Sequencing the Peace

Abstract

Once a set of civil war actors reach a final peace agreement, a number of different implementation sequences are possible as the negotiated provisions are put into practice. We focus on a key but threatening stepping stone in the post-accord period—the holding of the first post-accord election—which has the capacity to be a stabilizing or destabilizing force. We identify effective accommodation provisions that civil war actors can negotiate and implement before the first post-accord election to reduce the chances of renewed violence. Utilizing new longitudinal data on the implementation of comprehensive peace agreements between 1989 and 2012 and a series of survival models, we find that if the first post-accord election is preceded by the implementation of accommodation measures, elections can have a peace-promoting effect. However, in the absence of preelection accommodation measures, elections are much more likely to be followed by peace failure.

Keywords

civil war peace agreement implementation accommodation demobilization bargaining sequencing recurrence

When civil war adversaries negotiate and implement a peace agreement, the first post-accord election has the potential to function as a watershed event that can consolidate the peace process or trigger renewed violence. Elections have become the primary mechanism since the Cold War for regulating political contestation among conflict parties emerging from civil war through a negotiated settlement.¹ Recent empirical studies emphasizing institutional development have shown that early elections are more destabilizing than later elections in the aftermath of civil conflict (Brancati and Snyder 2013; Flores and Nooruddin 2012). These findings also suggest that electoral violence—like civil war violence more generally—is rooted in credible commitment problems and information failures which are not resolved quickly, but are gradually overcome through participation in confidence building measures. In this article, we investigate how civil war actors can reduce the inherent dangers in holding the first post-accord election by sequencing the implementation of accommodation measures before elections.

We treat post-conflict environments as characterized by a condition of “multiple sovereignty,” that is, the existence of competing groups who wish to rule the state or part of it, each backed by a military wing, and each advancing an agenda which is rejected by the other side (Tilly 1978). We refer to these three core aspects as contestation to rule, military rivalry, and incompatibility. These represent the main challenges, we argue, that have to be dissolved in war to peace transitions and they can be tackled in different order or sequence. There are three common mechanisms for unraveling this triad of peacebuilding challenges, respectively, elections regulate contestation to rule, demobilization manages the military rivalry, and accommodation lessens the incompatibility of interests. Elections and demobilization are potentially disempowering, each carrying a risk that the outcome could be very different, and possibly more damaging, than what is anticipated. Their disempowering nature is often an obstacle to their successful implementation.

In the current article, we develop a sequencing argument that proposes ways in which former adversaries can first sequence the implementation of accommodation measures to reduce the dangers associated with the holding of the first post-accord election. We empirically test our argument utilizing new data on the implementation of comprehensive peace agreements (CPAs) and their major provisions negotiated between 1989 and 2012. We find that if elections are first preceded by the implementation of accommodation measures, elections can serve their democratic purpose without increasing the risk of renewed violence and war. Inversely, elections held without preceding accommodation are significantly more likely to be followed by renewed armed conflict.

This article contributes to several literatures. First, the article contributes new knowledge to the large body of literature on civil war recurrence and peacebuilding. Second, the accommodation provisions we identify represent tangible policies that civil war actors can negotiate to produce more peaceful elections and more durable peace agreements. As such, the article also informs the mediation and negotiation literatures. Third, the sequencing argument and findings contribute to the vast body of work on democratic transitions and democracy survival. Fourth, the article represents a follow-up on several recent works on issues of timing and sequencing associated with a country’s first election out of civil war.

In the next section, we review prior research on the holding of a country’s first election after a civil war. Applying a bargaining perspective to the problem of first elections, we develop a sequencing argument on peace agreement implementation and derive testable hypotheses from it. In the Research Design section, we specify the dependent and independent variables, data sources, and methods. The findings and analysis section presents empirical results from a series of survival analyses and two illustrative case examples from Tajikistan and Angola. Finally, we conclude by discussing implications of the findings.

Prior Research on Sequencing First Post-conflict Elections

The level of interest in topics such as the proper preconditions, sequencing, and timing of elections following civil war violence has drastically increased since the 1990s. It was during this decade that peacebuilding efforts reached a tipping point in which civil wars were ending at a faster rate than new conflicts were beginning for the first time in at least seventy-five years. At the center of most of these peacebuilding efforts were elections. Manning (2004, 60) notes that “[s]ince 1990 elections have been a key part of virtually all negotiated agreements to end civil wars.” Descriptive statistics in Flores and Nooruddin (2012, 560) show that the number of elections held the first year following a civil conflict in the 1990s was double that of the 1980s and quintuple that of the 1970s. Writing about the peacemaking experiences in the early 1990s, López-Pintor (1997) coined the term “reconciliation elections” to describe the new emphases being put on elections in civil war peace processes at the time. Building on López-Pintor’s work, Pagani (1998, 224) describes how “the electoral process” in the aftermath of civil war was perceived as containing “all the expectations of the peace process” and constituting “its political and psychological culmination point.” In particular, “election day” is seen as a key visible event in the peace process and to the extent that the elections are perceived to be free and fair, and the outcome is accepted, the resulting government attains a greater legitimacy than was previously possible.

Pagani notes several factors that have complicated first elections: the former warring parties often lack democratic and electoral experience and have little understanding of what scholars refer to as democratic political culture. Paraphrasing Clausewitz, Pagani (1998, 227) holds that in a recent war context, actors will likely view “the political confrontation and the electoral competition” as “the continuation of the war by other means.” He notes that governments, despite having agreed to elections in the peace agreement, often show a strong political and psychological reluctance in even discussing the details of the administration of the election with its former enemy, since such issues normally fall under the prerogatives of a sovereign government. Meanwhile, the threat of “returning to the bush” is a key political card played by rebel groups leading up to elections (pp. 229-230). For the former warring parties, the elections are the moment of truth in the sense that it will be decided who will rule the country, and often the whole political struggle that motivated the violent conflict is perceived to be at stake. To complicate matters further, the electoral outcome tends to be more difficult to foresee (at least the victory margins) than in elections in other contexts since the parties have little or no previous experience with electoral competition and the polls are less scientific and less reliable than in developed nations (p. 234).

In response to the high levels of expectation surrounding first elections in conflict settings and recognizing the magnitude of complications, advocates of early elections argue that, even if held under imperfect conditions, it is preferable to start the democratic process earlier rather than later. Having leaders in place with greater legitimacy assists in the process of cultivating greater democratic norms in these difficult contexts (see Kumar 1998; Lindberg 2003; Diamond 2006). Several early successes in the 1990s were seen as confirming the arguments of early election proponents. However, a series of first election failures led to the development of a scholarly camp opposed to the idea.

Drawing from Huntington’s (1968) thesis that the political development of a society—if it is to be mostly peaceful—should follow the right sequence, skeptics of early elections argued for a wider acknowledgment and appreciation of the role of “preconditions.” While there was general agreement that it was “dangerous to push states to democratize before the necessary preconditions are in place” (Mansfield and Snyder 2007, 5; 2005), there was less consensus on what those preconditions were and how to go about getting them in place. Paris (1997, 2004) argued against countries attempting to liberalize their political systems or economies until the right mix of institutions were established. Barnett (2006, 89) argued that elections should wait until the state has the capacity “to channel, organize, and respond” to the rising tide of popular demands. Quite often the preconditions put forth were themselves large-scale complex processes, presumably, with their own preconditions (e.g., “civil society,” “law and order,” “stability,” “capacity”).

