Political party formation by former armed opposition groups after civil war

Under what conditions are rebel groups successfully incorporated into democratic politics when civil war ends? Using an original cross-national, longitudinal dataset, we examine political party formation by armed opposition groups over a 20-year period, from 1990 to 2009. We find that former armed opposition groups form parties in more than half of our observations. A rebel group’s pre-war political experience, characteristics of the war and how it ended outweigh factors such as the country’s political and economic traits and history. We advance a theoretical framework based on rebel leaders’ expectations of success in post-war politics, and we argue that high rates of party formation by former armed opposition groups are likely a reflection of democratic weakness rather than democratic robustness in countries emerging from conflict. ARTICLE HISTORY Received 19 January 2016; Accepted 3 February 2016


Introduction
The end of the Cold War presaged the end of many civil wars around the globe, with more such conflicts ending in the 15 years after 1989 than during 45 years of the Cold War. 1 Democratization has formed the cornerstone of most peace settlements in this period of "liberal peacebuilding," with the introduction or expansion of electoral politics providing an important and highly visible marker of the war to peace transition. 2 Liberal peacebuilding presumes that armed opposition groups will lay down their weapons and replace violent conflict with political competition. To what extent have armed opposition groups embraced this model?
This article presents a systematic exploration of the factors that influence whether or not an armed opposition group forms a political party following civil war. We constructed a cross-national, longitudinal dataset that tracks the incorporation of armed opposition groups into democratic politics after civil war over a 20-year period, from 1990 to 2009. We include conflicts that ended during this period; these include some that began as early as the 1960s and as late as 2009, bridging the Cold War and post-Cold War periods.
The dataset includes information on the characteristics of the armed conflict and the armed opposition groups, as well as on the characteristics of the post-war political system and the post-war electoral performance of former armed opposition groups. 3 Following Sartori, we define a political party as "any political group that presents at elections, and is capable of placing through elections, candidates for public office". 4 Party formation occurs when an armed opposition group registers as a legal party eligible to compete in elections.
The transformation of rebel groups into parties in no way guarantees either peace or democracy. When armed opposition groups register as legal political parties they are taking the necessary first step toward accepting political competition as a substitute, or at least a supplement, to violence as a route to political power.
A growing literature explores the transition of armed opposition groupsincluding both terrorist organizations and rebel groupsto political parties. 5 Most of these are empirically rich single case or small-n comparative studies focusing on a handful of key cases, and they have yielded important insights that deserve further exploration.
This project seeks to build on that base. Our independent variables are drawn from it and from the vast literature on party development, civil wars and post-war politics. A survey of this literature suggests possible correlations between post-rebel party formation and pre-war political experience of the rebel group; conflict intensity and duration; how the war ended; and the political and economic environment at war's end.
In the section that follows, we construct hypotheses about these relationships that we then test against our data. We find that rebel groups form parties more often than not (54.8% of the time). Rebel groups with pre-war political and especially electoral experience are significantly more likely to form parties and that both comprehensive and partial peace agreements make post-rebel party formation more likely. We also find that more violent conflicts reduce the likelihood that rebel groups will form parties after the war. Variables relating to the post-war political and economic environment were not significant, an unexpected and intriguing result.
Our results must be interpreted with caution. Large-n comparative research on postconflict politics presents a number of challenges, including multicollinearity, potentially complex interactions and considerable variation across a relatively small number of country cases. While most statistical treatments assume that independent variables are largely independent of one another, we know that empirically the variables we identify are linked to one another in complex ways, and the impact of one factor is likely to be contingent on the presence of others.
For example, a negotiated peace might give warring parties that participate in negotiations opportunities to shape post-war democratic institutions in ways beneficial to themselves. A country's pre-war experience with democracy might also affect the specific types of democratic institutions that are created in the post-war period. The intensity of the war, the history of pre-war politics and the nature of the peace settlement might all affect the confidence of party leaders in the durability and reliability of post-war institutions, thus affecting their calculations regarding the value of participating in politics.
