2.1 Peace through military victory

Most people may have a negative image of military victory, such as higher causalities or winner-takes-all. Therefore, ‘peace through military victory’ is difficult to understand intuitively. Why does military victory tend to achieve durable peace?

Wagner (Reference Wagner and Roy1993) compares the characteristics of military victory and peace settlement and offers a convincing argument for the relationship between civil war outcomes and post-civil war peace. According to Wagner (Reference Wagner and Roy1993: 255), negotiated peace settlement is more likely to result in a balance of power relationship between former combatants and in unstable peace because the former combatants can continue to retain the military capability required to resume civil war in the future. On the other hand, military victory destroys the organizational capability and identity of the loser, and allows only the winner to hold military weapons and thus it is possible to prevent the loser from re-igniting civil war. In sum, he argues that military victory is more likely to lead to stable peace than is negotiated peace settlement because of capability asymmetry between the winner and the loser.

Wagner successfully conducted the pioneering work on the relationship between military victory and the risk of conflict recurrence. This proposition is called the ‘the Wagner hypothesis’ (Licklider, Reference Licklider1995: 685). Luttwak (Reference Luttwak1999) takes a similar stand, contending that the international community should not mediate combatants to end by negotiated peace settlement, but rather should ‘give war a chance’ (Luttwak Reference Luttwak1999: 44).

Besides Wagner's argument, military victory can ease the information problem after conflict. When civil war ends in negotiated peace settlement, warring parties not only preserve their own military forces but also have an incentive to disguise information about their military capacities and political will in the peace process (Walter, Reference Walter2002). Such incomplete information about military strength and political will can be a critical cause for the re-occurrence of conflict (Fearon, Reference Fearon1995). In the case of military victory, the winning side can display its relative strength and political will to the losing side and potential rebel groups (Walter, Reference Walter2004: 374). Therefore decisive military victory can prevent them from (re)starting civil war.

Wagner's research lumps government military victory and rebel military victory together as military victory and does not investigate the differences between the two. After military victory, the winning side can control the bureaucracy, national economy, and territory. Whereas a government can continue to use its regime after government military victory, a rebel group has to first overthrow the former government regime before establishing its new regime to run the country. Since a rebel group tends not to have much experience in state management, the new regime the rebel group establishes after military victory may not be as strong as that of a government regime. Civil war is less likely in states with enough capacity (Fearon and Laitin, Reference Fearon and Laitin2003). States with enough capacity can resolve grievances and prevent insurgencies by using competent bureaucracy and deploying a well-organized army. In other words, the lack of adequate capacity is positively related to the likelihood of conflict recurrence. Therefore, military victory by a rebel group leads to less stable peace than does military victory by a government.

In spite of its importance in academic and policy fields, Wagner's hypothesis has not been directly studied, with a few exceptions. Licklider (Reference Licklider1995), using a simple cross tabulation analysis, shows empirical evidence supporting Wagner's hypothesis that military victory is likely to lead to more durable peace than is negotiated peace settlement. Toft (Reference Toft2010) also examines the relationship between civil war termination and conflict recurrence. She finds that military victory, especially rebel military victory, is more likely to lead to durable peace in post-conflict period, whereas negotiated settlement and cease-fire increase the risk of civil war recurrence. Although the researches cited above are the few empirical studies which deal with civil war outcomes as main independent variables, they do not deal with duration analysis and use time-varying covariates. The shortcomings of the analytical method render the reliability of empirical results doubtful.

DeRouen and Berocovitch (Reference DeRouen and Bercovitch2008), using duration analysis for civil wars between only enduring internal rivalries, partially confirm the relevance of Wagner's hypothesis. Mukherjee (Reference Mukherjee2006) demonstrates that military victory does not always have a statistically significant effect on the longevity of peace. He examines all civil war cases and shows that military victory decreases the risk of civil war recurrence only when there is a political power-sharing agreement.