For civil wars ending in negotiated peace agreements where the implementation process can be designed and planned, three of the more widely discussed preconditions that have been put forth concern interim governments, demobilization, and third-party guarantees. Lyons (2002, 216) argues that negotiators and conflict actors “can increase the chances that elections will both end a war and begin a transition toward democracy” by establishing “interim institutions” that “create confidence that the vote will be managed effectively” as well as “programs to transform militias into political parties.” Stedman identifies “demobilization” as “the single most important sub-goal” within a series of implementation processes geared toward “the transformation of warring armies into political parties” (2002, 668). In Walter (1997), we also see an emphasis on demobilization and third-party enforcement as well as power sharing (1999, 141) in the context of transitions involving elections.

Brancati and Snyder (2013) and Flores and Nooruddin (2012) represent the leading empirical works on the topic of post-conflict election timing and preconditions and test a number of these prior claims. Brancati and Snyder (2013) examine a sample of 136 time periods following civil war and the amount of time until the first post-conflict election. They find strong supportive evidence that postponing the first post-conflict election significantly lengthens the duration of peace. They also found that initiating demobilization before the first election contributes to longer peace spells. They conclude that first post-conflict elections can be peaceful when “institutionalized administrative and legal safeguards reduce the risk of the elections being riddled with fraud and elected officials arbitrarily exploiting power once in office” (p. 844). Flores and Nooruddin (2012) find large and significant conflict-inducing effects for elections held in the first year for “new” democracies but not in countries with prior exposure to democratic institutions. They conclude, “unless elections are delayed at least two years in new democracies and one year in more established democracies, renewed violence is likely” (p. 566).

Although the findings of Brancati and Snyder (2013) and Flores and Nooruddin (2012) provide strong and generalizable cross-national evidence that election timing and institutional maturity are important dynamics, their results do not translate easily, in our view, into specific policy recommendations for civil war actors engaged in negotiations toward a peace agreement. First, both studies argue in favor of postponing elections, but as demonstrated by Radin (2013), Carothers (2007), Reilly (2002), and others, this is often impractical as the conflict actors and public usually demand early elections. Second, both studies assume that, even if the conflict actors could be persuaded to postpone the first election, institutional development will occur in the meantime. Third, it is not clear from either study precisely what the exact institutional developments or preconditions are—only that they take time to develop. In sum, the recent literature, in our view, has focused more on timing, that is, early versus late elections, rather than on sequencing, that is, the best order of implementation regarding elections. In this article, our goal is to inform peace agreement design and agreement implementation by identifying a small set of accommodation measures that civil war actors can negotiate and implement immediately to reduce the destabilizing effects of elections. We next develop a theoretical argument based on bargaining and commitment theories of civil war violence that provide a rationale and framework for selecting a set of accommodation provisions for this task.

A Bargaining Perspective on the Sequencing of Implementation

Following previous scholarship that analyzes civil war violence using a bargaining framework, we explicate two main drivers of violent confrontation surrounding elections, namely, the lack of information about one’s adversaries’ true capabilities and preferences (information failure) and the difficulty of convincing one’s adversaries that you will not exploit a future position of strength, if attained (the commitment problem, or the time inconsistency problem; Fearon 1995; Powell 1996; Walter 1997; Wagner 2000; Filson and Werner 2002; Slantchev 2003). In the context of elections, violence often results from uncertainties regarding how one’s opponent will rule if elected and the inability of each side to convince the other that they will not exploit them if given the chance. Walter (1999, 140) thoroughly covers the fears that must be overcome for elections to be peace inducing:

It is easy to imagine why parties would willingly submit to elections in situations where they might benefit from winning in the future, where they will not be maltreated if they lose, and where it would be costly to subvert the system. Under these conditions, they have every incentive to cooperate. But what if elections and institutions could promise none of these things? What if the losers of the first postwar election could not count on another opportunity to regain power? What if no limits were set on what the winning party could do once elected? Most important, what if a loss in the first election could be permanent? These are the conditions that characterize countries emerging from civil wars.

We concur with previous studies that strong third-party guarantees in the form of peacekeeping operations and monitoring/verification missions (Walter 1997; Doyle and Sambanis 2000, 2006; Fortna 2003, 2004; Mattes and Savun 2010; Joshi 2013) can help parties mitigate security dilemmas—especially surrounding demobilization. However, we see elections as being associated with different kinds of fears than demobilization, and it is highly unlikely that a third party would be willing to step in and stop one side in an election from cheating, or stop the winning side from passing discriminatory policies, or establishing one-party rule after some time in office. In the rare instance that a third party would be willing to intervene in the domestic politics of a nation in defense of one group over another group, the probable reason for that resolve is that the third party has great stakes in the outcome and favors one of the parties. Hence, it is likely that the resolved third-party enforcer is going to be perceived as self-interested, biased, and bullying, rather than a neutral guarantor who can help both sides overcome the commitment problems that are threatening the electoral process (Roeder and Rothchild 2005). Thus, problems of credible commitment posed by elections between former enemies require different and tailored solutions.

The decisive factor in our view regarding whether elections are stabilizing or destabilizing, is the acceptance or rejection of the electoral outcome. The key determinant of whether all significant parties will accept the outcome is the level of trust that has been established between the parties, and the level of confidence that has been established in the country’s electoral institutions to deliver a fair result. If the losing side concedes, accepts that it has lost, and remains committed to the democratic process, this moment represents a major step forward in the peace process (López-Pintor 1997; Pagani 1998; Höglund 2009; Höglund, Jarstad, and Kovacs 2009; Collier 2009). Conversely, if the losing side refuses to accept the outcome, there is a serious risk that there will be a return to war. The path back to war may result from the losing side taking up arms again after the election, however, because of first-strike advantages, either side may attack the other at any time in the electoral process if they are uncertain how the other side may react in the future.

We add to existing theoretical arguments about the conditions for successful peacebuilding by considering the role of sequencing the implementation of costly accommodation provisions before the electoral process begins. We argue that holding elections, which are potentially disempowering, without first building confidence through the implementation of accommodation measures raises fears of being cheated or abused. On the other hand, we argue that if accommodation provisions are implemented first, then elections can reduce the risk of renewed war by regulating the underlying contestation over who should rule.

Accommodation assumes the central role in our argument. Accommodation can include the implementation of any measure that reduces the incompatibility of the former warring parties’ preferred policies and signals a strong willingness to shrink the bargaining space by moving toward one’s opponents’ ideal policy position. When one side implements a policy that they formerly considered incompatible with their preferred outcome, the remaining incompatibility shrinks, and the benefitting counterpart will be more satisfied with the new status quo. This lessening of the incompatibility is a key part of accommodation. Being on the receiving end of accommodation legitimizes your standing, increases your stake, and gives you more confidence that the remaining incompatibility can be bridged. Accommodation also functions as costly signaling, in line with the arguments by Kydd (2000), Hartzell and Hoddie (2003), and Jarstad and Nilsson (2008), where adversaries can exchange signals designed to persuade the other side that they are trustworthy by virtue of the fact that the signals are costly in a way that one would hesitate to send them if untrustworthy. A failure to implement accommodation measures could signal a large remaining incompatibility and intransigence. Here we go beyond previous theorizing by noting the critical importance of sequencing accommodation before engaging in implementation processes that are disempowering for one side-such as elections. In the next section, we discuss accommodation in more detail and put forth five criteria in an effort to identify the best possible set of preelection accommodation measures that conflict actors can negotiate and implement after the signing of a peace agreement.

