Abstract
Efforts to strengthen democratic institutions have become a nearly universal feature of peacebuilding efforts in states recovering from civil war, yet such institutions often become dominated by a single actor or party. Why are some post-conflict incumbents able to dominate politics in this way, while others face more robust electoral competition? This study finds that post-conflict electoral competition depends on the types of capabilities that combatants develop during civil war. Where rebels develop capabilities that can be converted from military to electoral use, incumbents face stronger electoral challenges, but the effects are conditioned by the number of actors involved.
Introduction 1
Nearly every civil war that has ended since 1989 has involved efforts at implementing or strengthening democracy as part of the peace process. Democratic institutions are believed to be a mechanism for accommodating and resolving social conflict without violence, a key transformation in societies that have been through the horrors of war. Constitutions are re-written, political parties are trained, and elections are organized in an effort to normalize politics and move competition from the battlefield to the ballot box. This view of democracy as a pacifying force has become a centerpiece of development aid after civil wars, traceable at least as far back as the United Nations’ seminal document on peacebuilding, Agenda for Peace (Boutros-Ghali, 1992). “Institution building” after conflict now frequently includes prioritization of building “democratic governance” and strengthening the rule of law (see, e.g. United Nations Peacebuilding Commission, 2009).
On the surface, these efforts seem to have met with some success. Yet as scholars have emphasized in recent years, elections and other formally democratic institutions do not necessarily lead to democracy in practice (Diamond, 2002; Levitsky and Way, 2010). These formal structures must be accompanied by de facto political competition and protection of a basic set of individual rights in order to meet the standard criteria for “full” or “liberal” democracy (Dahl, 1971). Most scholars and practitioners implicitly or explicitly treat this more robust version of democracy as a key to normalizing politics and channeling social conflict.
Despite the ubiquity of democratic institutions after civil war, there is still wide variation in the de facto competitiveness of post-conflict politics. 2 While these institutions have been successful in promoting peaceful political competition in some cases, in others they have clearly failed, leading to renewed violence or the entrenchment of highly authoritarian regimes. In Sierra Leone and El Salvador, for example, the transition from war to electoral politics did actually seem to lay the foundations of a system of robust competition and electoral turnover. In cases such as Tajikistan, on the other hand, post-conflict elections quickly became a rubber stamp for the incumbent government, leaving the opposition little or no chance for success.
Why do countries emerging from conflict with democratic institutions exhibit such variation in their electoral competitiveness? This question is especially important in the post-conflict context because the failure of democratic institutions risks not only unaccountable government, but also a return to war. Citizens frustrated with political corruption may engage in destabilizing protests, or former rebel groups who come to realize that elections offer them no real prospect of governing may choose to again take up arms in an effort to secure a revised settlement. Incumbents may even invoke the preservation of peace to increase political repression where accountability is limited. Highly competitive elections may only rarely be in the interests of those already in power, but they are generally in the interests of society. When do such elections actually emerge?
I argue that certain types of capabilities developed by combatants during civil war are more useful than others in subsequent elections. Capabilities such as the ability to promote a populist ideology or mobilize support among ethnic networks are “convertible” from use on the battlefield to use on the campaign trail. Other capabilities, such as access to military hardware or technical combat skills, are more specific to conflict and less valuable in the transition from war to electoral politics. Whether post-conflict elections are competitive depends fundamentally on the number of actors who emerge from conflict having developed the first type of capabilities, which I call “convertible capabilities”. Specifically, rebel groups with convertible capabilities may be especially important. Often rooted in significant bases of support among the population, these former combatants will be able to more effectively challenge incumbents at the polls.
I test this theory using survival analysis of post-conflict episodes initiated between 1946 and 2007. The analysis provides evidence supporting the argument that not all combatant capabilities are alike in their influence on post-conflict politics. I find support for the hypothesis that rebel capabilities that are convertible between war and politics, such as the ability to mobilize support among the population or the maintenance of a distinct political wing, increase political competition (measured by electoral turnover) when the shooting stops, but that these effects are conditional on the number of actors involved and the length of the peace. Capabilities that are less convertible, such as a rebel group’s fighting capacity, are less influential.
Civil war and democratization
Much scholarship on civil war and subsequent democratization has focused on post-conflict stability. While some have argued that democracy provides a means of stabilizing peace after civil war (Wantchekon, 2004; Wantchekon and Neeman, 2002; Wood, 2000), others have found the connection to be more complex or even problematic (Flores and Nooruddin, 2009; Jarstad and Sisk, 2008; Lyons, 2005; Metternich, 2011; Paris, 2004; Reilly, 2008). More recently, scholars have begun to explore variation in the quality or success of post-conflict democratization (Hartzell and Hoddie, 2015). Some studies have turned to the mode of conflict termination (Gurses and Mason, 2008; Toft, 2010) or structural conditions (Fortna and Huang, 2012) to explain why long-term democratization succeeds, while others have focused on the role of the international community (Doyle and Sambanis, 2000). Despite the frequent intensity of international democracy promotion efforts, recent research suggests that post-conflict democratization depends at least as much on whether it serves the interests of local elites as it does on the efforts of international actors (Zürcher et al., 2013).
