual cooperation (C,C) is larger than the net reward under unilateral defection (D,C) and (C,D). Similarly, if at the same time net rewards under unilateral defection is larger than the net reward under mutual defection, then we again reach the mutual cooperation outcome. In other words, the distribution of electoral support across the moderates can lead to a cooperative outcome even though the military might be tempted to intervene.

So far I only looked into possibility of cooperation between the civilian parties as a dominant strategy solution. A dominant strategy need not be present if a Nash equilibrium exists. What is the condition where (C,C) is the Nash equilibrium of the game presented above? If the reversed inequalities (1) and (3) are satisfied, whereas (2) and (4) are satisfied, it is enough to have (C,C) to be the unique Nash equilibrium of the above game. In other words it is enough to have military oppression on the defector to be greater than the difference between the defector's net reward and net reward under mutual cooperation. Clearly this condition can be satisfied unilaterally by the military increasing its punishment of the only defector. On the other hand, keeping the oppression of the military on the only defector and the net reward of the only defector constant as the net reward of mutual cooperation for both parties increase the condition for (C,C) as Nash equilibrium tends to be satisfied. In other words, in both the dominant strategy solution for mutual cooperation and in the Nash equilibrium solution two sources for inducing cooperation exist. One is the unilateral adjustment of the level of oppression upon the civilians as a credible threat to cooperate. The other is the structure of electoral reward that induces the political parties to rather cooperate than to defect.

The significance of this solution is that there exists both an internal regulatory mechanism to the democratic process defined by the structure of electoral support as well as an external regulatory possibility in the form of military oppression upon the defectors. So far as Cohen's (1994) PD framework is concerned our conclusion was that the military intervened because the civilians could not or simply did not cooperate with each other. The reason for their lack of cooperation was not related to their electoral incentives. Moreover, the decision of the military was a non-strategic reaction to a PD game upon which it had no influence. However, the simple full information model depicted above includes the military as a strategic player and gives an explanation to the observation of non-cooperative civilians. The civilians do not cooperate because their electoral incentives do not allow them to benefit from such cooperation. In other words, the way the electoral support in a given country context is shaped determines whether or not the civilians will be able to find a democratic solution to rising political conflict. In addition, the model above also allows for a strategic military action that may or may not choose to let the democratic solution find a way out of the political conflict in a PD situation by simply committing itself to a given level of oppression upon the defectors. By the same token it may choose to let the system destroy itself by not meeting the necessary level of regulatory oppression on the defectors.

The above model is one of full information. Very rarely agents in the real world have full information. This is certainly true for the case of destabilized democracies that are about to collapse. Therefore dynamic models of interaction between the military and civilians in an incomplete information game context should help us gain further insights into the calculus of coup making. In order to keep the model simple for the start I first develop a model where a homogenous body of civilians is assumed to have asymmetric information advantage over the military. In short, civilians are assumed to hold private information about the degree of democratic support and may use this advantage of asymmetric information to further their chances of keeping the military out of the regime. Next the asymmetric information advantage is given to the military about the degree of conflict that the country is facing. Revealing the true state of affairs or simply distorting the truth yields advantage to the military over the civilians in their interaction within the coup model framework. Below, different variants of these games are presented and their implications are shortly discussed.

2. The Coup Game under Incomplete Information

As a first attempt to capture the dynamics of interaction in an incomplete information coup game I assume again that the stylized military faces this time a homogenous body of civilians that can coordinate their actions against the military and who hold private information about their electoral support. Nature starts the game by choosing whether or not the electoral support for the civilians is strong (S) or weak (W). The probability of strong electoral support is commonly known to be p whereas the probability of weak support for the civilians is 1-p. Civilians observe the nature's selection of their electoral support type and then move to either cooperate with the military or to defect. Here a cooperative move would be one of reconciling military's demands on key policy areas with government polices so as to keep the military out of the executive office by simply taking over. Defection, on the other hand, is to follow policies that are contradictory to military's preferences on key policy areas.