What Is Preelection Accommodation?

We argue that accommodation should be sequenced as the first step in a post-accord implementation process. The way we understand accommodation draws on bargaining theory and costly signaling theory as applied to peace processes (Kydd 2000, 2005). Although we have outlined the broad purpose of accommodation and sketched a definition, we need to put forth specific criteria for identifying effective preelection accommodation provisions. We propose that the kind of preelection accommodation most likely to moderate the dangers of elections must meet five criteria. The implementation of effective preelection accommodation measures must be swift, verifiable, costly, facilitative, and non-disempowering.

First, preelection accommodation implementation must be swift in order to be in place well before the first election is held. Ideally, the implementation of preelection accommodation needs to begin almost immediately after the accord is signed and be implemented to a large degree within six to twelve months. The second criteria is that preelection accommodation must be costly and this logic is derived from signaling theory (Fehr and Gächter 2002; Gintis, Smith, and Bowles 2001; Kydd 2000, 2005).² Since talk is cheap, only behavior that is costly enough to deter opportunists or free riders will send the right message. The third criterion that effective preelection accommodation must be verifiable stems also from signaling theory. In order to be an effective signal of honest intention, the costly activity has to be “hard to fake.” Parties cannot update their beliefs about each other’s peaceful intent from behaviors that reportedly happened but are difficult to verify. Our fourth criterion is that preelection accommodation needs to be directly facilitative to the holding of elections in a way that reduces fears and suspicions about the process by directly strengthening the electoral process. This criterion is also related to signaling theory which recognizes that not any costly signal will do. To be effective, the costly signal needs to be tied directly to the sender’s participation in the very activity where they stand to benefit by cheating (i.e., the electoral process). For our purposes, facilitative means that the accommodative act should increase the preparedness and integrity of the electoral process and increase the viability of opposition parties. This criterion separates relevant preelection accommodation measures from the many other conceivable ways of signaling peaceful intent that have little or nothing to do with strengthening elections. Lastly, our fifth criterion is that preelection accommodation must be non-disempowering. It would be self-defeating to use disempowering provisions as accommodation measures because they themselves would require preceding acts of accommodation in order to persuade or elicit actor participation in their implementation.

We assessed each of the fifty-one provision types in the Peace Accords Matrix Implementation Data Set (Joshi, Quinn, and Regan 2015) according to our five criteria. While numerous provisions meet some of the five criteria, only three provisions, in our view, were strong candidates across all five fields, namely, transitional power sharing government, amnesty, and prisoners’ release. Before discussing the three that we selected in greater detail, we briefly discuss a short list of prominent provisions from the peacebuilding literature that are unsuitable preelection accommodation measures (military power sharing, territorial power sharing, peacekeeping operation, and verification/monitoring). Military power sharing and territorial power sharing are strongly correlated with a durable peace (Hoddie and Hartzell 2003; DeRouen, Lea, and Wallensteen 2009, 370) but cannot be swiftly implemented (Jarstad and Nilsson 2008, 218). Verification and monitoring provisions (Mattes and Savun 2010) are critically important provisions for overcoming security-related commitment problems, however, they can be established at relatively low cost and their effectiveness (which matters most) is difficult to observe. The provision that comes closest to meeting all five criteria in a strong way is a peacekeeping operation which has been tested in previous studies of first elections following civil war (Brancati and Snyder 2013). For these reasons, we include peacekeeping as a control in our upcoming analysis. Now we turn to the three types of accommodation provisions that we argue best meet the five criteria of swift, verifiable, costly, facilitative, and non-disempowering implementation.

Based on our five criteria, we argue that the establishment of a transitional power sharing government is a strong candidate for a preelection accommodation measure. Our definition shares the major elements in Walter (2002) and Mattes and Savun (2009), a guaranteed share of cabinet positions or quotas in one of the three main branches of government. The implementation of a transitional power sharing government meets the first criteria of swiftness, that is, officially appointing the nominees to their positions can be established in a matter of weeks or months.³ Despite its swiftness, it is politically and psychologically costly to implement. In a civil war context where one or both sides have made past claims that they represent the only legitimate power who should rule unconstrained, sharing power with one’s enemy signals a strong commitment to peace precisely because it is so distasteful. Hoddie and Hartzell (2003), Jarstad and Nilsson (2008), and DeRouen, Lea, and Wallensteen (2009) applied costly signaling theory to the implementation of power sharing arrangements.⁴ Third, the coexistence of former antagonists under the same institutional roof is highly verifiable, and the participants themselves are the monitors of implementation. Fourth, the implementation of a transitional power sharing government is strongly facilitative to the practicalities of holding competent elections. One of the primary tasks of a power sharing transitional government is to prepare the nation for free and fair elections and this mandate is often highlighted in the peace agreement itself as a principal function.

In Cambodia’s Framework for a Comprehensive Political Settlement, for example, a twelve-member national council was established with the authority to “ensure a neutral political environment conducive to free and fair general elections.” Nepal’s November 2006 CPA established an “Interim Government,” consisting of the eight major political parties in the country. Only one function was outlined, that is, “to have the elections … held by the Interim Government in a free and fair manner.” While the interim or transitional governments serve a variety of functions, they are designed to be temporary and their chief task is to implement their own withdrawal and replacement in an orderly manner. As Lyons (2002, 222) argues, “Interim governments are by their nature not legitimated by democratic processes but derive their authority from the extent to which they prepare the country for meaningful elections and turn power over to the winners.” Fifth, the implementation of transitional power sharing—while costly—is not disempowering for either actor. Arguably, such arrangements “keep your enemies closer” and thus increase levels of transparency, accountability, and compliance with the peace agreement, making each actor more secure.

Next, we argue that granting amnesty, which exempts former combatants from liability for criminal or political offenses committed during the conflict, signals a willingness to accommodate the other and live together without vengeance and retribution. Writing and passing a general amnesty bill can be accomplished in a matter of months, and thus it meets the criteria of swift implementation. Implementing amnesty is also quite costly to politicians because significant segments of the population are often firmly opposed to it. The strength of amnesty as a costly act of accommodation lies in the fact that most everyone knows that the authorities would have preferred not to do it (see Snyder and Vinjamuri 2003; Darby and Mac Ginty 2008). The granting of amnesty is also highly verifiable in that it requires the drafting of legislation and the successful passage of that legislation. Conversely, if former rebel combatants are being pursued in military offenses or arrested at their homes, the lack of amnesty is clearly observable. The implementation of amnesty is also highly facilitative to the electoral process, particularly, in the area of increasing the viability of the emerging oppositional parties. Without amnesty, potentially thousands of rebel combatants and supporters are likely to choose to remain in hiding or in the field, depressing turnout to the polls. The reluctance to leave the field and participate in politics for fear of being prosecuted or repressed makes it unlikely that the rebel group will view itself as a viable political party and disillusionment with the process will be the likely result. Under these conditions, an electoral loss will be seen by many as the result of unfair treatment rather than the true choice of the electorate. Lastly, amnesty is not disempowering, that is, it does not increase the vulnerability of the participants. Quite the opposite, arguments could be made that amnesty leaves both sides more secure by eliminating uncertainty regarding criminal liability and the possibility of being prosecuted or targeted.