One of the goals of this paper is to tease out these relationships in order to contribute to theory building in this area. Though our aim is primarily theory-building rather than theory testing, we do explore the explanatory power of competing explanations. In the absence of clear theoretical reasons to privilege one set of explanations over another, we give equal weight to each of the competing arguments we present.
Understanding post-war political incorporation of armed groups The literatures on party development, political transition and civil war suggest that factors related to party history, war characteristics, the war to peace transition and the post-war political environment might be expected to influence post-rebel party formation. 6 In this section we construct hypotheses about these relationships, which we then test against our data. Our inquiry begins with two assumptions, both drawn from the comparative literature on party organization and development. First, rebel groups and political parties are not unitary actors but, as Sartori has argued about parties, both are comprised of sub-groups whose membership and interests may shift over time. These sub-groups might be based on deeply rooted ideological differences or on differences over tactics or strategy arising from the functional role of some sub-groups over others. 7 Second, like most organizations, rebel groups and political parties seek both to achieve particular goals and to sustain themselves as organizations. 8 Their leaders face a similar dual task as they seek to maintain their leadership positions and preserve the organization itself. These goals are sometimes in tension with one another. We hypothesize that leaders will make decisions about whether or not to form a party and participate in post-war electoral politics based on whether or not doing so threatens or bolsters the survival of their organization, their leadership role in the organization, or both.

Party history
The case for considering a party's prior political experience is straightforward. If an armed opposition group was a political party prior to the war, it might have a level of organizational structure and coherence, as well as experienced political cadres and a collective identity, that could help sustain it in post-war politics. 9 In the face of the very different set of challenges posed by democratic politics, compared to military imperatives, such organizations may be more resilient. There is considerable evidence in the broader comparative literature that parties that played a substantial role in the period preceding the transition to democracy tend to predominate afterwards, unless they are formally banned from participation in politics. 10 For example, the literature on successor parties in the former Soviet bloc and in post-authoritarian regimes in Africa suggests that successor parties have proven quite durable. 11 To explore whether parties with prior experience as political organizations do better than those that first came into existence as armed opposition groups, we code whether a rebel group had existed as a political party immediately prior to the outbreak of war.
Of course, this is a simplification of a complex reality. Rebel groups that were prewar parties differ from one another in important ways. Some were narrowly based while others were broad fronts; their social bases differed, as did the competitiveness of the political environment they faced and the rules and tools of competition. Organizational characteristicshow centralized and how institutionalized the party wasare also likely to be relevant. 12 Some used violence as an ancillary strategy, some did not. Nevertheless, as a first cut it is worth grouping together armed opposition groups (AOGs) with a past organizational life in politics as a group distinct from those that were formed as rebel groups per se. As much of the broader comparative party literature has established, an organization's past is likely to affect its future. Pre-war experience as a party creates human and organizational capital that can be useful in forming a new party. 13 Similarly, we expect that parties with pre-war experience in electoral environments are likely to perform even better in post-war electoral politics than those with non-electoral political experience. Adjusting to the challenges of electoral politics should be easier for parties that have previously had to develop decision-making mechanisms to select candidates, to adjudicate between competing ideas about party platform and to maintain a voting base. Indeed, De Zeeuw and his contributors find that past experience with democracy is an asset to rebel groups in the transition to party politics. 14 We thus code parties according to whether or not they competed in elections prior to the post-war period. Both this variable and prior political experience are dummy variables.