On the other hand, several studies find weak or no connection between military victory and durable peace. Quinn et al. (Reference Quinn, Mason and Gurses2007), analyzing the effect of rebel military victory and negotiated peace agreement on the risk of civil war recurrence, show mixed results that civil war, which ends in a rebel victory or a negotiated peace agreement supported by international peacekeeping operations, is less likely to recur than is the civil war that ends in a government victory. Doyle and Sambanis (Reference Doyle and Sambanis2000) demonstrate through cross-sectional analysis that there is no statistically significant relationship between military victory and stable peace. Walter (Reference Walter2004) also argues that military victory and other characteristics of previous civil war (e.g., cost of war) do not substantially impact the risk of recurrence. In her research, the characteristics of countries (e.g., living conditions) have a significant effect on the likelihood of renewed war.

Disagreement among existing empirical studies on Wagner's hypothesis and the lack of comparative studies suggest the need for further investigation.

2.2 Negotiated peace settlement, commitment problem, and durable peace

Although some studies demonstrate that negotiated peace settlement is more likely to lead to less durable peace than does military victory in a post-civil war country, not all negotiated peace settlements necessarily lead to unstable peace after civil war. In fact, 57% of the peace settlements between 1940 and 1992 were successfully implemented and contributed to durable peace in post-civil war societies (Walter, Reference Walter2002: 6).

According to the peace settlement school, the lack of credible commitment of advantaged groups on implementation increases the likelihood of peace settlement failure and civil war recurrence.Footnote ³ This hypothesis suggests that the failure of peace settlement implementation results from the combatants’ inability to provide a credible commitment not to renege on possible compromise agreements because of two structural features: quasi anarchy and the shadow of power transition. Quasi anarchy means that no central government exists to ensure implementation of order or law (Walter, Reference Walter1997: 338−40). Next, peace agreement requires a rebel group to disarm and demobilize and the creation of a new national army, and such a requirement makes the rebel group more vulnerable to surprise military attacks and coercion than before. In this situation, a government cannot offer the rebel group any credible commitment to implement the peace agreement. This encourages a rebel group to renege on the settlement and return to civil war. Even though both sides would prefer peaceful settlement to the cost of war, they cannot avoid the failure of the peace settlement if a credible commitment is lacking.Footnote ⁴

The main reasons for the failure of peace settlement implementation, and civil war recurrence, are the fear of military attack and excessive concessions, and the mistrust of the adversary's commitment to the implementation of peace settlement. Several scholars suggest that third-party interventions, usually in the form of peacekeeping operations, and the introduction of power-sharing pacts can guarantee combatants’ security and remove these obstacles to successful implementation (Hartzell, Reference Hartzell1999; Hartzell and Hoddie, Reference Hartzell and Hoddie2003; Hartzell et al., Reference Hartzell, Hoddie and Rothchild2001; Walter, Reference Walter1997, Reference Walter2002).

Most civil war researchers advocate that peace, in general, lasts longer when third parties intervene in post-civil war countries than when they do not intervene (e.g., Doyle and Sambanis, Reference Doyle and Sambanis2000; Fortna, Reference Fortna2004). Third parties can monitor and verify the compliance and implementation of a peace agreement better than the former warring parties, and thus provide reliable information to eliminate distrust between the combatants (Walter, Reference Walter2002). The information that is collected by third parties can play an important role at various stages of peace process − bargaining stage, negotiation stage, and implementation stage. In addition, third-party monitoring activity can reduce the incentive to cheat and renege on peace agreements (Werner, Reference Werner1999: 916–917). Peacekeeping operations increase the cost of surprise attack and promote disarmament and demobilization.

The introduction of a power-sharing mechanism has also been identified as one of the most effective ways to establish sustainable peace (e.g., Hartzell, Reference Hartzell1999; Hartzell and Hoddie, Reference Hartzell and Hoddie2003). The main reason for implementation failure is the fear of surprise attack and distrust in commitment on implementation. A power-sharing arrangement is designed to give both a government and a rebel group sufficient opportunity and ability to prevent the opponent from dominating politics, from launching a surprise attack, from occupying a territory, and from monopolizing welfare and resources. Thus, a power-sharing arrangement can ease the hostility between former combatants and lead to more stable peace in post-civil war society. Power-sharing between social groups is one of the most fundamental means of sustaining peace and resolving disputes without violence (Lijphart, Reference Lijphart1977).