Turning to our third candidate for effective preelection accommodation, we propose that conflict actors can implement the release of prisoners of war (POW) in order to signal a clear move toward the demilitarization of politics and normalized group relations. First, the release of war prisoners can be swiftly implemented. Even with a legal process of review in place, prisoners can be released gradually as their cases are approved. Second, releasing war prisoners is costly both politically and militarily. In addition to the psychological costs associated with the release of those considered by many to be killers, terrorists, or criminals, the implementation of this particular provision signals a strong commitment to move away from military rivalry. This is mainly due to the fact that the government is releasing combatants that they will have to fight against if the war is renewed. Third, releasing war prisoners from prison is a public, observable, and highly verifiable process. Prisoners are either going home or remaining in prison. Fourth, the release of POW is facilitative to the holding of elections. A political party is not likely to see itself as viable or perceived by the public as viable when many of its leaders and supporters are being held in prison. Finally, the release of war prisoners clearly meets the criteria of being non-disempowering in that releasing war prisoners does not decrease either side’s defensive capabilities. Based on the preceding arguments, we propose the following hypothesis:

Hypothesis 1: Holding elections without preceding accommodation implementation increases the risk of war recurrence, but holding elections with preceding accommodation implementation reduces the risk of war recurrence.

Research Design

Dependent Variable

We now turn to the testing of our sequencing hypothesis. The dependent variable to be used is the recurrence of armed conflict between the signatories of CPAs. By utilizing the dyadic version of the Uppsala Conflict Data Program (UCDP)/PRIO Armed Conflict Dataset (Gleditsch et al. 2002; Themnér and Wallensteen 2013), we identify whether the signatories to the accord resumed violence resulting in at least twenty-five battle deaths (coded 1 or 0). Once the recurrence of conflict is coded “1,” the data are right censored. Our unit of analysis is post-accord peace survival year and our dependent variable begins in the calendar year following the year the accord was signed so that every accord is allowed to survive one year.

Independent Variables

This study uses the Peace Accords Matrix Implementation Data Set (PAM_ID; Joshi, Quinn, and Regan 2015). This data set provides annual data on the implementation of fifty-one different types of provisions in CPAs signed between 1989 and 2012.⁵ Each CPA was read and coded according to the number and type of provisions contained in the document. In all, 724 provisions were identified across thirty-four CPAs (averaging 21 provisions in each accord). Research was undertaken on each provision, in each accord, and annual chronological narratives were written covering the major implementation events taking place each year. Content analysis was then performed on the annual narratives so that each provision received an ordinal implementation score, annually. Each year (for up to ten years) was coded according to the level of implementation achieved: never initiated (0), initiated and minimum (1), intermediate (2), or full (3).

Drawing from our theory, we operationalize accommodation based upon the implementation of three provisions, namely, the establishment of a transitional power sharing government, the granting of amnesty, and the release of war prisoners. Out of thirty-four CPAs examined, seven agreements contained all three provisions, thirteen contained two accommodation provisions, ten contained one provision, and four had no accommodation provisions at all. We constructed an accommodation implementation index that ranges from 0 to 100 and captures the degree to which the accommodation provisions that were agreed upon by the parties were implemented on an annual basis. To derive the index, the sum of the current level of implementation (each year) is divided by the highest possible level of implementation (full) for all commitments and normalized by multiplying by 100. As an example, consider an accord that contains all three accommodation provisions. If the implementation of all three provisions is considered minimal in year one, the accommodation index score is $[(1 + 1 + 1) / (3 \times 3)] \times 100 = 33 p e r c e n t$ . In year two, if the implementation of two provisions is intermediate and one still minimal, the accommodation score is 5/9 or 55 percent. If by the third year, two are fully implemented and one still minimal, the score is 7/9 or 77 percent. If all three are fully implemented in year four, the score is 9/9 or 100 percent. We then created the binary variable high accommodation which takes the value of “1” in every year that accommodation implementation is equal to or greater than two-thirds on the accommodation implementation index. High accommodation tests for the possibility that high levels of implemented accommodation measures have a direct, unmediated effect on war recurrence that begins to exert influence as soon as implementation begins (without any consideration of post-accord elections).

Our election-related variables are operationalized using data from the International Institute for Democracy and Electoral Assistance (IDEA 2013).⁶ The variable first post-accord election marks the year in which the first post-accord election was held (1,0) and tests for a direct, unmediated effect of elections on war recurrence (without any consideration of accommodation). To test our hypothesis on the effect of holding elections after high levels of implemented accommodation, we created the interaction (product) term High Accommodation × First Post-accord Election. This variable tests for a relationship between first post-accord election preceding high accommodation and peace survival time. This variable takes the value of “1” if high levels of implemented accommodation are in place as of the country’s first post-accord election. Some descriptive statistics are presented in Table 1.

Table 1.

Descriptive Statistics.

Variable	Observations	Mean	Standard deviation	Minimum	Maximum
High accommodation (0,1)	314	0.710	0.4544	0	1
First post-accord election (0,1)	314	0.098	0.2988	0	1
High Accommodation × First Post-accord Election (0,1)	314	0.531	0.4998	0	1
Armed conflict recurrence (0,1)	314	0.057	0.2328	0	1
Number of accommodation provisions (0 to 3)	314	1.640	0.9498	0	3
High demobilization (0,1)	314	0.328	0.4702	0	1
Conflict deaths (in 100,000)	314	1.907	4.7678	0	23
Demobilization implementation (0 to 3)	314	1.436	1.2904	0	3
War duration (months)	314	135.91	132.68	6	433
Executive constraint (5 to 7)	314	0.544	0.4988	0	1
Peacekeeping (0, 1)	314	0.334	0.4725	0	1
Prior CPA (0, 1)	314	0.095	0.2944	0	1
Infant mortality rate (in 1,000)	307	66.002	39.312	4.9	148.11
GDP per capita PPP international (in 1,000)	304	3.577	5.5001	0.28	35.87
Armed personnel (in 1,000)	300	162.77	390.90	2	2,400
Conflict type (0,1)	314	0.458	0.4991	0	1
Net foreign aid (in tens of millions)	297	56.037	53.883	3.2	260

Note: CPA = Comprehensive Peace Agreements; GDP = gross domestic product; PPP = purchasing power parity.

Control Variables

We include several control variables that may be systematically related to our independent variable or the recurrence of conflict. The variable number of accommodation provisions counts the number of accommodation provisions in the accord (ranging from 0 to 3). This allows us to distinguish between the impact of implementing accommodation provisions versus just having them in the agreement. Following Brancati and Snyder (2013), we control for demobilization. The variable demobilization implementation measures the range of implementation and takes a value from 0 to 3 each year. The dummy variable high demobilization takes the value “1” if demobilization was fully implemented by the end of the year and otherwise “0.” The costs of the preceding war should make peace accord signatories more wary of potentially disempowering elections and more reluctant to demobilize their combatants. On the other hand, higher costs could produce a weariness effect. For these reasons, we control for previous battle deaths and conflict duration of the civil war that produced the CPA using data from the UCDP (2013). We control for democracy using a binary indicator of high levels of executive constraint (levels 5–7) from the polity project (Marshall, Gurr, and Jaggers 2013). Following the peacebuilding literature, we control for peacekeeping operations in the implementation environment (Walter 1997; Fortna 2004; Doyle and Sambanis 2006; Quinn, Mason, and Gurses 2007; Joshi 2013). This variable is coded “1” when peacekeeping troops are deployed and “0” otherwise and taken from PAM_ID. We also control for prior CPA using the same data. This variable is coded “1” for a prior CPA and “0” otherwise. Following previous studies on state capacity and peace agreement implementation (DeRouen et al. 2010), we use infant mortality rate, gross domestic product (GDP) per capita (purchasing power parity international), and the number of armed state personnel to capture the influence of state capacity on implementation. Data for these variables come from the World Bank (2013). From the same data source, we also control for foreign aid since attracting foreign aid might create incentives that lead to higher levels of implementation. Our last control is a dummy variable indicating territorial conflicts from UCDP (2013).