War legacies
Where the literature on post-Communist successor parties focuses on the historical legacy of Communist rule, the post-conflict literature explores the implications of the war itself for post-war party development. 15 Finding theoretical guidance for this question is challenging. Much of the literature is focused on peace as an outcome. While certainly peace, at least in the short term, is essential for the creation of an electoral infrastructure that would incentivize and mobilize party formation by rebel groups, the wartime characteristics require further exploration in order to yield useful hypotheses on party formation, as we discuss below. New work is beginning to explore the impact of civil war on democracy and its component parts. 16 Findings are mixed. For example, Fortna and Huang find that for the most part "characteristics of the civil war have relatively little effect on the prospects for post-war democratization". 17 Drilling down from democracy to party system, Ishiyama finds that conflict intensity does have a significant impact, with more intensive conflicts leading to dominant-party systems. 18 The war characteristics examined in our analysis are conflict duration and conflict intensity (battle deaths). These variables are drawn from the Uppsala Conflict Data Program (UCDP) Conflict Termination and PRIO Battle-Related Deaths datasets. 19 Intensity is coded as the sum of battle deaths for a conflict episode based on estimates from the UCDP interval data on battle deaths. 20 We code conflict duration in months based on the start and end dates provided in the UCDP data. Other war-related characteristics, such as the wartime strategies of combatant groups, rebel groups' finance arrangements and support from external sponsors, would also be worth investigating for their impact on post-rebel party formation. Better data on these characteristics is now emerging from new case studies and small-n comparative work, allowing future research to build useful variables for some of these factors. 21 However, it remains difficult to collect accurate data on these variables for the full range of cases, and here we have opted to use battle deaths and duration as a first cut.
To understand the possible impact of war duration and intensity on post-conflict politics, we can start by building on the findings of the conflict recurrence literature, though the implications of those findings for post-rebel party formation might be quite different than they are for conflict recurrence. For example, there is considerable evidence that higher intensity warsthose that cause high numbers of deaths and largescale displacementare less likely to be followed by lasting peace. 22 One reason for this might be that more intense wars entrench wartime cleavages that are hard to overcome.
But the creation of entrenched wartime cleavages might actually improve the chances that an armed opposition group decides to form a party, since strong cleavages may increase cohesion within the parties on either side, and the ability to count on a clearly identified popular base would likely improve the party's electoral performance. 23 Ishiyama's finding that "bloodier civil wars promote the emergence of dominant-party systems with reduced electoral volatility" may offer support for the hypothesis that intense wars freeze wartime cleavages. 24 On the other hand, it is important to consider how violence has been used. Battle deaths may be a function of rebel or government strategy and tactics, of the "technology of violence" and the degree to which conflict is asymmetrical, since weaker groups may use more violence as a force magnifier. Without knowing who were the victims of violence and the relationship of those victims to different belligerents in the conflict, it is difficult to interpret the substantive impact of the correlation between battle deaths and party formation. We do not explore this intermediate relationship between violence and cleavage structures in this dataset, but it is worth examining in future research. For now, we simply offer an exploration of the relationship between battle deaths and party formation as a first cut on which to build a more nuanced investigation. We hypothesize that more violent conflicts are more likely to produce post-rebel party formation, because they are more likely to entrench cleavages around which parties could mobilize votes. Mobilization on each side might result from positive support for a side, or from security calculations that dictate cooperation in exchange for survival. 25 Since duration is closely related to other factors such as battle deaths, it is difficult to isolate its effects. However, we hypothesize that longer wars should produce a higher likelihood of post-rebel party formation for three reasons. First, long wars, like high intensity wars, may produce more entrenched cleavage structures that could prove advantageous in electoral politics. Second, longer wars may allow time for armed opposition groups to develop a more coherent organizational structure or collective identity. Third, though mutual distrust between belligerents may be greater in the wake of a long war, potential spoilers looking for a return to war may face more resistance from warweary civilians, and may decide that forming a party and playing the electoral game is the path of least resistance. This is consistent with the "war weariness" thesis found in the literature on duration and conflict recurrence. 26 In sum, we expect that more intense conflicts, and longer conflicts, are more likely to result in post-rebel party formation.