These studies have provided us with a valuable foundation for examining the complex relationship between conflict and democratization. Yet many of them look at democratization as a unitary, aggregate process. Building on this work, there is good reason to now focus on particular aspects of democracy and democratization. In addition to promoting general concreteness, this approach recognizes that not all aspects of democracy may go together in the wake of major conflict. Research has suggested, for example, that peacekeeping operations have different effects on civil liberties and political rights (Fortna, 2008). This approach also leverages the perspective gained from the literature on hybrid regimes, which highlights the ways that democratic institutions are often paired with authoritarian practices (Levitsky and Way, 2010), reducing the clarity of more aggregate approaches to democratization.
Accordingly, this study focuses on political competitiveness. While admittedly just one important aspect of post-conflict democratization, political competition is the mechanism that implicitly connects many theories of democracy and its outcomes. Political competition is believed to increase accountability for policy decisions (Gordon and Huber, 2007; Williams, 2013), motivate incumbents to respect civil liberties (Davenport, 2007), and (in some circumstances) protect minorities from communal violence as party leaders seek to maximize votes (Wilkinson, 2004: 137).
While these studies all identify the effects of political competitiveness on democratic politics, the literature has not yet produced a robust theory of the sources of political competitiveness in the post-conflict context. After all, violent conflict creates conditions under which actors can literally fight their way to a place on the ballot. How they are likely to perform once there, however, requires consideration of the sources of political competition and the ways in which conflict is likely to affect the electoral playing field.
The distribution of convertible capabilities
The literature on democratization and hybrid regimes has suggested that the emergence and survival of liberal democracy rely primarily on the inability of any one actor to monopolize power resources. Lucan Way, in particular, has argued that variation in the competitiveness of electoral regimes can be explained primarily by variation in the ability of actors to consolidate control over key political and economic resources (Way, 2005). Monopoly over the state, and particularly autonomous income such as resource wealth or some forms of foreign aid, enables incumbents to monopolize the political space and block out would-be competitors (Ahmed, 2012; Greene, 2010; Magaloni, 2006). Where the incumbent loses the monopoly on resources, liberalization is more likely to result.
In the post-conflict context, where institutions are so often weak or nonexistent, the distribution of combatants’ power resources during conflict should significantly affect the subsequent distribution of political power. Several scholars have emphasized the balance of power at the end of conflict as a key driver of post-conflict politics (Gurses and Mason, 2008; Toft, 2010; Wantchekon, 2004). Yet the relative capabilities of combatants are not determined by the same types of resources and organizations in all conflicts (Staniland, 2014; Weinstein, 2007). Some rebel organizations may rely on external sponsors and resources (Salehyan, 2009), while others develop deep ties and networks of support among the domestic population (Wood, 2003). Some groups may be strengthened by their superior knowledge of remote local terrain, while others draw on military hardware acquired via lootable resources.
Understanding post-conflict political competition thus requires attention to combatants’ capabilities, defined as the resources, skills and organizational structures used by a group to pursue its goals. We can distinguish between two broad types of combatant capabilities developed during conflict. Specific capabilities are those whose usefulness is largely limited to violence. Convertible capabilities are those which can be marshaled for use in various types of social mobilization. 3 Fighting skills, knowledge of remote terrain and military hardware are specific, whereas networks of ethnic support, a populist ideology, funding from a sympathetic diaspora or a highly developed political organization would be convertible capabilities.
Every combatant organization, whether state or rebel, has some mix of specific and convertible capabilities which are used to further that group’s political and military goals. During conflict, these various capabilities may be mutually reinforcing (as when co-ethnics in the diaspora provide funds for weapons purchases) or they may even be substitutes (as when rebels can make up for a lack of military strength by relying on superior intelligence from friendly populations). After conflict, however, specific capabilities are less useful in helping combatants achieve their ends via institutionalized political processes such as elections. 4 Under some circumstances, ex-combatants can force their opponent’s hand with threats of a renewed conflict, but this is distinct from success in adapting to new forms of competition. 5
Success in post-conflict elections thus depends on the convertible capabilities at a combatant organization’s disposal, assessed relative to other actors. Where only one actor has significant stocks of convertible capabilities, he will be able to dominate post-conflict elections and consolidate political power, even where he could not dominate the conflict militarily. Where multiple actors have electorally relevant capabilities, they will constrain one another’s ability to establish such dominance, leading to a sort of “pluralism by default” (Way, 2005: 232). This leads to the first hypothesis:
Hypothesis 1 focuses on the capabilities of rebel groups because the incumbent has the advantage of the state, which is the ultimate convertible resource where autonomous institutions are lacking, as they generally are in the post-conflict context. 6 Unless the distributions of convertible and non-convertible capabilities perfectly co-vary (which they do not), conflicts with a more equitable distribution of convertible capabilities between the government and rebels should result in more competitive politics, but the effect of specific capabilities should not be significant for electoral outcomes. Convertible capabilities may be useful in one of two ways. On the one hand, such capabilities may help rebels transform themselves organizationally into a political party. Where a rebel group already has members specializing in political activities, as with the separate military and political functions of the Irish Republican Army and Sinn Féin between 1919 and 1921, it has a set of leaders and officers already possessing some of the skills needed for organizing supporters and attracting votes. On the other hand, where a rebel group has depended on the local population for material support or has claimed to be fighting on behalf of a particular ethnic or religious group, it will have a core constituency of voters available to be mobilized at election time. Even if the rebel group was not organized to have an explicitly political wing, the existence of this core constituency may make it easier to build a competitive party after the fact.