Method

We use event history analysis to examine peace duration until a recurrence of conflict takes place. As Box-Steffensmeier and Jones (2004) suggest, the event history method is methodologically preferable when the dependent variable is the spell before an event. Within event history analysis, there are options for nonparametric, semi-parametric, and parametric models. The first two models compare subjects when failure occurs (Cleves, Gould, and Gutierrez 2004). Parametric models estimate the probability of what “occurs over the whole interval given what is known about the subject during this time” (Cleves, Gould, and Gutierrez 2004, 199). Because we expect the probability of armed conflict recurrence to decline over time, a Weibull model is most appropriate. The Akaike information criterion test statistics rule out the use of a Cox proportional hazard model. Among parametric models, tests statistics are similar. As the estimated parameter (p) for Weibull models are above 1 for every model, we use the Weibull model. Because we are interested in modeling time to event, we estimate using the accelerated failure time (AFT) metric (see Cleves, Gould, and Gutierrez 2004).

Findings and Analysis

Table 2 presents a series of Weibull regressions that examine peace survival time until a recurrence of conflict (failure event) takes place. Before we examine our main hypothesis regarding the effect of sequencing accommodation implementation before the holding of elections, we examine in model 1 the direct unmediated effects of implementing accommodation provisions. In model 2, we examine the direct unmediated effects of holding the first post-accord election. In model 1, the variable high accommodation is highly significant and positive, which indicates that implementing accommodation provisions at the start of a peace process has strong pacifying effects. In model 2, the variable first post-accord election is highly significant and negative, which indicates that the holding of the first post-accord election is indeed associated with a much higher subsequent risk of returning to war.

Table 2.

Sequencing High Accommodation before the First Post-accord Election and Peace Survival.

	(1)	(2)	(3)	(4)	(5)
High accommodation (0,1)	1.930**** (0.586)	1.254** (0.502)	0.904** (0.374)	0.837** (0.369)	0.864* (0.516)
First post-accord election (0,1)	—	−0.944**** (0.270)	−1.037** (0.408)	−1.018*** (0.389)	−0.894** (0.430)
High Accommodation × First Post-accord Election	—	—	1.187*** (0.423)	1.197*** (0.448)	1.270*** (0.455)
Deaths (in 100,000)	0.129* (0.067)	−0.058 (0.061)	−0.066 (0.043)	−0.054 (0.050)	−0.070 (0.047)
War duration (months)	−0.003 (0.002)	−0.001 (0.002)	0.001 (0.003)	−0.001 (0.003)	0.001 (0.003)
Executive constraint (5 to 7)	1.383*** (0.535)	0.568** (0.250)	0.745** (0.346)	0.634** (0.297)	0.489* (0.267)
Peacekeeping (0,1)	0.041 (0.334)	0.130 (0.570)	0.303 (0.322)	0.044 (0.381)	0.145 (0.408)
Conflict type (0,1)	0.693 (0.862)	0.890** (0.408)	1.785**** (0.392)	1.527*** (0.552)	2.078**** (0.498)
High demobilization (0,1)	—	1.557*** (0.499)	1.607*** (0.534)	1.548*** (0.544)	—
Demobilization implementation (0 to 3)	—	—	—	—	0.454** (0.177)
Number of accommodation provisions (0 to 3)	−0.191 (0.437)	—	—	—	—
Infant mortality rate (in 1,000)	—	−0.011*** (0.004)	—	—	—
GDP per capita PPP international (in 1,000)	0.370*** (0.142)	—	—	—	—
Armed personnel (in 1,000)	—	0.002 (0.003)	—	0.004 (0.004)	—
Prior CPA (0,1)	−0.975 (1.083)	—	—	—	—
Net foreign aid (in tens of millions)	—	0.004 (0.005)	—	—	—
Constant	0.400 (1.301)	1.357** (0.677)	0.144 (0.328)	0.173 (0.342)	0.077 (0.361)
Parameter (p)	1.847	2.646	2.133	2.351	1.878
Log pseudo likelihood	−17.822	−16.391	−19.930	−18.110	−20.975
Wald χ²	138.85	235.17	121.64	336.66	206.21
Probability of χ²	0.000	0.000	0.000	0.000	0.000
Observations/time at risk	217	204	225	213	225
Number of subjects at risk	33	33	34	34	34
Failures	9	10	10	10	10

Note: Robust standard errors in parentheses. CPA = Comprehensive Peace Agreements; GDP = gross domestic product; PPP = purchasing power parity.

*p < .10.

**p < .05.

***p < .01.

****p < .001.

Model 3 tests for the effect of sequencing high levels of implemented accommodation before holding the first post-accord election. In model 3, the variable High Accommodation × First Post-accord Election is highly significant and positive, which indicates that the unmediated effect of post-accord elections is generally hazardous, however, this negative effect can be neutralized if the implementation of accommodation comes first in the sequence of accord implementation. As such, the results provide strong support for our Hypothesis. Model 4 includes a control for the number of armed personnel in the national military, which controls for state capacity and the potential costs to be suffered by the opposition if war is resumed. Model 5 shows that the findings are robustly independent of any relationship with demobilization implementation.

Among the controls, the Number of accommodation provisions in the accord is found to be insignificant in Model 1. This provides evidence that the pacifying effect of implemented accommodation stems from the actual implementation of the accommodation provisions and not merely their presence in the agreement. In models 2, 3, and 4 we include high demobilization and find that achieving high levels of implemented demobilization has a highly significant pacifying effect. In model 5, we find demobilization implementation also positive and significant. Notably, executive constraint is consistently positive and significant in every model, suggesting that peace survival is very much aided by checks on executive power. Higher infant mortality rate is negative and significant in model 2. Higher GDP per capita leads to a significantly longer peace duration in model 1. Conflict type is positive and significant in models 2, 3, 4, and 5. We did not find consistent support for other controls.

Figure 1 depicts four possible paths that result from interacting high accommodation and first post-accord election (based on model 3). Each line in the figure represents a different sequential path and the predicted probability of peace survival over a ten-year period. The bottom line shows the probability of peace survival when elections are held without preceding high accommodation. As can clearly be seen, this is the lowest and steepest curve representing the most dangerous potential path depicted. The second-to-lowest curve shows the probability of peace survival in the absence of both elections and high accommodation. This shows to be an unstable path with a greatly reduced chance that peace will prevail. The second highest curve shows the probability of peace survival, given high accommodation and no election. The highest curve or the best possible path to long-term peace is achieved when sequencing high levels of accommodation implementation before the holding of the first post-accord election. It should be noted that the difference between the two best scenarios—elections preceded by high accommodation and high accommodation without elections—is slight, yet the predictions are very much in line with our argument in that holding elections without preceding accommodation is a dangerous sequence with a high risk of failure, but preceding accommodation implementation alleviates this high risk.

Figure 1.

The sequencing of high accomodation implementation before first post-accord election and peace survival.