Accordingly, I assume that besides the civilians' preferences that shape the nature of democratic competition there are forces within the country that seek to pursue their policy preferences through the use of the military's privileged position of having access to use of force if necessary. These third parties need not only be restricted to membership of the military institutions but they may include civilian groups coalescing with the military in realization of their policy preferences. The military's move comes after observing the civilians' actions. However, the military does not observe the nature's choice of the type of the civilians. In other words, when military moves it knows whether or not they are following a cooperative move by reconciling government policies with military's preferences or not. However, the level of public support behind civilians' move is uncertain to them whereas the civilians hold private information about this.

The military has two options. It either chooses to intervene (E) or it stays out (E). When it chooses to intervene the nature determines whether it will be a successful coup with probability q or unsuccessful coup with probability (1-q). The probability of success changes according to the type of the civilians' support. When the civilians' type is one of strong support behind them, the probability of successful coup q*, is lower than it is under weak civilian support case q. Here the asymmetric information advantage in the hands of the civilians allows them the opportunity to act as if they have a strong civilian back up behind their positions to defect with the military even when they know they do not have strong electoral backing. The problem of the military is to guess the type of the civilians since their dominant position is always to carry out a coup when civilians do not have strong electoral backing and always staying out when they do have a strong support base.

Since the above game is a typical game of incomplete information the Bayesian Nash equilibria needs to be computed. There are in total 16 possible combinations of pure strategies. These pure strategies for the civilians can be summarized as follows:

A1: Choose cooperation ( C) regardless of type.

A2: Choose defection (D) regardless of type.

A3: If the nature's choice is (S)trong civilian support then choose C, otherwise choose D.

A4: If the nature's choice is (S)trong civilian support then choose D, otherwise choose C.

Similarly, the Military has the following four strategies:

B1: Choose cooperation to carry out a coup (E) regardless of type.

B2: Choose not to carry out a coup (E) regardless of type.

B3: If the civilians' choice is (C ) then choose E, otherwise choose E.

B4: If the civilians' choice is (D) then choose E, otherwise choose E.

The complete strategic form filled with the expected payoffs in to the cells of a 4x4 game matrix is given in Table 1 below. To clarify the notation EijCoup (q) refers to the expected payoff for the actor i (either the military (M) or the civilians (C )) of carrying out a coup when the civilians choose j (either to cooperate or to defect) where nature determines the outcome with probability q of success and (1-q) of failure. Here nature's success rate changes again from the case of strong civilian support (q) to weak civilian support (q*). SQnij refers to the payoff of agent i (either the military (M) or the civilians (C )) when the military chooses not to carry out a coup and the civilians choose j (either to cooperate or to defect) under the nature's choice of n (either strong civilian support (S) or weak civilian support (W)).

Table 1: The Strategic Form of the Coup Game when there is asymmetric information advantage on the part of the unified civilians

< td>pEMDCoup(q*)+(1-p)EMDCoup(q) < /tr>
    B1 B2 B3 B4
A1 Military (p)EMCCoup(q*)+(1-p)EMCCoup( q) (p)SQSMC+(1-p)SQWMC (p)SQSMC+(1-p)SQWMC (p)SQS MC+(1-p)SQWMC
  Civilians (p)ECCCoup(q*)+(1-p)ECC Coup(q) (p)SQSCC+(1-p)SQWCC (p)SQSCC+(1-p)SQWCC ( p)SQSCC+(1-p)SQWCC
A2 Military (p)EMDCoup(q*)+(1-p)EMDCoup( q) (p)SQSMD+(1-p)SQWMD pEMDCoup(q*)+(1-p)EMDCoup(q)
  Civilians (p)ECDCoup(q*)+(1-p)ECD Coup(q) (p)SQSCD+(1-p)SQWCD PECDCoup(q*)+(1-p)ECDCoup(q)< /td> PECDCoup(q*)+(1-p)ECDCoup(q)
A3 Military (p)EMCCoup(q*)+(1-p)EMDCoup( q) (p)SQSMC+(1-p)SQWMD (p)SQSMC+(1-p)EMDCoup(q) ( p)SQSMC+(1-p)SQWMD
  Civilians (p)ECCCoup(q*)+(1-p)ECD Coup(q) (p)SQSCC+(1-p)SQWCD (p)SQSCC+(1-p)ECDCoup(q) (p) SQSCC+(1-p) SQWCD
A4 Military (p)EMDCoup(q*)+(1-p)EMCCoup( q) (p) SQSMD+(1-p) SQWMC (p)EMDCoup(q*)+(1-p)SQWMC (p)SQSMD+(1-p)SQWMC
  Civilians (p)ECDCoup(q*)+(1-p)ECC Coup(q) (p)SQSCD+(1-p)SQWCC (p)ECDCoup(q*)+(1-p)SQWCC (p)SQSCD+(1-p)SQWCC