Thus far, we have considered the distribution of convertible capabilities between an incumbent government and an opposing rebel group. Prior scholarship has taken a similar approach in studies of individual rebel groups’ transformation from rebels to political parties (Ishiyama and Widmeier 2013; Manning, 2008) and in post-conflict democracy (Huang, 2012). Yet 21% of the conflict episodes in the data (discussed below) involved more than one rebel group. The number of actors among whom resources are distributed should also matter for the translation of wartime capabilities into peacetime political strength. A bipolar distribution of capabilities should be highly competitive, but a wider distribution of capabilities may not be. If we assume that the individual or party in control of the state enjoys some incumbency advantage, several rebel groups with convertible capabilities may box one another out at election time. This divided opposition would dilute the benefits of developing convertible capabilities during conflict. This leads to a second hypothesis:
The literature on electoral authoritarianism suggests that the ability of an opposition to present a united front is essential to liberalizing the political system (Howard and Roessler, 2006). Where there are multiple opposition groups seeking access to the halls of power, incumbents can maintain political dominance by buying the support of some groups and excluding others (Lust-Okar, 2005; Weingast, 1997). Where there are many rebel groups, one actor (most likely the incumbent government) may be able to politically “conquer the divided” and (re)consolidate power in the wake of conflict. A bipolar distribution of convertible capabilities may thus be more conducive to post-conflict competition than a distribution that is more dispersed. 7
As presented, the theory is agnostic as to the outcome of the conflict. In one sense, its applicability is clearest in conflicts ended by ceasefire or negotiated settlement in which all combatant organizations have the option to make the transition to political parties. Cases of military victory may seem less straightforward. Where rebels win and become the incumbent power, they may be able to dominate the political system in ways similar to the Rwandan Patriotic Front or the Ethiopian People’s Revolutionary Democratic Front. Where the incumbent government defeats a rebel force, it may be able to establish similar dominance. Despite these possibilities, there is reason to think that the hypotheses presented may still be valid in cases of military victory. The individuals and networks that supported the defeated side do not generally disappear altogether, particularly if it developed the types of convertible capabilities discussed above. While a defeated government or rebel group may not make a wholesale transition into the post-conflict period, its convertible capabilities (if they exist) are not necessarily eliminated or co-opted altogether by the new government. A party or political wing affiliated with the losing side may be banned and its leadership exiled or executed, but this may still leave behind a body of operatives with skills and experience in political organization and mobilization. 8 Efforts at “national reconciliation” may be launched, but identities mobilized by conflict will probably remain politically salient cleavages. 9 A military defeat may delay the emergence of clearly competitive elections as supporters of the losing side reconstitute themselves into a viable political opposition, but the underlying process by which convertible capabilities might act as a counterweight to incumbent electoral advantage remains largely the same.
To summarize, prior research on democratization and civil war lays the foundations for a theory of post-conflict political competition. Scholars of democratization and electoral authoritarian regimes have found that the monopolization of political resources by incumbents and the division of opposition are crucial to incumbent survival over repeated election cycles. Studies of civil war processes indicate that combatants pursue their military and political goals via strategies based on widely varying resources and organizational capacity. Taken together, these bodies of research suggest two hypotheses: rebels with capabilities that are convertible from military to electoral competition should be better able to challenge incumbents in post-conflict elections; and bipolar distributions of such capabilities (between the incumbent and a single rebel group) should be more conducive to competition than other types of distributions.
Data and research design
The theory just presented applies to all post-conflict political systems. For purposes of the present study, I have defined a “post-conflict episode” as the period after an episode of intrastate violence, beginning in the last year of conflict (defined by the 25-deaths-per-year threshold coded by UCDP-PRIO) and ending in 2009, or the year that violence resumes. 10 The data includes all post-conflict episodes that began between 1946 and 2007 (N = 319), as defined by Kreutz (2010). The duration of post-conflict episodes in the data ranges from a minimum of 1 year to a maximum of 63 years, with a mean duration of 10.76 years. Turnover occurred in 27.27% of post-conflict episodes in the dataset. Among those episodes that experienced it, turnover occurred after an average of 9.36 years. Where incumbents manage to survive in power until the end of the observation period, they do so for an average of 19.09 years. 11
Dependent variable—post-conflict competitiveness 12
Prior studies of post-conflict democratization have tended to take democracy as a single, aggregate phenomenon. This study takes an alternative approach and instead focuses on one crucial element of democracy: political competition. Because of wide variation in institutional form, however, measuring competition cross-nationally is challenging. The margin of victory in an executive election and the distribution of seats in a legislature are not equivalent measures across electoral systems. To measure the level of political competition in a system, I thus turn to a more universal, if less precise, measure: whether or not a regime meets Przeworski’s criteria that “Democracy is a system in which parties lose elections” (Przeworski, 1991: 10). Consistent with the perspective highlighted earlier—in which democracy can be seen as the failure of would-be authoritarians to consolidate power—the following analysis examines incumbents’ ability to survive in office after a conflict has ended. 13 Although political competition may be difficult to measure cross-nationally, turnover is a clear indication of competitive elections. 14 Accordingly, the models in Tables 1 and 2 use survival analysis and as the dependent variable whether or not the incumbent party loses a national election during the post-conflict episode. 15 The emphasis here is on the party and not the individual incumbent. While heads of government may change relatively frequently, many electoral regimes come to be dominated by one party for periods that long outlast the professional life of any individual leader.