Some empirical studies successfully show the importance of third-party interventions and power-sharing arrangements. From a quantitative analysis of all peace settlement cases, Walter (Reference Walter2002) finds that third-party security guarantees or power-sharing pacts significantly increase the likelihood of successful implementation of peace agreements. Hartzell et al. (Reference Hartzell, Hoddie and Rothchild2001) also show, with empirical evidence, that third-party interventions reduce the risk of peace agreement failures and civil war recurrence. Hartzell and Hoddie (Reference Hartzell and Hoddie2003) divide power-sharing pacts into political, military, territorial, and economic power-sharing pacts and investigate their effects on peace duration after a peace settlement using a duration model. According to their findings, the addition of one more category of power-sharing terms reduces the risk of peace settlement failure by 53%.

Although the result of these studies indicates that third-party interventions and power-sharing pacts increase the likelihood of successful implementation of a peace settlement and lengthen the duration of post-civil war peace, they do not compare negotiated peace settlement cases with other post-civil war cases and discuss how these two important supportive factors influence the post-conflict peace after military victory.

2.3 Which is the best outcome for post-conflict peace?

Wagner (Reference Wagner and Roy1993) argues that military victory is more likely to lead to durable peace than peace settlement, and Licklider (Reference Licklider1995) and other scholars find empirical evidence in support of this hypothesis. The existing research on peace settlement contends that a peace agreement with third-party security guarantees or power-sharing pacts increases the likelihood of successful implementation compared to an agreement without such supports. Neither the research on military victory nor the research on peace settlement makes comparisons with the case that is considered a more durable peace by another school. Quinn et al. (Reference Quinn, Mason and Gurses2007) simultaneously investigate the effects of rebel victory and of peace settlement with peacekeeping operations on the risk of war recurrence. Mukherjee (Reference Mukherjee2006) includes a one-side military victory and military victory with a political power-sharing agreement in the same regression model. However, they do not explain which is the best route for post-conflict peace?

In civil war research, the motivation and feasibility of armed groups are considered crucial for the onset of civil war (Collier and Hoefller, Reference Collier and Hoeffler2004; Fearon and Laitin, Reference Fearon and Laitin2003). The motivation for rebellion derives from people's grievance against a government, which arises from political oppression, relative deprivation, social exclusion, ethnic and religious fractionalization, and inequality, or people's greed for accumulating private wealth. Sharing the common motivation for insurgency enables a potential rebel group to overcome the collective action problem during costly war. As rebel leaders have to organize their own groups and procure arms and belligerents to wage civil war, the feasibility of civil war is determined by the availability of financial and military resources. The logic of civil war onset is partially applicable to civil war recurrence.

Military victory can destroy the loser's organized capability, such as their military arsenals and bases, and political and military leaders can be arrested or executed. Mobilizing human and material resources to restart armed conflict, namely the feasibility of renewing insurgency, becomes more difficult for the loser than ever before. However, as much as decisive victory deprives the losers of their political, economic, and social opportunities, it aggravates their grievances and increases the motivation of potential belligerents for insurgency.

In contrast to military victory, negotiated peace settlement allows the warring parties to preserve the feasibility of their armed operations in the new integrated national army. If the peace settlement is supported by third-party peace operations or power-sharing pacts, these supports attenuate the fear and mistrust of belligerents, and provide former combatants the opportunities to achieve their political, economic, and social goals through non-violent means. In such an environment, while the motivation to rebel is likely to decrease, the feasibility of military operation still remains. In the case of a truce, the warring parties not only preserve their own armed forces, but also any disagreements are not resolved. Thus, the motivation for insurgency and feasibility of belligerents continues after truce.

In most cases, power-sharing arrangements are achieved when civil wars end in negotiated peace settlements. However, sometimes, combatants agree to power-sharing arrangements after military victory (Mukherjee, Reference Mukherjee2006), and third parties intervene in a post-conflict country to maintain internal order, even if civil war ends with a victory. Third-party interventions and power-sharing arrangements are not always a subset of negotiated peace settlements.