The only Bayesian Nash equilibrium of this game is composed of A1 and B1 pure strategy combination. The critical binding conditions for this equilibrium is given by the following inequalities:

p(E MC Coup (q*) - SQ S MC ) > (1 - p) (SQ W MC - E MC Coup (q))

According to my assumptions both sides of this inequality are negative. In other words, when the nature chooses strong civilian support as C's type, the expected payoff for the military under coup decision following cooperative civilians is less than the status quo payoff for the military when civilians cooperated. This is equivalent to saying that the Military does not intervene against civilians to carry out a coup when it sees them cooperate under strong (S) civilian support. Under weak support for civilians the dominant strategy is to carry out a coup.

p(E CC Coup (q*) - E CD Coup (q*)) > (1 - p)(E CD Coup (q) - E CC Coup (q))

The above inequality is satisfied since the right hand side is positive whereas the right hand side is negative. This means that the expected payoff for the military under strong civilian support when the civilians cooperated is larger than the case where civilians defected. The same condition holds both under S and W.

E CC Coup (q ) > E CD Coup (q).

This inequality represents actually an assumption that will be carried well into my ensuing analysis. All of the above conditions for both the civilians and the military for an equilibrium are thus satisfied. The significance of this Bayesian Nash equilibrium is that the civilians under the threat of a coup pool their strategies under all conditions down to A1 where it chooses to cooperate under all conditions. Accordingly, the civilians tend to choose C regardless of their types and the military always responds with a decision to carry out a coup. Accordingly independent of the type the cooperative move leads to a military that is ready to intervene with a coup. This result is contingent upon the beliefs of the military according to Bayes' rule.

The first condition above implies that (q*) will play a determinant role in the equilibrium outcome. After all, the expected payoff of a coup for the military when the civilians' type is strong (S) or weak (W) can be made small enough by rendering the q as low as necessary. Similar arguments also apply for the case of p.

The assumption above of a single body of civilians interacting with the military is restrictive at best for several reasons. It runs against our intuition that forms the basis of our modeling effort that military regimes come about because of the lack of cooperation among the civilians. In this particular sense the PD game was appealing as a suitable modeling framework. As Cohen (1994) insightfully put the civilians would like to cooperate with each other only if mutually binding commitments to cooperation could be established. Otherwise they both had the dominant preference for defection which leads to mutual lack of cooperation thus the collapse of the system. Our first model shows clearly that either an internal or an external regulatory mechanism for these self-destructive tendencies can be found. However the solutions do not by-pass the central problem of civil societies in the creation of a mutually cooperative environment. Once we have a single coordinated body of civilians the model implicitly assumes the central problem is one of eliminating the threat of the military. By positing a threat that is contingent upon weak civilian structure we gain the advantage of an endogenous player that can not only play the game of a regulator of conflict but also potentially become part of the conflict itself that may self-destruct itself eventually. Figure 5 shows the same game structure with civilians starting to move after observing the nature's selection of their type. However this time the civilians are differentiated into two groups of parties; one left (L) party and a right ( R) party. Both parties know what the nature's selection is but they do not observe each other's moves. The military moves last after observing the civilians' moves but not being able to see what nature has chosen the civilians' types to be. The extensive for of this game is given below but the solution of this new version will not be sought.