Electoral turnover (Cox proportional hazards models)
Robust standard errors in parentheses, clustered by country.
p<0.01, **p<0.05, *p<0.1.
Models all use a Cox proportional hazards model estimated using Stata and report hazard ratios.
The natural logs of GDP per Capita, Aid per Capita, Conflict Length and Battle Deaths are used.
Electoral turnover (Cox proportional hazards models)
Robust standard errors in parentheses, clustered by country.
p<0.01, **p<0.05, *p<0.1.
Models all use a Cox proportional hazards model estimated using Stata and report hazard ratios.
The natural logs of GDP per Capita, Aid per Capita, Conflict Length and Battle Deaths are used.
Independent variables—combatant capabilities
The hypotheses presented above suggest that the level of political competition faced by incumbents after conflict should be influenced by the development of various capabilities by rebel groups and the number of actors involved in conflict. The key independent variables, measured at the level of the post-conflict episode, indicate the number of rebel groups who develop high levels of certain convertible and specific capabilities during the preceding conflict period. Measures for each of these are included in the model as discussed below.
Testing the hypotheses requires that we identify rebel capabilities in a way that both gives a general sense of their strength and distinguishes between convertible and specific capabilities. To reiterate, the theory suggests that all participants in violent conflict utilize some mix of various resources and abilities. Some of these, such as combat skills or arms, are primarily or even uniquely useful in the prosecution of violent conflict. Others, which I have labeled convertible capabilities, are those which a former combatant may be able to transfer from the battlefield to the electoral arena.
Data available via the Non-State Actor (NSA) Dataset allows us to distinguish between different types of capabilities enjoyed by rebel groups, providing measures identifying their various capacities and organizational features (Cunningham et al., 2009). 16 In particular, the NSA data provides general measures for fighting capacity and mobilization capacity for each rebel group, as well as an indicator of whether the group had a separate political wing, which I take as a basic measure of political organization. 17 The first is a capability specific to conflict, while the latter two are more convertible to electoral competition. The NSA data uses categorical coding for fighting capacity, which is qualitatively assessed based on a group’s ability to present a military challenge to the incumbent government’s forces. Mobilization capacity is similarly assessed based on a group’s ability to garner support from the relevant population, serving as a generalized estimation of its ability to recruit fighters and rally support among civilians. 18
It is particularly helpful that the NSA data enables us to measure both political organization and popular mobilization capacity as distinct features of rebel groups. While both are convertible capabilities, the former provides an indication of a rebel group’s ability to re-orient its organization from military to electoral struggles, while the latter provides a measure of how robust that group’s core support base is likely to be. Nor do the two necessarily go hand-in-hand. For example, the Moro National Liberation Front in the Philippines is coded as having high mobilization capacity among Muslims in the country between 1972 and 1980, but lacked a political wing during this period. This distinction allows us to discern the various pathways by which convertible capabilities might be deployed after conflict: as a foundation for political party-building, a network for voter mobilization, or both. If convertible capabilities are indeed crucial for post-conflict political competition, distinguishing among the types of these capabilities will help us begin to understand how.
To test the first hypothesis—that conflicts in which any rebel group develops convertible capabilities are more likely to be followed by competitive elections—I use dichotomous indicators for the three capabilities, measured at the level of the conflict episode. Fighting Capacity indicates that at least one rebel group involved in the conflict demonstrated moderate or high fighting abilities vis-à-vis the government. Political Wing indicates whether any rebel group developed a political wing during the conflict. Mobilization Capacity indicates whether at least one rebel group involved in the conflict episode could be judged to have a moderate or high capacity for mobilizing support among the population, again relative to the government. If the first implication of the theory is correct, then Political Wing and Mobilization Capacity should be associated with an increased likelihood of electoral turnover, while Fighting Capacity should not be significant.
Yet not all conflicts are between the state and a single rebel group, and 21% of the peace spells in the dataset followed conflict episodes in which more than one rebel group was active. The second hypothesis is that the overall distribution of convertible capabilities should matter for post-conflict political competition, with bipolar distributions (between a single strong rebel group and the incumbent government) being most strongly predictive of competitive elections. Accordingly, conflict episodes can be thought of as marked by one, multiple or no rebel groups with fighting capacity, a political wing or mobilization capacity. Splitting these categorical variables into dichotomous indicators results in nine main variables of interest: NOfight, ONEfight, MULTIfight, NOpolwing ONEpolwing, MULTIpolwing, NOmobilize, ONEmobilize and MULTImobilize, which are included in the models in Table 2. This approach leaves the reference categories in model 6 of Table 2 as conflicts in which no rebel group has the relevant capability. Model 7 in Table 2 resets the reference category for the mobilization and political wing indicators. 19 If Hypothesis 2 is correct, both ONEmobilize and MULTImobilize should be positively associated with electoral turnover, compared with NOmobilize, and the positive effect of ONEmobilize should be greater than MULTImobilize. The same should be true of ONEpolwing and MULTIpolwing. In order to distinguish between the effect of multiple mobilizers or multiple political wings and that of simply having multiple rebel groups in a conflict episode, a dummy variable for conflicts with multiple rebel groups (Multiple Rebels) is included in all models in Table 2.