As mentioned above, military victory can reduce the feasibility of rebellion, but it cannot eliminate the source of motivation for rebellion. However, if third-party peacekeeping operations or power-sharing arrangements support a post-conflict country which ends in a military victory, its civil war outcome can decrease the loser's motivation for, as well as feasibility of, rebellion. In other words, military victory, supported by peacekeeping operations or power-sharing arrangements, can lead to longer peace spells than does negotiated peace settlement, supported by peacekeeping operations or power-sharing arrangements, and military victory without support.

From the above argument, it is anticipated that in the case of military victory with support, the motivation for, and the feasibility of, insurgency may be alleviated, and the risk of recurrence is lower than in other civil war outcomes. Previous works have not considered the case of military victory with support or conducted a comprehensive comparative study of different civil war outcomes and the durations of peace spells.Footnote ⁵

3.1 Data

In order to test the relationship between civil war outcomes and the durations of post-civil war peace, I use Fortna's (Reference Fortna2004) dataset to identify a sample of 114 civil war episodes for the period 1944–1999.Footnote ⁶ The definition of civil war followed by Fortna's (Reference Fortna2004) dataset is the same as the one given by Doyle and Sambanis (Reference Doyle and Sambanis2000: 783): a civil war ‘causes more than 1,000 battle deaths overall and in at least a single year; challenges the sovereignty of an internationally recognized state; occurs within the recognized boundaries of the state; includes rebels with the ability to mount organized armed opposition to the state; and has parties concerned with the prospect of living together in the same political unit after the end of the civil war’.

The dataset is preferable to others for the following two reasons. First, Hartzell and Hoddie (Reference Hartzell and Hoddie2003), Hartzell et al. (Reference Hartzell, Hoddie and Rothchild2001), and Walter (Reference Walter2002) focus only on negotiated peace settlement cases and exclude other civil war outcomes (i.e., military victory by one side and truce). For this research, the different termination types, government military victory, rebel military victory, negotiated peace settlement, and truce, are incorporated into the dataset to compare the effects of different civil war outcomes. Second, most researchers on civil war recurrence use time-constant covariates (e.g., Hartzell et al., Reference Hartzell, Hoddie and Rothchild2001). In this paper, since the characteristics of countries and third-party interventions vary with time after civil war, several time-varying covariates, such as peacekeeping operation, level of democracy, and economic condition, are incorporated into the dataset.

3.2 Dependent variable

The dependent variable is operationalized as the duration of peace after the end of civil war. Many former studies consider that a spell of peace is successful if the civil war does not resume during a certain period following the termination of civil war (e.g., Doyle and Sambanis, Reference Doyle and Sambanis2000; Licklider, Reference Licklider1995; Walter, Reference Walter1997, Reference Walter2004).Footnote ⁷ They operationalize the length of post-civil war peace as a dichotomous variable (either two or five years), but the arbitrary time period may be problematic. Therefore, the entire duration of peace spell is used here. Fortna's dataset includes detailed information about the cease-fire dates and the recurrence dates of civil wars. Such information enables one to count the exact duration of peace spells.Footnote ⁸ Unlike the existing studies, in which the duration of peace is measured by the number of months or years (e.g., Hartzell and Hoddie, Reference Hartzell and Hoddie2003; Hartzell et al., Reference Hartzell, Hoddie and Rothchild2001; Mukherjee, Reference Mukherjee2006), for this research it is measured by the number of days that peace endured following termination of civil war, through to 31 December 1999. A spell of peace is therefore the time between the termination of fighting (by a decisive military victory, negotiated peace settlement, or truce after a military stalemate) and the recurrence of civil war, if any, between the same protagonists.Footnote ⁹ When a civil war does not recur, the duration of peace is right-censored on 31 December 1999. With this operationalization of dependent variable, the 114 post-civil war peace events are at risk of failure for a total of 581,648 days.