Lastly, the coup game is transformed into one where the military instead of the civilians is the party with private information that can be used to its advantage. In Figure 6 the nature selects the country's security situation. The military has private information about whether or not country faces serious bad conditions into the future or it has good security conditions. Going back to the discussion of the first model above I again assume that on both sides of the ideological spectrum there are extremists who tend to use violent means like urban terrorism or rural armed conflict not only against each other but also against the regime forces. It is not uncommon to find these groups engaged in international trafficking of drugs and arms. Similarly some of these groups may be in touch with foreign powers that may support them in their territorial conflicts with the existing regime. All of these connections typically create an informational asymmetry in favor of the military who may or may not use them honestly and in cooperation with the civilian holders of power in elected governments. Civilian governments in coup contexts I discussed above for the first model are typically short-lived. Moreover they tend to be coalition governments due to fractionalized electoral support. This means that parties with backing of extremist parties of various ideological orientations may come to power. Such a development tend to push the military to see themselves as the protectors of the regime and they may choose to conceal information from the civilians for the sole purpose of pushing them to a more favorable condition from the perspective of the military increasing their popularity. 6 It is commonly known that bad conditions are likely with probability p whereas the good conditions' probability of occurrence is (1-p). After observing the nature's choice the military decides whether to reveal the true state of affairs or simply to lie about it.

Since the above game is also a typical game of incomplete information the Bayesian Nash equilibria needs to be computed. There are in total 64 possible combinations of pure strategies. These pure strategies for the civilians L and R can be summarized as follows:

L1: If the Military declares good conditions then choose C L . Otherwise choose to defect (D L ).

L2: If the Military declares bad conditions then choose C L . Otherwise choose to defect (D L ).

L3: Choose cooperation (C L ) regardless of type.

L4: Choose cooperation (D L ) regardless of type.

R1: If the Military declares good conditions then choose C R . Otherwise choose to defect (D R ).

R2: If the Military declares bad conditions then choose C R . Otherwise choose to defect (D R ).

RL3: Choose cooperation (C R ) regardless of type.

R4: Choose cooperation (D R ) regardless of type.

Similarly the Military's strategies are given by:

M1: If the Nature selects good conditions then choose G M . If the Military's declaration is bad conditions then choose B M . (Tell the truth).

M2: If the Nature selects bad conditions then choose G M . If the Military's declaration is good conditions then choose B M . (Lie about the nature's selection).

M3: Choose cooperation (G M ) regardless of type.

M4: Choose cooperation (B M ) regardless of type.

The complete strategic form filled with the expected payoffs in to the cells of a 4x4x4 game matrix is given in Table 2 below. To clarify the notation E(Mi)j refers to the expected payoff for the actor j (either the military (M), the left (L) or right ( R) party) at the terminal information node i (i=1,..16) where nature determines the outcome with probability q of success and (1-q) of failure after the military chooses to carry out a coup or to stay out. Here nature's success rate changes again from the case of strong civilian support (q) to weak civilian support (q*). ERk(i,j) refers to the payoff of agent k (either the leftwing party (L) or the rightwing party (R )) where the actions of I (of L) and j (of R) that can take the form of C or D. OP(i,j) k refers to the oppression by the military in case of a successful coup to agent k (either the leftwing party (L) or the rightwing party (R )) following the actions of i (of L) and j (of R) that can take the form of C or D. Mm(i,j) refers to the payoff of the military when it takes action m (either to stay out (E) or to carry out a coup and succeed (S) or to fail (F)) following the actions of i (of L) and j (of R) that can take the form of C or D.

Similar to the discussion on the magnitudes of payoff for the first complete information model above I assume the following relative magnitudes of the payoffs:

ER L(D,C) > ER L(C,C) > ER L(D,D) > ER L(C,D)
ER R(C,D) > ER R(C,C) > ER R(D,D) > ER R(D,C)

which are the typical PD game payoff structure.

OP (D,D) L = OP (D,D) R > OP (C,D) R = OP (D,C) L > OP (D,C) R = OP (C,D) L > OP (C,C) R = OP (C,C) L > 0
M E(C,C) > M S(D,D) > M S(C,D) = M S(D,C) > M S(C,C) > M E(C,D) = M E(D,C) > M E(D,D) > 0
0 > M F(C,D) > M F(D,C) > M F(D,D) > M F(C,C)

The relative magnitudes of all of these payoffs are assumed to stay the same under both the good and the bad security conditions that the nature determines but the military's payoffs under bad conditions are larger than they are under the good conditions. Similarly, the civilian's payoffs are larger under the good conditions than they are under the bad one.