The distinction that should be of primary importance to scholars and practitioners seeking the optimal sequence for achieving an enduring peace after civil war is the difference between holding a post-accord election with high levels of accommodation having been implemented and holding a post-accord election without high levels of accommodation having been implemented. We explored this difference using the lincom function in STATA^® to compute a test of the null hypothesis that the difference between the two coefficient estimates is equal to zero. The results confirmed the highly significant nature (P < .000) and magnitude of the distinction between the two sequences (difference between coefficient estimates = 2.133, standard error = 0.477, 95 percent confidence interval = [1.198, 3.06]). Our interpretation of the range of findings shown in the analysis is that accommodation measures have a strong direct peace-promoting effect that immediately begins with their implementation. The cumulation of these effects builds an atmosphere of increased trust in the competence of the electoral process and results, thereby allowing elections to serve their important democratic purpose of regulating contestation to rule through nonviolent participatory competition.

A Note on Endogeneity and Epiphenomenality

Although the findings strongly support our argument, endogeneity could pose a potential threat to our findings if the same factors that influence the presence of accommodation provisions in peace agreements also make the recurrence of armed conflict less likely. If this were the case, then both accommodation and a greater probability of peace stem from these antecedent conditions, and we would be wrong to attribute causality to the implementation of accommodation. We investigated this question empirically and we also controlled for the conditions that might influence the presence or absence of accommodation provisions in the analysis. The general issue of endogeneity—as it relates to accord content—has been explored in prior studies (Fortna 2003, 339). Specifically, Hartzell and Hoddie (2007) found that political power sharing was more likely in tougher cases (i.e., protracted conflicts) where one would presumably expect less implementation and lower-quality elections. We also examined which accords produced political power sharing provisions for our population of CPAs. The average number of previous battle deaths for accords with a transitional power sharing government provision was triple that of accords without the provision. We also examined conflict duration and found that accords with and without transitional power sharing came out of civil wars that lasted an average of 12.7 versus 9.7 years. We also looked at what types of accords contain all three accommodation provisions versus those that contain zero. We found that our three accommodation provisions are far more likely in tough cases. Accords that contain all three accommodation provisions had a previous death toll from the civil war four times higher than those accords with none of the three and three times higher than accords with only one of the three provisions. Duration of the civil war follows the same pattern, that is, accords with all three accommodation provisions had an average duration of 120 months versus 84 months for accords without any accommodation provisions. Based on this evidence, and the direction of our hypotheses, we do not see endogeneity as threatening our findings in a serious way. On the contrary, had accommodation provisions been negotiated in the easier cases, the peace-promoting effects of their implementation would have been greater than our findings indicate.

Before we conclude the article, two brief case examples from Tajikistan and Angola further illustrate the prominent role given to the three accommodation provisions at the center of our analysis. In both cases, fights over the implementation of accommodation provisions represented the main axes of contention.

Tajikistan: June 1997 to December 1999

After the general agreement between the Tajik government and the United Tajik Opposition (UTO) in 1997, two years of extreme volatility followed. After the accord, ceasefire violations, including bombings, were commonplace.⁷ In 1998, clashes near the city of Romit killed numerous people and a top UTO leader on the joint commission was assassinated (United Nations Mission of Observers in Tajikistan [UNMOT] 1998). The US Ambassador to Tajikistan, at the time (who had been a participant in the talks), reported that “the two sides were slow to implement the amnesty and POW release, to establish and fill the assembly areas with opposition fighters, and to set procedures for return of opposition fighters from Afghanistan” (Smith 1999, 245). As a result, “the security situation continued to be unstable” (Smith 1999, 245) and demobilization was moving at “a slow grind” (Kannangara, Solijonov, and Khoshmukhamedov 2004, 9).

As for the root cause of these problems, Smith (1999, 245) pointed to delays in the implementation of the political power sharing arrangement, stating that, “the UTO asserted that its forces could not give up their weapons and demobilize before the opposition received the 30 per cent of governmental positions due to it.” The second major problem, according to Smith, involved the government’s delay in implementing amnesty, “The UTO also argued that the amnesty process should be completed prior to demobilization and disarmament, so that UTO fighters would know where they stood with the law before surrendering their weapons” (1999, 245). UNMOT reported that the release of POWs was delayed by the government and that this was a source of great concern for the UTO and its fighters. Out of the 700 UTO cases submitted to the government for amnesty in the first year (1997), the government only released seventy-eight prisoners (UNMOT 1997). In 1998, the government announced that all charges against UTO leaders were dropped but did not include a blanket amnesty for UTO fighters, an omission that observers cited as causing low rates of participation and major delays in the disarmament, demobilization, and reintegration process (Abdullaev and Barnes 2001). As of May 1998, of the 1,370 cases submitted to the government, 399 had been granted amnesty (UNMOT 1998).

As the due date for the first post-accord election approached, both UTO and government leaders directly expressed their concerns about the current impasse in implementation and “[i]t was recognized that the failure to address this issue in an expeditious manner would probably jeopardize the successful implementation of the electoral process in 2000 and thereby negatively impact the conclusion of the peace process” (Kannangara, Solijonov, and Khoshmukhamedov 2004, 9). Things began to improve in Tajikistan only after the level of implementation for power sharing, amnesty, and POW release began to tick upward. Kannangara, Solijonov, and Khoshmukhamedov (2004, 9) describe this shift in priorities as stemming directly from the mutual fear that if these delays in implementation were not quickly resolved, UTO fighters would begin to exit the peace process, putting the upcoming elections in peril.

Angola: November 1994 to July 1996

The Lusaka Protocol signed between the Angolan government and the National Union for the Total Independence of Angola (UNITA) in 1994 contained provisions for a transitional power sharing government, amnesty, and mutual POW release. By this time, Angola had already returned to civil war once, after UNITA contested the results of the 1992 election.⁸ In this, the second attempt to negotiate a peaceful settlement, political power sharing would not be implemented for over two years, the release of POWs would take two years, and amnesty would become the central problem in the ensuing breakdown that led back to war.

The Angolan parliament immediately passed an amnesty bill as part of the Lusaka Accord, but UNITA leaders argued that it contained too many exceptions, allowed for civil prosecutions against UNITA members, and did not legalize UNITA as a political party—as stipulated by the accord.⁹ Most importantly, the bill only covered up to the signing of the Lusaka accord, although fighting took place days after. UNITA leaders made repeated statements that they could not come forward under the current amnesty bill. Just weeks after the bill, General Gato, UNITA’s Deputy Secretary of Foreign Affairs, told reporters that the Lusaka peace process was breaking down over how “amnesty was being handled.” Three months after the accord, UNITA released a formal resolution asking the government for “general and total amnesty embracing all the period of the Angolan conflict.”¹⁰ Over one year later, it was reported that “no further progress has been achieved with respect to the release of prisoners” (United Nations Angola Verification Mission III 1995). In February 1996, Savimbi publicly stated that he and his forces would demobilize if the government would pass a new amnesty bill.¹¹ The government responded by withdrawing from the joint commission on the implementation of the peace accord and ramped up its offensive attacks.^12,13 Jorge Valentim, who negotiated for UNITA in the Lusaka peace talks, explained the breakdown to diplomats in 1996, “No one could expect a UNITA general to come out of the bush and enter one of the UN’s demobilization camps if, as a consequence, he would be prosecuted by the government.”¹⁴

At every turning point in the Lusaka peace process, the accommodation measures that were mutually agreed upon were bypassed and neglected rather than implemented. The political power sharing arrangement was not implemented and a general amnesty bill was not passed. In 1998, four years after the accord, and in the midst of full civil war,¹⁵ the parliament passed legislation, making UNITA a legal political party with Jonas Savimbi as party head.¹⁶ Savimbi was later killed in battle.