3.3 Independent variables

Civil war outcome. The main independent variables, in this research, are civil war outcomes that are based on the Fortna's (Reference Fortna2004) dataset.Footnote ¹⁰ The variable Military Victory was measured here by a dummy variable coded 1 if civil war ended in a military victory of a government or a rebel group, 0 if otherwise. More detailed military victory variables, Government Victory and Rebel Victory, are also created for this study. Government Victory is a war termination in which a government decisively defeats a rebel group in the battlefield. Rebel Victory is a war termination in which a rebel group militarily defeats a government's forces in the battlefield, overthrows the incumbent regime, and establishes a new regime. Moreover, it includes secessionist civil wars in which a rebel group secedes from the government and builds a new state or autonomous region in the territory controlled by the defeated government. Peace Settlement is a formal peace treaty, in which a government and a rebel group agree to stop fighting and promise to resolve the issue of conflict through mutual cooperation in a new political order with a formal peace agreement. Therefore, an informal settlement or a cease-fire is not included in Peace Settlement. In Truce, there is no formal settlement between a government and a rebel group, but there is an informal settlement or a cease-fire which stalemates any further large-scale fighting. All variables are dummy variables.

Third-party peacekeeping. In this analysis, a third-party peacekeeping operation in a post-civil war country is used as a third-party intervention. To identify third-party peacekeeping operations, the websites of several International/Regional Organizations (UNDPKO, OSCE, and ECOWAS) are used, besides the Doyle and Sambanis’ (Reference Doyle and Sambanis2000) dataset, Fortna's (Reference Fortna2004) dataset, and UCDP database. Peacekeeping Operation is coded 1 for each peace year when there is at least one third-party peacekeeping operation (UN or non-UN) in post-civil war countries and 0 if otherwise. In addition, interaction terms between third-party peacekeeping operation and military victory and negotiated peace settlement (Peacekeeping Operation × Military Victory and Peacekeeping Operation × Peace Settlement) are calculated to estimate the effect of military victory and negotiated peace settlement with third-party peacekeeping operations on the post-civil war peace.

Power-sharing arrangement. Power-Sharing Arrangement is obtained from Hartzell and Hoddie (Reference Hartzell and Hoddie2003) and Walter (Reference Walter2002). Hartzell and Hoddie define power-sharing as ‘those rules that, in addition to defining how decisions will be made by groups within the polity, allocate decision-making rights, including access to state resources, among collectives compositing for power’ (Harzell and Hoddie Reference Hartzell and Hoddie2003: 320). Three types of power-sharing arrangement are identified here: political power-sharing, military power-sharing, and territorial power-sharing. Power-Sharing Arrangement is coded 1 if a government and a rebel group agree with more than one power-sharing pact, 0 if otherwise. Interaction terms are created between power-sharing arrangement and military victory and negotiated peace settlements (Power-Sharing Arrangement × Military Victory and Power-Sharing Arrangement × Peace Settlement) to unravel the effects of military victory and negotiated peace settlement with power-sharing arrangements on peace spells.

3.4 Control variables

Once civil war terminates and a peace spell begins, several factors may influence whether the peace will last long or end soon. To identify these factors and to understand their influence on the duration of peace spell, first three characters of a previous civil war are taken into account: cost of civil war, duration of civil war, and war type. Next, it is expected that some factors, which are thought to be the causes of civil war onset, may also become the causes of civil war recurrence. Therefore, several characteristics of countries, such as the level of democracy, economic conditions, oil export, and ethnic heterogeneity are included in the analysis models.

Cost of civil war. To estimate the impact of civil war cost on the duration of peace spell, the number of causalities, that is the total number of the dead (both civilian and battle deaths) during the civil war, is used in this analysis (Cost of Civil War). The variable is obtained from Doyle and Sambanis (Reference Doyle and Sambanis2006) and is measured using the natural log of the number of the dead people to ease the problem of outliers and skewness. Intuitively, high war costs may have a positive impact on the duration of peace spells. However, several scholars find that the higher the civil war cost, the greater is the risk of civil war recurrence (Doyle and Sambanis, Reference Doyle and Sambanis2000: 787; Fortna, Reference Fortna2004: 287; Hartzell et al., Reference Hartzell, Hoddie and Rothchild2001: 198). High casualty rates may create high levels of hostility between the combatants. High levels of hostility can cause more hatred and distrust between combatants and people, and thereby render conflict resolution and peace maintenance very hard. The increase in the cost of civil war is expected to increase the probability of civil war recurrence.