War to peace transition
De Zeeuw and his contributors find that "perhaps the most important exogenous factor impacting rebel-to-party transformation is the extent to which the civil war has ended". 27 Wars that have come to a definitive end, and more specifically those that have ended in a peace agreement, are more conducive to rebel to party transformations. This is consistent with Doyle and Sambanis who, among others, argue that a peace agreement significantly boosts the chances of "participatory peace" after civil war. 28 Fortna and Huang find that negotiated settlements have a positive impact on postwar democratization, but only in the short term (2-5 years). 29 Just as a peace agreement is expected to increase the chances of successful transition from battlefield to political arena for armed opposition groups, a military victory by government is expected to hurt the chances of such a transition. In the case of military victory by the government, repression of opposition is not uncommon, and formal provision is less likely to be made for the rebel to party transformation. We would thus expect post-rebel party formation to be less likely in such cases. In the case of military victory by rebels, on the other hand, if the post-war political system is based on elections, we would expect to see such groups form parties and participate in elections.
Another possible outcome of civil war is a "separate peace", in which a peace agreement is signed that leaves one or more armed opposition groups still fighting. We theorize that even such a partial peace provides an incentive to the rebel groups included in the agreement to form parties and participate in politics in order both to maximize their chances of gaining a share of the political spoils and to distinguish themselves from insurgent groups still fighting. Indeed, we might expect such groups to be more likely to form parties than rebel groups subject to a comprehensive peace. Groups willing to negotiate a separate peace might have done so in part because they calculated that their chances of survival were better in the political arena than on the battlefield, and they may even have negotiated guarantees of their ability to participate in politics.
In sum, we expect that, all else equal, armed opposition groups are more likely to form parties when war ends in some sort of a negotiated peace agreement, whether comprehensive or "separate". Where there is provision for democratic post-war politics, military victory by the rebel group should also be conducive to party formation and persistence.

Post-war environment
The post-war institutional structure might well affect party formation. We explore the influence of post-war institutional incentives by examining both political openness at the end of the conflict and whether the electoral system is proportional vs. majoritarian. We assume that electoral systems based on proportional representation offer a greater promise of success for new parties, since entry barriers are lower. We code these as "permissive" electoral systems. Similarly, more open systems promise parties that contest elections greater assurance that the results of the poll will both reflect the will of the electorate and will be implemented. We code openness of the political system in terms of Freedom House political rights scores.
Finally, some scholars argue that structural conditions in the pre-war, wartime or post-war periods shape both the party and the political environment in which that party operates. These conditions are both beyond the reach of institutional design and independent of the organizational capacities or motivations of armed opposition groups. For example, pre-existing socioeconomic, political or cultural cleavages may continue to define the post-war political system and its dominant political actors. After the war, socioeconomic factors such as level and rate of economic growth, or regional economic inequalities, may impact the expected and actual performance of political parties, including post-rebel parties. In addition, the presence of armed conflict in neighbouring states, or the existence of "safe havens" or other exit strategies for would-be spoilers, might also affect the decisions of rebel leaders about whether or not to form a political party and under what circumstances to participate. We include controls for region, level of economic development and past history of democracy in the country. However, in this iteration we do not control for conflict in neighbouring countries or other factors related to the post-war security environment.

The dataset
Our dataset builds on the UCDP Conflict Termination and Actors datasets. It covers all intrastate conflict episodes ending and holding post-war elections between 1990 and 2009 and includes conflicts in 50 countries. Using these criteria, we coded 98 unique conflict episodes and identified 133 distinct armed opposition groups. 30 In our dataset, a single observation is an AOG-conflict episode, yielding 155 observations. 31 In 19 conflict episodes, armed opposition groups returned to fighting before the first post-conflict election. Armed opposition groups that returned to war before elections were held may have done so after calculating that they could not succeed politically, so it was important to include them in the dataset. However, we also want to know how many groups that do not return to conflict also do not decide to form parties or participate in electoral politics. When we exclude observations of groups that went back to fighting before the first election from the analysis, we get the same number of conflict-episodes and countries, but a smaller number of AOG-conflict episodes to 136 observations of 127 distinct rebel groups. We ran a separate analysis of these cases, but there was little difference from the full set thus the smaller set is not included below. Appendix 1 contains the list of countries, AOGs and parties in the dataset.