Control variables
In addition to the main independent variables of interest above, I include two sets of control variables in the main models. Both sets of variables are included because they have been found to influence the prospects for the overall quality of democracy, to which political competition is closely related. The first set of control variables relates to characteristics of the conflict, which have received a great deal of attention in the literature on post-conflict democratization, as well as research on post-conflict stability. The models thus include variables for the natural log of the length and battle-related deaths of the preceding conflict. 20 Conflicts involving multiple rebel groups have been found to be more difficult for governments to win (Akcinaroglu, 2012), for example, and so the inclusion of these controls allows us to distinguish between the effects of the conflict itself and the effects of the characteristics of the non-state actors involved. 21 Other research has indicated that international involvement can also influence the prospects for a more democratic peace (Doyle and Sambanis, 2006), although only multidimensional missions are believed to have this effect. The final models thus also control for United Nations involvement, distinguishing between traditional and multidimensional UN missions along the lines set forth in Doyle and Sambanis (2006). 22 Because violence may end and resume multiple times within a single conflict, the measure Prior Episodes counts the number of prior peace episodes within a given conflict.
In addition to these, I also include control variables for the economic and contextual aspects of democratization more generally. Research on democratization in the post-conflict context and elsewhere has emphasized the importance of economic variables. Higher levels of economic development are associated with democratization in the post-conflict context (Fortna and Huang, 2012) and more generally (Przeworski et al., 2000). In addition to overall levels of development, the structure of the economy is believed to be important, with high levels of natural resource dependence being considered by many to inhibit democratization under many circumstances (Dunning, 2008; Ross, 2001). Access to autonomous sources of income (such as revenue from natural resources) is also found to enable incumbents in electoral systems to build and maintain dominance (Greene, 2010; Magaloni, 2006), bearing even more directly on the question of electoral competition. Foreign aid, although sometimes used effectively by incumbent governments for its stated purposes (Girod, 2012), has also been suggested to serve a similar function of power consolidation (Ahmed, 2012). Accordingly, for each year of the post-conflict episode, the full models all include a log-transformed measure of per capita gross domestic product (GDP), a measure of natural resource rents as a percentage of GDP, and a log-transformed measure of the per capita net development assistance received by a country. 23 Because the end of the Cold War is widely viewed as a watershed moment in the recent history of democratization, I include a variable indicating whether the preceding conflict episode ended after 1988. Finally, the full models include a dummy variable indicating whether or not the regime qualified as a democracy in the year prior to the onset of conflict in order to distinguish cases with a pre-conflict history of electoral turnover, such as India, from those without, such as Angola. 24
Estimation strategy
We can think of the primary question at the core of this enquiry as whether or not incumbents who have come through the experience of civil war are able to retain their position and stay in office. Our theoretical orientation is thus well suited to survival analysis. We do not have strong theoretical priors about the form that the baseline hazard takes over time, so a Cox proportional hazards model is appropriate because it makes no assumptions in this regard. The choice of survival analysis is appropriate to account for the role that time plays, but care must also be taken in the model specification. Standard Cox models assume that hazards are proportional. In other words, the effect of covariates on survival is assumed to be consistent over time. However, we might expect the impact of conflict-era variables to change over time, so the possibility of non-proportional hazards must be considered. 25 Testing for this indicates that the effect of Political Wing, ONEpolwing, and MULTImobilize are indeed time-dependent. Accordingly, the interactions of these variables with the natural log of time are included in the final models. 26
Analysis and results
The results of the survival models are presented in Tables 1 and 2. The results in Table 1 utilize the independent variables indicating whether any rebel group in a conflict episode displays high fighting or mobilization capacity or develops a political wing. This tests Hypothesis 1. All models in the table report hazard ratios, so that ratios over 1 indicate an increase in the likelihood of electoral turnover, while ratios less than 1 indicate a lower risk. The results in Table 1 suggest that conflicts involving rebel groups with high levels of fighting capacity are less likely to result in electoral turnover than those whose rebels are militarily weak. This finding, however, disappears both substantively and in its statistical significance when we control for other relevant characteristics of the conflict and post-conflict periods. More consistent is the finding that the presence of at least one rebel group with a political wing slightly decreases the prospects for electoral turnover in the short run, but significantly increases the likelihood of turnover as time goes on. The same effect does not hold for mobilization capacity.
While Table 1 shows the effects of the presence of various rebel capabilities, Table 2 shows the effects of the distribution of capabilities, distinguishing among conflict episodes featuring no, one or multiple rebel groups with a political wing or mobilization capacity. Model 4 includes only the capabilities variables and Multiple Rebels, model 5 introduces the interaction with ln(t) for the relevant variables, model 6 includes the full battery of controls and model 7 changes the reference categories for the Political Wing and Mobilization Capacity variables. 27 This tests Hypothesis 2, which suggested that the positive relationship between convertible rebel capabilities and electoral competition should be strongest where these capabilities are concentrated with only one rebel group and decline as they are distributed across a larger number of rebel organizations.
The presence of rebel political wings improves the long-term prospects for electoral competition, but the results in Table 2 indicate that this is largely being driven by conflicts in which there is only a single rebel group with a political wing. Consistent with the theory, ONEpolwing has significant effects on the prospects for turnover, both relative to conflict with no rebel political wings (model 6), and when the reference category is set as MULTIpolwing (model 7). Regardless of whether a conflict includes multiple rebel groups or only one, long-term competition in subsequent elections will be highest when only one group developed political organizational capacity during the preceding conflict episode.