Conclusion

Several important conclusions emerge from the results. The findings indicate that it is always better (given the analyses we have conducted) to implement accommodation provisions as soon as possible after the signing of a peace agreement (and this may be true of other civil war outcomes as well). Implementing accommodation provisions seems to be a highly effective strategy in getting a peace process started along the right path and this appears to be the case with or without elections. If elections are to be held, failing to implement accommodation provisions beforehand is a dangerous sequence that is associated with a much higher risk of returning to war. If accommodation is implemented first, however, elections can serve their democratic purpose without increasing the risk of returning to war. Sequencing, then, appears to be an important part of designing an overall implementation process around elections. The results also suggest that it is the process of implementing the accommodation provisions—and not simply having them in an accord—that makes a real difference. There appears to be nothing to gain by agreeing to accommodation measures apart from implementing them—providing further evidence of accommodation’s facilitative power.

Results from this study can be expanded in a variety of ways. The key contribution of the article is the advancement and testing of a sequencing argument concerning the potential of accommodation. Many other implementation sequences in peace processes are possible and much more attention should be focused on how the experiences of the conflict actors in the earliest weeks and months of the implementation process act as priming experiences that are likely to influence their perception of the probable outcomes of future implementation processes. Future research could expand upon the sequencing approach to implementation used in this article to partition peace processes into smaller periods of time in order to explore the important downstream effects of implementation in one area on subsequent implementation processes. Logical arguments and case evidence, in combination with longitudinal data on peace agreement implementation, should further our understanding of peace processes as complex causal chains.

Footnotes

Acknowledgments

Authors are listed alphabetically. We thank Peter Wallensteen for comments on an earlier version of the manuscript. We are grateful to Paul Huth for his many excellent suggestions on improving the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was partially supported with grants from the United States Institute of Peace (149-06F) and the National Science Foundation (0921818).

Notes

References

Abdullaev

Kamolidin Nadzhmidinovich

Barnes

Catherine

. 2001. Politics of Compromise: The Tajikistan Peace Process. London, UK: Conciliation Resources.

Barnett

Michael

. 2006. “Building a Republican Peace: Stabilizing States after War.” International security 30 (4): 87–112.

Box-Steffensmeier

Janet M.

Jones

Bradford S.

. 2004. Event History Modeling. A Guide for Social Scientists. Cambridge, UK: Cambridge University Press.

Brancati

Dawn

Snyder

Jack L.

. 2013. “Time to Kill: The Impact of Election Timing on Postconflict Stability.” Journal of Conflict Resolution 57 (5): 822–53.

Carothers

Thomas

. 2007. “The ‘Sequencing’ Fallacy.” Journal of Democracy 18 (1): 12–27.

Cleves

Mario Alberto

Gould

William W.

Gutierrez

Roberto G.

. 2004. An Introduction to Survival Analysis Using Stata. College Station, TX: Stata Press.

Collier

Paul

. 2009. Wars, Guns, and Votes: Democracy in Dangerous Places. New York: Harper Collins Publishers.

Darby

John

Ginty

Roger Mac

. 2008. “Introduction: What Peace? What Process?” In Contemporary Peacemaking: Conflict, Peace Processes and Post-war Reconstruction, 2nd ed., edited by Darby

John

Ginty

Roger Mac

, 1–10. London, UK: Palgrave Macmillan.

DeRouen

Karl

Lea

Jenna

Wallensteen

Peter

. 2009. “The Duration of Civil War Peace Agreements.” Conflict Management and Peace Science 26 (4): 367–87.

10.

DeRouen

Karl

Ferguson

Mark J.

Norton

Samuel

Park

Young Hwan

Lea

Jenna

Streat-Bartlett

Ashley

. 2010. “Civil War Peace Agreement Implementation and State Capacity.” Journal of Peace Research 47 (3): 333–46.

11.

Diamond

Larry

. 2006. “Promoting Democracy in Post-conflict and Failed States.” Taiwan Journal of Democracy 2 (2): 93–116.

12.

Doyle

Michael W.

Sambanis

Nicholas

. 2000. “International Peacebuilding: A Theoretical and Quantitative Analysis.” American Political Science Review 94 (4): 779–801.

13.

Doyle

Michael W.

Sambanis

Nicholas

. 2006. Making War and Building Peace: United Nations Peace Operations. Princeton, NJ: Princeton University Press.

14.

Fearon

James D.

1995. “Rationalist Explanations for War.” International Organization 49 (3): 379–414.

15.

Fehr

Ernst

Gächter

Simon

. 2002. “Altruistic Punishment in Humans.” Nature 415 (6868): 137–40.

16.

Filson

Darren

Werner

Suzanne

. 2002. “Bargaining and Fighting: The Impact of Regime Type on War Onset, Duration, and Outcomes.” American Journal of Political Science 46:296–313.

17.

Flores

Thomas Edward

Nooruddin

Irfan

. 2012. “The Effect of Elections on Postconflict Peace and Reconstruction.” Journal of Politic 74 (2): 558–70.

18.

Fortna

Virginia Page

. 2003. “Scraps of Paper? Agreements and the Durability of Peace.” International Organization 57 (2): 337–72.

19.

Fortna

Virginia Page

. 2004. “Does Peacekeeping Keep Peace? International Intervention and the Duration of Peace after Civil War.” International Studies Quarterly 48 (2): 269–92.

20.

Gintis

Herbert

Smith

Eric Alden

Bowles

Samuel

. 2001. “Costly Signaling and Cooperation.” Journal of Theoretical Biology 213 (1): 103–19.

21.

Gleditsch

Nils Petter

Wallensteen

Peter

Eriksson

Mikael

Sollenberg

Margareta

Strand

Håvard

. 2002. “Armed Conflict 1946-2001: A New Dataset.” Journal of Peace Research 39 (5): 615–37.

22.

Hartzell

Caroline

Hoddie

Matthew

. 2003. “Institutionalizing Peace: Power Sharing and Post-civil War Conflict Management.” American Journal of Political Science 47 (2): 318–32.

23.

Hartzell

Caroline

Hoddie

Matthew

. 2007. Crafting Peace: Power Sharing and the Negotiated Settlement of Civil Wars. University Park: The Pennsylvania State University Press.

24.

Hoddie

Matthew

Hartzell

Caroline

. 2003. “Civil War Settlements and the Implementation of Military Power-sharing Arrangements.” Journal of Peace Research 40 (3): 303–20.

25.

Högbladh

Stina

. 2012. “Peace Agreements 1975-2011—Updating the UCDP Peace Agreement Dataset.” In States in Armed Conflict 2011, edited by Therése

Pettersson

Themnér

Lotta

, 39–56. Uppsala, Sweden: Department of Peace and Conflict Research Report 99, Uppsala University.

26.

Höglund

Kristine

. 2009. “Electoral Violence in Conflict-ridden Societies: Concepts, Causes, and Consequences.” Terrorism and Political Violence 21 (3): 412–27.

27.

Höglund

Kristine

Jarstad

Anna K.

Kovacs

Mimmi Söderberg

. 2009. “The Predicament of Elections in War-torn Societies.” Democratization 16 (3): 530–57.

28.

Huntington

Samuel P.

1968. Political Order in Changing Societies. New Haven, CT: Yale University Press.

29.

IDEA (International Institute for Democracy and Electoral Assistance). 2013. International Institute for Democracy and Electoral Assistance. Accessed June 14, 2013. http://www.idea.int/uid/index.cfm.

30.