Table 2: The Strategic Form of the Coup Game when there is asymmetric information advantage on the part of the Military facing non-unified civilians

L1 R4 pE(M3)R+(1-p)E(M16)R
L   L1 L1 L1
R   R1 R2 R3
  Military pE(M4)M+(1-p)E(M15)M pE(M3)M+(1-p)E(M16)M pE(M3)M+(1-p)E(M15)M pE(M4)M+(1 -p)E(M16)M
M11 Left pE(M4)L+(1-p)E(M15)L pE (M3)L+(1-p)E(M16)L pE(M3)L+(1-p)E(M15)L pE(M4)L+(1-p)E(M1 6)L
  Right pE(M4)R+(1-p)E(M15)R pE(M3)R+(1-p)E(M15)R pE(M4)R+(1-p) E(M16)R
  Military pE(M9)M+(1-p)E(M8)M pE(M10)M+(1-p)E(M7)M pE(M9)M+(1-p)E(M7)M pE(M10)M+(1- p)E(M8)M
M12 Left pE(M9)L+(1-p)E(M8)L pE( M10)L+(1-p)E(M7)L pE(M10)L+(1-p)E(M7)L pE(M10)L+(1-p)E(M8 )L
  Right pE(M9)R+(1-p)E(M8)R pE(M10)R+(1-p)E(M7)R pE(M9)R+(1-p)E(M7)R pE(M10)R+(1-p)E (M8)R
  Military E(M9)M+E(M13)M E (M14)M+E(M10)M E(M9)M+E(M13)M E(M14)M+E(M10)M
M13 Left E(M9)L+E(M13)L E(M14)L+ E(M10)L E(M9)L+E(M13)L E(M14)L+E(M10)L
  Right E(M9)R+E(M13)R E(M1 4)R+E(M10)R E(M9)R+E(M13)R E(M14)R+E(M10)R
  Military E(M4)M+E(M8)M E( M3)M+E(M7)M E(M3)M+E(M7)M E(M4)M+E(M8)M
M14 Left E(M4)L+E(M8)L E(M3)L+E( M7)L E(M3)L+E(M7)L E(M4)L+E(M8)L
  Right E(M4)R+E(M8)R E(M3) R+E(M7)R E(M3)R+E(M7)R E(M4)R+E(M8)R

L2 R4 pE(M1)R+(1-p)E(M16)R pE(M12)R+(1-p)E(M5)R
L   L2 L2 L2
R   R1 R2 R3
  Military pE(M2)M+(1-p)E(M15)M pE(M1)M+(1-p)E(M16)M pE(M1)M+(1-p)E(M15)M pE(M2)M+(1 -p)E(M16)M
M11 Left pE(M2)L+(1-p)E(M15)L pE (M1)L+(1-p)E(M16)L pE(M1)L+(1-p)E(M15)L pE(M2)L+(1-p)E(M1 6)L
  Right pE(M2)R+(1-p)E(M15)R pE(M1)R+(1-p)E(M15)R pE(M2)R+(1-p) E(M16)R
  Military pE(M11)M+(1-p)E(M6)M pE(M12)M+(1-p)E(M5)M pE(M11)M+(1-p)E(M5)M pE(M12)M+( 1-p)E(M6)M
M12 Left pE(M11)L+(1-p)E(M6)L pE (M12)L+(1-p)E(M5)L pE(M11)L+(1-p)E(M5)L pE(M12)L+(1-p)E(M 6)L
  Right pE(M11)R+(1-p)E(M6)R pE(M11)R+(1-p)E(M5)R pE(M12)R+(1-p )E(M6)R
  Military E(M15)M+E(M11)M E(M16)M+E(M12)M E(M15)M+E(M11)M E(M16)M+E(M12)M
M13 Left E(M15)L+E(M11)L E(M16)L +E(M12)L E(M15)L+E(M11)L E(M16)L+E(M12)L
  Right E(M15)R+E(M11)R E(M 16)R+E(M12)R E(M15)R+E(M11)R E(M16)R+E(M12)R
  Military E(M2)M+E(M6)M E( M1)M+E(M5)M E(M1)M+E(M5)M E(M2)M+E(M6)M
M14 Left E(M2)L+E(M6)L E(M1)L+E( M5)L E(M1)L+E(M5)L E(M2)L+E(M6)L
  Right E(M2)R+E(M6)R E(M1) R+E(M5)R E(M1)R+E(M5)R E(M2)R+E(M6)R