Figure 1 shows the predicted survival rates for two hypothetical incumbents. In both cases, the incumbent has survived a conflict episode against a single rebel group with low mobilization capacity and low fighting capacity. All controls are set at zero. The first, baseline scenario reflects an incumbent who has survived a conflict episode against a rebel group that has no political wing. For such an incumbent, the likelihood of survival in office is quite high. The second scenario, however, is one in which the rebel group did develop a political wing during the conflict. As is evident from the figure, the development of this convertible organizational capacity during the conflict episode has major ramifications for the incumbent’s prospects in subsequent elections. While conflicts with politically institutionalized rebel groups are slightly less likely to lead to incumbent electoral defeat in the short term, after 5 years there is a negative impact on his ability to retain office. Twenty-one years after surviving a conflict against a politically institutionalized rebel group, the incumbent’s survival chances slip below 50%, while they remain near 75% for the incumbent who survived conflict against a rebel group that did not have a political wing. Political organizational capacity, a highly convertible asset, is thus clearly valuable to rebel groups who go on to compete in post-conflict elections. Consistent with the theory, a bipolar distribution of this capability (between the incumbent government and a single rebel organization) is most likely to result in highly competitive elections.
Incumbent survival after conflict with a single rebel group: rebel political wings.
Mobilization capacity, however, does not have quite the same effect. Instead of a bipolar distribution of convertible capabilities being most likely to lead to turnover in the long run, conflict episodes featuring only one rebel group with high mobilization capacity are least likely to lead to competitive elections (as defined by turnover). A wide distribution of mobilization capacity among many groups appears to be most likely to lead to incumbent turnover in the long run. Figure 2 shows the prospects of post-conflict political survival for three hypothetical incumbents, differing only in whether they survived a conflict episode featuring no, one or multiple rebel groups with high mobilization capacity. In the initial post-conflict years, there is almost no difference between “multiple-mobilizer” and “single-mobilizer” episodes, but the difference becomes significant as time goes on, with the more fractured field having a greater chance of turnover. A similar pattern can be seen between no-mobilizer and multiple-mobilizer episodes. This surprising finding is discussed in the following section.
Incumbent survival after multi-actor conflict episodes: rebel mobilization capacity.
The control variables all behave largely in the manner expected. Electoral turnover became more likely with the end of the Cold War and, not surprisingly, countries in which democracy was established prior to conflict are more likely to see electoral turnover subsequently. Economically, overall wealth increases the likelihood of electoral turnover, while incumbents in countries that are heavily dependent on rents from natural resources are slightly more likely to survive. Interestingly, foreign aid also reduces the prospects of electoral turnover and facilitates incumbent survival. This is consistent with prior research suggesting that aid can be misdirected by incumbents in some (although not all) circumstances (Bearce and Tirone, 2010; Girod, 2012). As for the controls related to the characteristics of the conflict, United Nations intervention has a positive effect on electoral competition, regardless of whether it is traditional or multidimensional. Longer conflicts are also positively associated with turnover, while there is limited evidence from models 6 and 7 that more violent conflict episodes dampen the political competition that would lead to turnover.
Discussion
The relative irrelevance of fighting capacity
Consistent with the theory’s expectations, the distribution of fighting capabilities during wartime had no consistent, statistically significant effect on the likelihood of electoral turnover after conflict. While both rebel organizational capacity and rebel mobilization capacity increased the likelihood of turnover in some way, rebel fighting capacity did not. At most, high rebel fighting capacity may reduce the prospects for turnover, but this effect failed to reach significance once the control variables were added. Military success provides no guarantee of subsequent electoral success. Once conflict has ended, rebel groups who specialized solely in the use of violence may find that their reputations have been relatively fixed among voters and that their members lack the skills needed to campaign or rally voters, leaving them at a distinct disadvantage compared with the incumbent. Incumbents in control of the state by necessity nearly always have some level of organizational and mobilization capacity, and they carry this onto the post-conflict political playing field. Compared with a rebel group scrambling to build a political party after conflict ends, the incumbent effectively has a head start.
Convertible capabilities: organization and mobilization
The results provide support for the notion that the distribution of convertible capabilities during civil war influences the competitiveness of post-conflict elections. Precisely how convertible capabilities affect post-conflict political competition is less straightforward. The results indicate that building political organizational capacity may be one of the best investments that a rebel group can make with an eye to the post-conflict future. Incumbent turnover via elections is significantly more likely when rebels institutionalize politically and establish a distinct political wing during a civil war. This is particularly true where the conflict episode involves a single politically institutionalized rebel group. Students of democratization should find it as no surprise that an organized, unified opposition is more likely to unseat an incumbent.
While the political wing variables behave largely as the theory would predict, mobilization capacity seems to play a different role. The results indicate that having the opposition’s mobilization capacity concentrated with a single rebel group is actually the condition least conducive to incumbent turnover after civil war. Wider distributions of mobilization capacity, as when multiple rebel groups enjoy high levels of this capacity, are far more conducive to electoral competition once conflict ends. The more fractured the field of voters, it seems, the less likely it is that an incumbent will be able to (re)consolidate his position in the face of electoral challenges.