Jarstad

Anna K.

Nilsson

Desirée

. 2008. “From Words to Deeds: The Implementation of Power-sharing Pacts in Peace Accords.” Conflict Management and Peace Sciences 25 (3): 206–23.

31.

Joshi

Madhav

. 2013. “United Nations Peacekeeping, Democratic Process, and the Durability of Peace after Civil Wars.” International Studies Perspectives 14 (3): 362–82.

32.

Joshi

Madhav

Darby

John

. 2013. “Introducing the Peace Accords Matrix (Pam): A Database of Comprehensive Peace Agreements and Their Implementation, 1989–2007.” Peacebuilding 1 (2): 256–74.

33.

Joshi

Madhav

Michael Quinn

Regan

Patrick

. 2015. “Annualized Implementation Data on Comprehensive Peace Accords, 1989-2012.” Journal of Peace Research 52 (4) 551–62.

34.

Joshi

Madhav

Lee

Sung Yong

Ginty

Roger Mac

. 2014. “Just How Liberal Is the Liberal Peace?” International Peacekeeping 21 (3): 364–89.

35.

Kannangara

Solijonov

Khoshmukhamedov

. 2004 Outcome Evaluation Report: Sustainable Disarmament, Demobilization and Reintegration of Ex-combatants and Conversion of Military Assets to Civilian Use Dushanbe, Tajikistan: UNDP.

36.

Kumar

Krishna

, ed. 1998. Postconflict Elections, Democratization, and International Assistance. Boulder, CO: Lynne Rienner Publishers.

37.

Kydd

Andrew

. 2000. “Trust, Reassurance, and Cooperation.” International Organization 54 (2): 325–57.

38.

Kydd

Andrew

. 2005. Trust and Mistrust in International Relations. Princeton, NJ: Princeton University Press.

39.

Lindberg

Staffan I.

2003. “The Democratic Qualities of Competitive Elections: Participation, Competition and Legitimacy in Africa.” Commonwealth and Comparative Politics 41 (3): 61–105.

40.

López-Pintor

Rafael

. 1997. “Reconciliation Elections: A Post-cold War Experience.” In Rebuilding Societies after Civil War: Critical Roles for International Aasiatance, edited by Kumar

Krishna

, 43–62. Boulder, CO: Lynne Rienner.

41.

Lyons

Terrence

. 2002. “The Role of Postsettlement Elections.” In Ending Civil Wars: The Implementation of Peace Agreements, edited by Stedman

Stephen J.

Rothchild

Donald

Cousens

Elizabeth M.

, 663–71. Boulder, CO: Lynne Rienner.

42.

Manning

Carrie

. 2004. “Elections and Political Change in Post-war Bosnia and Herzegovina.” Democratization 11 (2): 60–86.

43.

Mansfield

Edward D.

Snyder

Jack L.

. 2005. Electing to Fight Cambridge, MA: MIT Press.

44.

Mansfield

Edward D.

Snyder

Jack L.

. 2007. “The Sequencing ‘Fallacy’.” Journal of Democracy 18 (3): 5–9.

45.

Marshall

Monty G.

Gurr

Ted Robert

Jaggers

Keith

. 2013. Polity IV Project, Political Regime Characteristics and Transitions, 1800-2012. Accessed May 1, 2013. http://www.systemicpeace.org/polity/polity4.htm

46.

Mattes

Michaela

Savun

Burcu

. 2009. “Fostering Peace after Civil War: Commitment Problems and Agreement Design.” International Studies Quarterly 53 (3): 737–59.

47.

Mattes

Michaela

Savun

Burcu

. 2010. “Information, Agreement Design, and the Durability of Civil War Settlement.” American Journal of Political Science 54 (2): 511–24.

48.

Pagani

Fabrizio

. 1998. “The Peace Process at its Culmination: The Reconciliation Elections.” In The Psychology of Peacekeeping, edited by Langholtz

Harvey J.

, 223–38. Westport, CT: Praeger Publishers.

49.

Paris

Roland

. 1997. “Peacebuilding and the Limits of Liberal Internationalism.” International Security 22 (2): 54–89.

50.

Paris

Roland

. 2004. At War’s End: Building Peace after Civil Conflict. New York: Cambridge University Press.

51.

Powell

Robert

. 1996. “Uncertainty, Shifting Power, and Appeasement.” American Political Science Review 90 (4): 749–64.

52.

Quinn

J. Michael

David Mason

Gurses

Mehmet

. 2007. “Sustaining the Peace: Determinants of Civil War Recurrence.” International Interactions 33 (2): 167–93.

53.

Radin

Andrew

. 2013. “The Impracticality of Sequencing: Domestic Opposition and Election Timing after Civil Wars.” Paper Presented at the American Political Science Association Annual Meeting, Chicago, IL.

54.

Reilly

Benjamin

. 2002. “Post-conflict Elections: Constraints and Dangers.” International Peacekeeping 9 (2): 118–39.

55.

Roeder

Philip G.

Rothchild

Donald

, eds. 2005. Sustainable Peace: Power and Democracy after Civil Wars. Ithaca, NY: Cornell University Press.

56.

Slantchev

Branislav

. 2003. “The Principle of Convergence in Wartime Negotiations.” American Political Science Review 97 (4): 621–32.

57.

Smith

Grant R.

1999. “Tajikistan: The Rocky Road to Peace.” Central Asian Survey 18 (2): 243–51.

58.

Snyder

Jack

Vinjamuri

Leslie

. 2003. “Trials and Errors: Principles and Pragmatism in Strategies of International Justice.” International Security 28 (3): 5–44.

59.

Stedman

Stephen John

. 2002. “Policy Implications.” In Ending Civil Wars: The Implementation of Peace Agreements, edited by Stedman

Stephen J.

Rothchild

Donald

Cousens

Elizabeth M.

, 663–71. Boulder, CO: Lynne Rienner.

60.

Themnér

Lotta

Wallensteen

Peter

. 2013. “Armed Conflicts, 1946-2012.” Journal of Peace Research 50 (4): 509–21.

61.

Tilly

Charles

. 1978. From Mobilization to Revolution. New York: Mac-Graw Hill.

62.

UNAVEM III (United Nations Angola Verification Mission III). 1995. Report of the Secretary-general to the Security Council. (S/1995/1012, December 7, 1995).

63.

UNMOT (United Nations Mission of Observers in Tajikistan). 1997. Report of the Secretary General to the Security Council (S/1997/859, November 5, 1997).

64.

UNMOT (United Nations Mission of Observers in Tajikistan). 1998. Report of the Secretary General to the Security Council (S/1998/374, May 6, 1998).

65.

Uppsala Conflict Data Program. 2013. “UCDP Conflict Encyclopedia.” Uppsala University. Accessed December 1, 2012. www.ucdp.uu.se/database.

66.

Wagner

R. Harrison

. 2000. “Bargaining and War.” American Journal of Political Science 44 (3): 469–84.

67.

Walter

Barbara F.

1997. “The Critical Barrier to Civil War Settlement.” International Organization 51 (3): 335–64.

68.

Walter

Barbara F.

1999. “Designing Transitions from Civil War: Demobilization, Democratization, and Commitments to Peace.” International Security 24 (1): 127–55.

69.

Walter

Barbara F.

2002. Committing to Peace: The Successful Settlement of Civil Wars. Princeton, NJ: Princeton University Press.

70.

World Bank. 2013. “World Development Indicators.” Accessed May 1, 2013. http://data.worldbank.org/.