L3 R4 pE(M1)R+(1-p)E(M14)R
L   L3 L3 L3
R   R1 R2 R3
  Military pE(M2)M+(1-p)E(M13)M pE(M1)M+(1-p)E(M14)M pE(M1)M+(1-p)E(M13)M pE(M2)M+(1 -p)E(M14)M
M11 Left pE(M2)L+(1-p)E(M13)L pE (M1)L+(1-p)E(M14)L pE(M1)L+(1-p)E(M13)L pE(M2)L+(1-p)E(M1 4)L
  Right pE(M2)R+(1-p)E(M13)R pE(M1)R+(1-p)E(M13)R pE(M2)R+(1-p) E(M14)R
  Military pE(M9)M+(1-p)E(M6)M pE(M10)M+(1-p)E(M5)M pE(M9)M+(1-p)E(M5)M pE(M10)M+(1- p)E(M6)M
M12 Left pE(M9)L+(1-p)E(M6)L pE( M10)L+(1-p)E(M5)L pE(M9)L+(1-p)E(M5)L pE(M10)L+(1-p)E(M6) L
  Right pE(M9)R+(1-p)E(M6)R pE(M10)R+(1-p)E(M5)R pE(M9)R+(1-p)E(M5)R pE(M10)R+(1-p)E (M6)R
  Military E(M9)M+E(M13)M E (M14)M+E(M10)M E(M9)M+E(M13)M E(M14)M+E(M10)M
M13 Left E(M9)L+E(M13)L E(M14)L+ E(M10)L E(M9)L+E(M13)L E(M14)L+E(M10)L
  Right E(M9)R+E(M13)R E(M1 4)R+E(M10)R E(M9)R+E(M13)R E(M14)R+E(M10)R
  Military E(M2)M+E(M6)M E( M1)M+E(M5)M E(M1)M+E(M5)M E(M2)M+E(M6)M
M14 Left E(M2)L+E(M6)L E(M1)L+E( M5)L E(M1)L+E(M5)L E(M2)L+E(M6)L
  Right E(M2)R+E(M6)R E(M1) R+E(M5)R E(M1)R+E(M5)R E(M2)R+E(M6)R

L4 R4 pE(M3)R+(1-p)E(M16)R pE(M12)R+(1-p)E(M7)R
L   L4 L4 L4
R   R1 R2 R3
  Military pE(M4)M+(1-p)E(M15)M pE(M3)M+(1-p)E(M16)M pE(M3)M+(1-p)E(M15)M pE(M4)M+(1 -p)E(M16)M
M11 Left pE(M4)L+(1-p)E(M15)L pE (M3)L+(1-p)E(M16)L pE(M3)L+(1-p)E(M15)L pE(M4)L+(1-p)E(M1 6)L
  Right pE(M4)R+(1-p)E(M15)R pE(M3)R+(1-p)E(M15)R pE(M4)R+(1-p) E(M16)R
  Military pE(M11)M+(1-p)E(M8)M pE(M12)M+(1-p)E(M7)M pE(M11)M+(1-p)E(M7)M pE(M12)M+( 1-p)E(M8)M
M12 Left pE(M11)L+(1-p)E(M8)L pE (M12)L+(1-p)E(M7)L pE(M11)L+(1-p)E(M7)L pE(M12)L+(1-p)E(M 8)L
  Right pE(M11)R+(1-p)E(M8)R pE(M11)R+(1-p)E(M7)R pE(M12)R+(1-p )E(M8)R
  Military E(M15)M+E(M11)M E(M16)M+E(M12)M E(M15)M+E(M11)M E(M16)M+E(M12)M
M13 Left E(M15)L+E(M11)L E(M16)L +E(M12)L E(M15)L+E(M11)L E(M16)L+E(M12)L
  Right E(M15)R+E(M11)R E(M 16)R+E(M12)R E(M15)R+E(M11)R E(M16)R+E(M12)R
  Military E(M4)M+E(M8)M E( M3)M+E(M7)M E(M3)M+E(M7)M E(M4)M+E(M8)M
M14 Left E(M4)L+E(M8)L E(M3)L+E( M7)L E(M3)L+E(M7)L E(M4)L+E(M8)L
  Right E(M4)R+E(M8)R E(M3) R+E(M7)R E(M3)R+E(M7)R E(M4)R+E(M8)R