Our dependent variable, post-rebel party formation, is defined as whether a former armed opposition group that fought in the most recent conflict episode formed a legal political party. Our key independent variables relate to pre-war political experience; war characteristics; how the war ended; and post-war political characteristics. Table 1 provides a list of independent variables, definitions and sources.
Coding for this project presented some challenges. First, as other scholars have observed, AOGs with pre-war experience as political parties vary widely in the nature and length of that experience and in their organizational coherence and structure, as well as in their approach to the use of violence prior to the beginning of the conflict. 32 Our set includes well-institutionalized parties with decades of experience, personalized "cliques of notables," loosely organized fronts and everything in between. AOGs in the set reflect similar diversity. Prior electoral experience means the party participated in elections, regardless of how free and fair those elections might have been. Future studies could refine coding of parties' prior organizational experience to account more directly for organizational and ideational capital that post-rebel parties might inherit from a previous political incarnation.
We looked for post-war party formation in every post-war election, so we include parties that formed to contest elections immediately as well as those that formed later on. Patterns of politics that led to and followed war varied widely, from cases in which armed groups cycled into politics and back to the battlefield with some regularity, to those that followed a more linear path. Some armed opposition groups produced a single party that proved durable (UNITA, Renamo), while in other cases AOGs formed multiple parties at the first elections. Sometimes several rebel groups united to form a single party after the war (Democratic Republic of Congo).
Rebel groups are not the only organizations that use violence. Political parties may have armed wings or youth militias. In some post-conflict countries, violence around elections, by groups associated with particular parties, may be seen by these parties as politics as normal, rather than as armed opposition activity. Our definition of political party accommodates this possibility. For the purposes of this study, an AOG becomes a political party when it registers as a legal party and runs in its first election; a party is an organization that presents candidates for election and strives, as its primary goal, to place its candidates in office. If violence remains an ancillary strategy to achieving this goal, it does not necessarily disqualify an organization as a party ( Table 2).

Data analysis: post-rebel party formation
To analyse our data, we use a logit model, as our dependent variable is a simple dichotomous option of whether a party forms or not. Each case in this model refers to the end of a conflict episode reported in the UCDP Actors dataset that was followed by some form of elections during the observation period between 1990 and 2009. A party was formed in 54.8% of the cases. DV: Political party formation: Was a legal political party formed after the conflict episode? Sources: Nohlen et al. 2001;Nohlen et al. 1999;Nohlen 2005; Europa Regional Yearbooks H1: Prior experience as a political party makes post-rebel party formation more likely. IV: Did rebel group emerge from a political party formed before start of war? Sources: UCDP Actor Dataset v. 2. 1-2012;Nohlen et al. 2001;Nohlen et al. 1999;Nohlen 2005; Regional Yearbooks H2: Prior experience as an electoral political party makes post-rebel party formation more likely. IV: Did rebel group emerge from a political party that had contested elections before the start of the war? Sources: UCDP Actor Dataset v. 2. 1-2012;Nohlen et al. 2001;Nohlen et al. 1999;Nohlen 2005 IV: Freedom House Political Rights score Source: Freedom House data, available at https://freedomhouse.org/report/freedom-world-aggregate-andsubcategory-scores H10: More permissive rules of political competition increases likelihood of post-rebel party formation. IV: PR electoral system Sources: Nohlen et al. 2001;Nohlen et al. 1999;Nohlen 2005; Europa Regional Yearbooks H11: International supervision of peace implementation increases likelihood of party formation. IV: Multilateral peace operation in place Contact group or other committee of bilateral actors oversees implementation Source: UCDP Peace Agreement dataset v. 2.0, 19752.0, -20112.0, , presented in Harbom et al. 2006 We used multiple model specifications to examine multivariate correlations between a number of independent variables culled from the literature. The models were robust to changes in specifications, with only very slight changes between specifications. The models shown are estimates using clustered standard errors based on country-level factors, but the results were not significantly different when standard errors were clustered on conflict episodes or when non-clustered errors were utilized. 33 We also ran a simplified second model in which variables with little impact on the overall model fit were removed. 34 This second model is used to generate substantive statistics from the estimates.