This finding appears to challenge the theory’s implication that competition should be most robust under bipolar distributions of convertible capabilities. A unified opposition should present a particularly strong electoral challenge to incumbents, while a more fractured opposition may offer opportunities for “divide and conquer” strategies by the incumbent. Particularly given the findings related to rebel groups with political wings, the relative insignificance of unified mobilization capacity is surprising. Why is it that bipolar distributions of organizational capacity facilitate post-conflict electoral competition, while bipolar distributions of mobilization capacity do not have a significant effect? The first possibility is that unified rebel organizations with high levels of mobilization capacity are more likely to win a military victory. While some research has suggested that this type of conflict termination is most likely to lead to long-term democratization (Toft, 2010), prominent cases of rebel victory such as in Ethiopia in 1991 and Rwanda in 1994 suggest that rebel groups who win victory are particularly capable of establishing political systems dominated by a single party. While there is strong case evidence for this argument, dominant rebel victors are not driving the cross-national results, which are robust even when cases of rebel victory are dropped from the sample. Instead, two mechanisms are likely to be behind the superior importance of organizational over mobilization capacity.
First, the modeling strategy adopted in this chapter makes the assumption that the incumbent government has both organizational and mobilization capacity. A fractured opposition is thus assumed to be facing off against a unified, capable incumbent who should be able to dominate the electoral field. Yet one of the most prominent effects of civil war is that it generates an enormous strain on states and the incumbents who rule them. Conflicts with multiple mobilizing rebel groups may exacerbate this strain. A highly fractured field may offer opportunities for new coalitions or new parties where the incumbent’s capacity is limited. While the available data requires that we assume a certain level of incumbent capacity, qualitative analysis might allow us to relax that assumption. The complete wartime collapse of the state and repeated military coups d’état in Sierra Leone, for example, strain our traditional notions of incumbent advantage when considering conditions at the end of the 1991–2002 conflict.
The second possible explanation is that organizational capacity simply matters more than mobilization capacity in the long run. While parties with an established base of support among voters have an advantage over those that do not, organizational capacity is crucial in expanding that base, getting voters to the polls, preventing leadership splits and reclaiming support from voters who may have moved away from the party. High mobilization capacity is not sufficient, although a combination of high mobilization capacity and high organizational capacity is likely to lead to electoral success. An equitable distribution of mobilization capacity may not sustain political competition over time where there is not also an equitable distribution of organizational capacity. Conversely, organizational capacity may be adequate to sustain a party though “dry spells” in which mobilization proves difficult. Future research may be able to leverage alternative approaches for testing this explanation.
The long-run effects of conflict
The effect of time on the results is also surprising. Across the models, the difference between conflicts with and without convertible capabilities is quite minor in the early years after the war. We only begin to see a significant difference in turnover rates after about 5 years of peace. This is counterintuitive if we generally expect legacies to fade over time. One plausible explanation is that the substantive effect of rebel capability is simply overwhelmed by other factors in the early, fragile years of peace. Foreign aid flowing in to assist with reconstruction may tip the scales in favor of the incumbent in a way that cannot be immediately combated by rebels-turned-opposition. The finding related to aid supports this explanation. Power-sharing agreements may also freeze a particular configuration of political offices for the immediate post-war period. It may be that the instability of those early years turns what should be immediate effects into latent ones, as concern for the stability of peace overwhelms any other interests.
While the effects of convertible capabilities on electoral turnover emerge in the longer term, the findings are limited in their implications for the trajectory of a political system once turnover has occurred. As noted earlier, electoral competition is a necessary but not sufficient component of democratization and it is difficult to assess whether convertible capabilities reduce the risk of autocratization once the opposition achieves electoral victory. It may be that they do. A party that comes to power relying on the types of convertible capabilities addressed here may be more confident in its competitiveness in future elections, reducing the incentive to secure its position by engaging in autocratization and facilitating other processes of democratization. 28 Moreover, if turnover occurs in cases where the overall distribution of convertible capabilities is indeed equitable, then the likelihood of autocratization may be further reduced. Just as politically organized rebels or a highly destructive conflict limit the incumbent’s monopoly on power, an organized and mobilizing incumbent party may provide a similar counterweight should it find itself newly in opposition. 29 Limitations in the data do not allow rigorous investigation of these possibilities at present, but the issue does provide a potentially fruitful avenue for future research.
Conclusions
The findings above indicate that some incumbents who survive military challenges are more able than others to survive subsequent electoral challenges. Where conflict features one rebel group with an established political wing or multiple groups with networks of popular support, incumbents are less likely to be able to retain their position over the long term. In addition to being substantively important in their own right, the findings have several implications for both scholarship and policy. First, they highlight the importance of moving away from analyses that make use of aggregate variables such as “democracy” or “balance of power”. Instead, theory-building and theory-testing should rest on more concrete conceptualization and operationalization. By focusing on one key component of post-conflict democratization (electoral turnover), this study contributes to the general movement towards this type of theory in the field. Similarly, the findings suggest that qualitative differences in the capabilities of conflict combatants have effects on post-conflict political outcomes.
This analysis has practical implications for both former rebels and the international community. Rebels who develop a separate political wing during conflict are better able to maintain high levels of electoral competition during peacetime than those who build more informal support networks among the population. This institutionalization offers rebels an opportunity not only to connect with civilians, but also to have members who specialize in pursuing their agendas in non-military ways. Political organization is thus the key to subsequent electoral effectiveness. The conclusion is not, of course, to suggest that rebel groups should forego developing the capacity to mobilize popular support during conflict. Particularly where there are multiple non-state actors engaged in conflict, there is a significant electoral advantage to more than one of them developing this capacity. Moreover, it is likely that the consequences of the war itself will be lower in terms of lives lost and civilians abused where rebel groups are actively seeking to mobilize popular support.