Duration of civil war. The duration of previous civil war also has an impact on the risk of civil war recurrence (Duration of Civil War). The duration of war varies from 1 month (Uganda 1966–1966 and Yemen–South/Peoples’ Rep. 1986–1987) to 600 months (Israel–Palestine 1947–1997). This variable is also obtained from Fortna (Reference Fortna2004) and measured using the natural log of monthly duration. According to previous research, a long duration of civil war offers combatants a chance to learn the opposing group's military ability, interests, and political will, and thus diminishes the former combatants’ incentive to restart civil war (Hartzell et al., Reference Hartzell, Hoddie and Rothchild2001: 198). The duration of civil war is therefore negatively related to the likelihood of civil war recurrence.

Identity war. Identity War is a variable that distinguishes ethnic, religious, and identity conflicts from ideological, revolutionary, and other wars. Identity War is coded 1 if it is an ethnic, religious, or identity conflict, 0 if otherwise. The variable is obtained from Fortna (Reference Fortna2004). The stakes in identity conflicts, involving ethnic, religious, racial, and linguistic issues, are less divisible than those in politico-economic conflicts, concerning ideological and socioeconomic interests (Hartzell et al., Reference Hartzell, Hoddie and Rothchild2001: 189). Identity conflicts are assumed to involve higher stakes, and thus compromise and accommodation are much harder to achieve than they are with politico-economic conflicts. The post-civil war peace is, therefore, expected to be more difficult to sustain after an identity civil war.

Level of democracy. Polity IV democracy–autocracy scale (–10 to 10) is used to measure the level of democracy after a peace spell starts (Democracy). Democratic countries are less likely to experience a civil war because the democratic institution and process give people an opportunity and a tool to express their grievance without violence. Therefore, the level of democracy after civil war is expected to lead to longer durations of peace spells following civil war termination.

Economic condition. To find the relationship between the duration of peace spells after civil war and economic conditions, GDP per capita, from Fearon and Laitin's (Reference Fearon and Laitin2003) dataset, is included in this empirical analysis (GDP per capita). Fearon and Laitin (Reference Fearon and Laitin2003) and others find that economic conditions are strongly and negatively correlated to susceptibility to civil war. In addition, low economic and living conditions mean low opportunity cost to becoming involved in civil war (Collier and Hoeffler, Reference Collier and Hoeffler2004). GDP per capita is therefore expected to have a negative impact on the risk of civil war recurrence.

Oil export. Natural resource dependence is one of the most important indicators for the risk of armed conflict. Especially, oil dependence is closely related to weaker stake capacity and civil war onset (Fearon and Laitin, Reference Fearon and Laitin2003). Oil Export is coded 1 when states receive more than one-third of their export revenues from oil exports, 0 if otherwise. This variable is obtained from Fearon and Laitin (Reference Fearon and Laitin2003). Several scholars suggest that oil dependence is positively related to the risk of civil war onset (de Soysa, Reference de Soysa2002; Fearon and Laitin, Reference Fearon and Laitin2003). A similar relationship is expected to hold for civil war recurrence. Thus, oil dependence is expected to increase the probability of civil war recurrence.

Ethnic heterogeneity. Vanhanen's (Reference Vanhanen1999) dataset of ethnic heterogeneity is included in the empirical model to test the impact of racial, religious, and linguistic divisions on the chances for recurrence of civil war (Ethnic Heterogeneity). This index ranges from 0 (minimum heterogeneity) to 144 (maximum heterogeneity). Greater ethnic heterogeneity may increase the opportunity for mutual contact among different ethnic groups. More contacts can exacerbate ethnic divisions and lead to enhancement of hostilities after civil war. Therefore, ethnic heterogeneity is expected to have a positive impact on the recurrence of civil war.