The multivariate model is intended primarily as an exploratory analysis. This brings two cautions to bear. The first concerns the large number of variables. With so many variables it is likely that there are complex interactions between them, which will need to be accounted for in causal models developed from this analysis. Second, since this is an exploratory analysis of new data, we look not only at what is "confirmed" but also at those values that are close to significance, using graphical representations. 35 The largely heuristic nature of our analysis leads us to be concerned both with statistical consistency and with the substantive impact on party formation, which may be of greater value to consumers of political research. Table 3 presents the results of our logit model. Figure 1 provides a graphical interpretation of the substantive impact of the key dichotomous independent variables. It shows the average marginal effects calculated for a change from 0 to 1 for each. 36 For each one, the other variables are held to their mean value. (Estimates for political rights and battle deaths, our continuous variables, are not shown in the figure.)

Analysis
In both specifications of the model, we find variables from each of our conceptual areas to be influential. Consistent with our expectations, prior party experience, both electoral and non-electoral, was important. Whether or not a rebel group had a pre-war incarnation as  an electoral party was consistently the most statistically and substantively significant variable in our model. Experience as a party, but not an electoral party, prior to the conflict episode is statistically significant at the 90% level in both specifications, but is not as influential as pre-war electoral experience. Our findings regarding the post-war political environment were most surprising. We expected party formation to be more likely in more open political systems and where electoral rules meant lower barriers to entry. Instead, we found that AOGs were actually less likely to form parties in proportional representation systems than in those that utilized single-member systems. And parties formed from AOGs more often in countries where respect for political rights was lower (as measured by Freedom House scores).
These findings suggest that rebel leaders' decisions about party formation were not heavily influenced by calculations of their likely electoral success. If they were, we might expect to see higher rates of party formation in countries with a higher level of democratic openness (where new parties could expect a level playing field), or in cases with PR electoral systems (coded as "permissive"), where the barriers to entry for elected office would be lower. Perhaps merely participating in elections has instrumental value to these groups, even if the outcome is not electoral victory.
Alternatively, our assumption that proportional electoral systems mean lower barriers to office may be inaccurate for rebel groups. In fact, single member districts may well benefit rebels more, particularly if they have a territorial base that they controlled during wartime. Ishiyama also finds that electoral systems have different effects in post-war cases, noting that, unlike in other new democracies "these institutional factors had no effect on party systems characteristics among countries that had experienced a civil war". 37 We also found that in countries where there had been a multiparty election within 10 years of the start of the conflict episode (coded as "recent multiparty elections"), armed opposition groups were significantly less likely to form political parties. It may be that older, pre-war parties are able to crowd out newer entrants, including post-rebel parties. However, this finding must be interpreted with caution. First, the holding of pre-war elections tells us nothing about the age of the party system or of the dominant parties therein. Moreover, 10 years is a relatively broad timespan, and the quality of these pre-war elections varied considerably.
It is not surprising that AOGs with prior electoral experience are more likely to form post-rebel parties. Perhaps the polarizing effects of war reinforce or create entrenched cleavages that define the post-war party system regardless of post-war institutions.
But why should prior electoral experience matter when the likelihood of a level playing field appears unimportant? One possible answer is that parties with electoral experience might understand that electoral victory matters less than participation itself. The ability to field candidates for office is important even if electoral outcomes are not, as participation in elections allows groups to demonstrate their potential importance as allies or power-brokers even if they earn relatively small vote shares. For example, a party may use an election to demonstrate that it has deep support in a strategic region of the country, or that it has the support of a particular socioeconomic sector. Parties with prior electoral experience are likely to make a better showing regardless of how meaningful elections are. An alternative answer is that parties with prior electoral experience may be better positioned to manipulate electoral processes than those who have never contested an election.