Instead of giving the full equilibrium analysis of the 4x4x4 payoff matrix I checked whether some interesting strategy combinations appear to be Bayesian Nash equilibria. For example the separating strategy for the military of telling the truth (M1) combined with separating strategies of the civilians of cooperating under military's declaration of good conditions and defecting under the bad (L1 and R1) is not an equilibrium. Similarly (M1, L2, R2), (M3, L3, R3), (M2, L3, R3) and (M1, L3, R3) also do not appear to form an equilibrium under the above assumptions. In other words, military's option of telling the truth about the security situation of the country does not appear to be in equilibrium under sensible reaction combinations of the civilians like always cooperating with each other (L3, R3). However similarly lying strategy for the military also does not appear in equilibrium. An interesting equilibrium however arises when the military always declares good conditions and as a response the civilians always defect (M3, L4, R4). This equilibrium means that the military always declares good conditions. Such a situation may arise due to a legitimacy concern of the military. If the conditions are always good then the military does not need to defend its performance to the public. "All is under control" is the signal that the military is trying to give here. However, the civilians can not break the PD context's dominant strategy of defection under all conditions.

CONCLUSION AND PROSPECTS FOR FUTURE RESEARCH

The occurrence of a military coup is a result of a complex interaction between the structural characteristics of a country and strategic decisions by the principal actors in the political system. By explicit modeling of the actors' decisions based on their evaluations of structural conditions and the effect of other players' possible strategies, one is able to diagnose the conditions under which a coup is to be undertaken. Following Cohen's (1994) pioneering work on the modeling of coups I formulated a two-stage game where the effects of the military action are explicitly taken into account. The results of the analysis reveal two important dynamics that have important policy implications. First, I find that coups are intrinsically related to the degree of military oppression upon the political parties under military regimes. Cooperative results can be imposed by the military by threatening to punish the defectors at a sufficiently high level. Accordingly, the role of the military is not seen here as a simple threat to democratic rule but also as an active participant in the regulation of political conflict that may lead to the breakdown of democracy. It may be suggested that such a benevolent military is hard to find. However, the costs of military regimes are not trivial upon the military itself. The popularity of the military under mutual cooperation and no intervention can actually be greater than the payoff of intervention under unilateral or mutual defection. If this is the case although the military prefers to enter under all conditions other than the mutual cooperation, the first choice for the military is not to intervene if it can persuade the civilians to cooperate. Accordingly the military prefers first to regulate political conflict by simply threatening to intervene and punish the defector(s) at a sufficiently high level. The option of intervention is preferred if this threat does not work and unilateral or mutual defection occurs. 7

Second, above discussion clarifies conditions under which electoral rewards for different strategy combinations lead to cooperative outcomes. Although these electoral conditions are partially determined by structural distribution of electoral support across different parties in the political system, an important determinant can also be found in different election systems that allow electoral fractionalization and ideological polarization to increase. Election systems can be adjusted to allow net electoral gains to support cooperative conditions diagnosed above. Similarly, rewards for coalition building as well as control and minimization of losses due to unilateral or mutual cooperation will increase the likelihood of cooperative outcomes. The above discussion illustrates the usefulness of explicit formulation of electoral incentives in the explanation of coups. It helps us diagnose the underlying electoral forces behind democratic collapse and thus provides insights as to how these structural bottlenecks can be avoided.

When information asymmetries are introduced the results are far from being encouraging for a stable democratic rule in immature democratic environments with PD context dominating the civilians competition with each other and the military as a significant player. First, I find that when a unified civilian body plays against the military the only equilibrium outcome is one where the civilians always cooperate with the policy demands of the military but the military never chooses to stay out. The decision of the military to carry out a coup critically depends on the nature's determination of the rate of success for a coup. If the probability of success is high enough, even under the case when the civilians carry out the demands of the military by cooperating with them, then the coup occurs under those cases where cooperative concessions are already given to the military. This is clearly an empirically testable hypothesis that needs to be checked with the historical evidence. More specifically the model leads us to conclude that the civilian side will cooperate with the military independent of their electoral strength and will face military intervention under both strong and weak electoral support for the civilian rule. All of these of course, are assumed to take place under PD context defined by the payoff structure.