For the international community, the findings suggest that one-size-fits-all programs to support post-conflict democratization are unlikely to be effective. While a turn toward tailored solutions is already clearly underway at the major institutions involved in facilitating conflict recovery, the above analysis suggests that it is a trend that should be strengthened. Equally important, however, is the longevity of democracy support. Assisting former combatants in the transition from armed rebel group to political party may require more than a brief commitment to demobilization and the provision of limited resources for party leadership. International organizations and non-governmental organizations may need to consider more comprehensive programs of support for political party-building, even where former rebel groups enjoyed a significant degree of popular support and even as these organizations navigate the line between democracy support and interference in domestic politics.
This is important not only for the development and preservation of democratic legitimacy, but also for providing opposition with a motive to remain inside the system. Former combatants who find themselves perpetually shut out of power may decide that their only hope of effecting change is a return to arms. Establishing and preserving de facto political competition, and not just electoral institutions, is vital to post-conflict stability. This is not only because it is a democratic value in its own right, but also because it provides incentives for key elites to preserve the peace.
Footnotes
Appendix
Variable descriptions and sources
| Variable | Source | Description |
|---|---|---|
| Electoral Turnover | Hyde and Marinov (2012) | Coded from the variable nelda24, indicating whether or not the incumbent party lost the election. NELDA codes both legislative and executive elections, as well as constituent assembly elections. The data here code for executive elections, unless no executive elections are held (as in a Westminster-style parliamentary democracy), in which case legislative elections are coded. Failure is coded as the first election in or after the last year of conflict in which the incumbent party loses. |
| Multiple Rebels | Cunningham et al. (2009) | Dummy variable identifying whether multiple non-state actors were active in the preceding conflict episode. |
| Political Wing | Cunningham et al. (2009) | Coded from the variable rebelpolwing. Indicates whether or not rebel groups involved in the conflict had an established political wing. |
| ONEpolwing | ||
| MULTIpolwing | ||
| NOpolwing | ||
| Mobilization Capacity | Cunningham et al. (2009) | Coded from the variable mobcap. The original variable had three categories (low, moderate, and high). For purposes of this analysis that is broken into a dummy variable for moderate and high mobilization capacity, leaving low capacity as the reference category. ONEmobilize indicates that only one non-state actor active in the conflict had moderate/high mobilization capacity. MULTImobilize indicates that more than one non-state actor active in the conflict had moderate/high mobilization capacity. NOmobilize indicates that no non-state actors active in the conflict had moderate/high mobilization capacity. |
| ONEmobilize MULTImobilize | ||
| NOmobilize | ||
| Fighting Capacity | Cunningham et al. (2009) | Coded from the variable fightcap. The original variable had three categories (low, moderate, and high). For purposes of this analysis that is broken into a dummy variable for moderate and high fighting capacity, leaving low capacity as the reference category. ONEfight indicates that only one non-state actor active in the conflict had moderate/high fighting capacity. MULTIfight indicates that more than one non-state actor active in the conflict had moderate/high fighting capacity. |
| ONEfight | ||
| MULTIfight | ||
| UN—Traditional, | Doyle and Sambanis (2006), Stockholm International Peace Research Institute (n.d.) | Coding for UN missions is taken from Doyle and Sambanis’s unmandate variable and supplemented with data from SIPRI. The dummy variable UN-Traditional indicates that there was an observer mission or traditional peacekeeping mission present at the end of the conflict episode. The UN-Robust variable indicates that there was a multidimensional peacekeeping mission, enforcement mission, or transitional administration in place. For SIPRI supplemental data, the Doyle and Sambanis rules were approximated using Chapter VI missions as traditional missions and Chapter VII missions, or Chapter VI missions with two dimensions beyond the provision of protection, as robust missions. |
| UN—Multidimensional | ||
| Prior Democracy | Cheibub et al. (2010) | A dummy variable indicating whether or not a regime qualified as a democracy by Cheibub, Gandhi, and Vreeland’s criteria for their democracy variable in the year before conflict onset. |
| GDP per capita | World Development Indicators (World Bank, 2013) | Gross Domestic Product, per capita. Constant 2005 US Dollars. Natural log transformed. |
| Aid per capita | World Development Indicators (World Bank, 2013) | Net development aid and assistance per capita. Constant 2011 US Dollars. Natural log transformed. |
| Rents | World Development Indicators (World Bank, 2013) | Rents from natural resources as a percentage of GDP. |
| Conflict Length | Kreutz (2010) | The length of the preceding conflict episode in years. Natural log transformed. |
| Battle Deaths | Lacina and Gleditsch (2005) | The number of battle-related deaths during the previous conflict episode. The measure uses the bdeadhigh estimate from the data in order to minimize missing data. |
| Prior Episodes | — | This is a count variable of the number of prior episodes of the same conflict. |
Acknowledgements
Thanks to Marc Morjé Howard, Desha Girod, Thomas Flores, Lars Svasand, Stefan Malthaner, Anjali Dayal, Michael Weintraub, Megan Stewart, and several anonymous reviewers for comments and suggestions. All errors remain those of the author.
Authors’ note
Earlier versions of this project were presented at the annual meetings of the Midwest Political Science Association, the American Political Science Association, and the International Studies Association.
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.