Method of war termination had a significant and substantive impact on party formation. In our analysis, we use general peace agreement as the omitted reference category. Our first finding is that armed opposition groups that make a separate peace with the government are also more likely to form parties. This is not surprising, since parties that negotiate a separate peace are more likely to do so if they expect to survive the transition from battlefield to political arena. Not surprisingly, conflicts that end with the defeat of the rebel group by government forces or due to inactivity on the battlefield are less likely to lead to party formation. Wars ended through government victory or inactivity likely offer few if any assurances to rebel groups regarding their chances of survival in the political arenas.
Rebel victories were not significantly more likely to lead to party formation than a general peace agreement. This was counter to our expectations. The number of cases in this category is relatively small, so it is difficult to generalize. But the literature on organizational change would suggest that parties, like other organizations, adapt only when necessary to achieve their goals: political power, survival or policy change. Victorious rebel groups may have little reason to make the transition to party. 38 Measures of the intensity and duration of the conflict had differing impacts. Contrary to our expectations, more intense conflicts significantly reduce the likelihood that armed opposition groups make the transition to political parties. However, our measure of intensity (battle deaths) does not allow us to make distinctions about the targets and types of violence, both of which might affect the impact on cleavage formation and rebel groups' expectations about popular support after the war. And violence might itself be a reflection of the strengths and weaknesses of rebel groups. 39 Finally, higher levels of wartime violence (and longer wars) might be a function of government preferences regarding whether and when to attempt negotiated peace.
War duration had no significant impact in the first model shown and was dropped due to a high degree of collinearity with the indicator variable for Cold War era conflicts. Party formation was significantly more likely in Cold War era conflicts (those that began before 1990). These conflicts frequently involved more ideologicallyfocused combatants with a greater degree of external support. This degree of organization and financial support provides a base upon which to build a party. Also, such conflicts were almost all ended through a negotiated peace agreement overseen by Western donors and/or multilateral organizations. In such conflicts, the barriers to new party formation were extremely low, as provisions for participation by ex-belligerents were often written into the peace agreement, and the opportunity cost of not forming a party was high. In such cases, party formation was the price of admission to both formal and informal struggles for power and resources.
In light of this possibility, we decided to test the influence of external involvement in the supervision of peace processes more directly. Interestingly, we found that the presence of a peacekeeping mission per se had no meaningful impact on party formation by AOGs. However, if external involvement included a committee of donors to oversee the agreement, party formation was significantly more likely. When supervision committees included only domestic actors, there was no notable impact on party formation. Recent research has shown that bilateral actors like major donors can be more effective than multilateral peacekeeping operations in exercising conditionality in peacebuilding processes. 40 These findings suggest that these actors likely provide carrots and/or sticks that affect the calculations of armed opposition groups regarding party formation after war.

Conclusion
This article has examined the question of when and why rebel groups form political parties at war's end. We theorized at the outset that rebel leaders would establish parties when they believed that doing so would contribute to their goal of organizational survival and retention of their leadership positions within the organization. In other words, parties would form where the barriers to success are low, and/or where the opportunity costs of not forming a party are high. Our findings offer some support to this theory. We found that parties were especially likely to form: when rebel groups had prior political and especially electoral experience; when there was a peace agreement, and especially where rebels signed a "separate peace" with the government; and when external guarantors were present. These factors likely increase rebel leaders' confidence in their ability to compete and survive in the post-war system. Interestingly, post-war political openness did not affect the chances of party formation. This suggests that post-rebel party formation may have little to do with a party's expectations of electoral success in free and fair elections. Participation in flawed elections may be sufficient for AOGs to reap the expected benefits. Where democratic institutions are weak, as in many post-conflict countries, where there are few strong political challengers to the former belligerents, or where the participation of the rebels is viewed by donors or other influential actors to be important to peace, the risks associated with the transition from rebel group to party are reduced, and the opportunity costs of not participating in elections increase, regardless of whether the party is winning or losing.

Disclosure statement
No potential conflict of interest was reported by the authors.