When information asymmetry is in favor of the military facing a divided civilian body under PD constraints the outcome is again not very encouraging. It seems that the military has no equilibrium strategy that pushes them to tell the truth about the insecurity threats that faces the country and that the civilians can not break the dominant strategy constraint of defection thus leading to a breakdown of the system. Accordingly when the payoff structure is given by the PD context the breakdown of the democratic system seems inescapable.

Remedying several aspects of the phenomena that are consciously ignored in its present form can further develop the above modeling framework. First and foremost among these ignored aspects is the exchange between the moderates and the extremists of their own party or the extremists of their own ideological camp. This exchange takes place under perfect recognition of the fact that the result of this particular game will lead to an outcome that will shape the incentives for the moderates to play with each other in the above depicted cooperation-defection game. What makes the extremists push the moderates towards cooperation rather than defection? If this question can be satisfactorily answered then a complete diagnosis of party system attributes leading to cooperation can also be identified.

Another line of research can be further developed to account for the uncertainty the civilians face when competing with each other. Figure 5 comes closest to defining such a game structure. The civilians can be depicted as divided into left and right ideological camps and each may hold private information as to their electoral strength. When the civilians move under this mutual uncertainty they know that they will face the decision of the military which may have perfect information about one side (like the right wing ideological camp) and may be uncertain about the other side and it will have to make a decision to intervene or to stay out. Such a modeling framework opens a new line of thinking by simply allowing the military to take sides or simply coalesce with one camp against the other. Moreover, as such we will be in a position to model the coup game as a repeated game. The military in Turkey for example has intervened three times in the last nearly four decades. In the last two the civilian leadership was the same. Due to institutional rigidity the military leadership can be assumed to have remained more or less constant, at least in terms of their preferences. I short the real world question that is of interest to us is one of a repeated game with players of long term interests with possibility of building coalitions not only among the civilians but also between the civilians and the military.

Perhaps the most important feature of these models is however their use in designing new institutional structures, new incentive schemes to achieve desirable outcomes. In this particular case the desirable objective is a stable democracy. In this sense such a modeling approach anchors into one of the oldest and most debated literatures in political science and thus offers a new approach to an old question. All of these however remain to be dealt with in another paper.

References

Notes:

Note 1: See Oye, 1986; Grieco, 1990; Haas, 1990, Putnam and Bayne, 1987; Milner, 1992, 1997 for a similar conception of cooperation. Back.

Note 2: Prisoner's dilemma is one of the most extensively studied game theoretical concepts. A good selection of the main characteristics of the literature can be found in Olson, 1965; Mueller, 1989; Taylor, 1987, Geddes, 1994; Axelrod, 1984; Hardin, 1968; Rappaport and Chammah, 1965; Argyle, 1991. Back.

Note 3: The model developed below relies heavily on Cohen (1994) for the description of the economic and political environment within which it is supposed to be played as well as key behavioral assumptions about the players. Back.

Note 4: The two parties will be treated as Cohen's (1994) ideological moderates. However no specific ideological commitment is necessary for the actual functioning of the model since their limitations imposed by the extremists will be explicitly taken into account by the way their electoral pay-offs are shaped. Back.

Note 5: This asymmetric punishment by the military however, should not be taken as siding with one party against the other. For such biased intervention one would have to assume that the oppression for one defecting party is higher than the oppression when the other party defects. OpDLD,C > OpDRC,D would then mean that L is getting punished more than R when it unilaterally defects. For simplicity, I assume that both parties will get oppressed the same way as the other defects. This assumption also means that under similar conditions the oppression required for the military to obtain the results it wants by applying oppression does not change. Back.

Note 6: See Carkoglu (1996) for a lengthy discussion of the military's incentives to conceal information in the specific case of Turkish coups of 1960, 1971 and 1980. Back.

Note 7: See Ziegenhagen (1986) for a thorough discussion of issues in the regulation of political conflict. Back.