ADVANCING THE VETO-PLAYER FRAMEWORK

(1)

ADVANCING THE VETO-PLAYER FRAMEWORK

A STUDY OF THE CONDITIONS UNDER WHICH FRAGMENTATION INFLUENCES THE LIKELIHOOD OF A PEACE AGREEMENT

ALANNA SMART Master's Thesis

Spring 2021

Department of Peace and Conflict Research, Uppsala University Supervisor: Isak Svensson

Word Count: 22999

(2)

1

Abstract

Recent research has recognised the complexity of fragmented conflicts; however, debate exists regarding when and how fragmentation hinders peace processes. This study addresses this lacuna, by investigating the conditions under which fragmentation affects the likelihood of a peace agreement.

Advancing Cunningham’s (2011) veto-player framework with insights from political parties’ literature, creating a novel causal mechanism, I argue that two conditions affect how fragmentation influences the likelihood of a peace agreement: veto players and outbidding. As veto players can ‘veto’ a peace agreement, a reduction in their number increases the likelihood of a settlement. Where fragmentation increases veto-players, I hypothesise that whether a peace agreement is likely or not, depends on the level of preference divergence, which is dependent on their engagement in outbidding. The hypotheses are tested through a qualitative structured-focused comparison approach, examining the cases of the Philippines (1990-1996), Uganda (1986-1988), Ethiopia (1973-1983), and Afghanistan (1980-1990).

Overall, this thesis finds mixed support. Limited support is found for the hypotheses, with half of the cases concurring with expectations. However, the cases also reveal significant support for the mechanisms, although with qualifications, suggesting further refinement is required.

Key words: armed conflict, fragmentation, rebel groups, veto players, outbidding, preferences.

Acknowledgments

First, I would like to thank my supervisor, Isak Svensson, for providing me with guidance and support in my endeavors. I am also extremely grateful for the unwavering support of my parents, Chris and Mas, without whom this would not have been possible. Lastly, I would like to extend thanks to my friends, in Uppsala and abroad, whom have shared with me their knowledge, sympathies, and joy.

(3)

2

List of Abbreviations

ANLF Afghanistan National Liberation Front

CRA Coordination de la Résistance Armée

ELF Eritrean Liberation Front

EPLF Eritrean People's Liberation Army

FLAA Front de libération de l'Aïr et de l'Azawagh FORGE Foundations of Rebel Group Emergence Dataset Harakat-i Inqilab Movement of the Islamic Revolution

HIG Islamic Party of Afghanistan

HIG-Khalis Islamic Party of Afghanistan - Khalis faction

HSM Holy Spirit Movement

IGO Inter-Governmental Organisation

ISA Islamic Society of Afghanistan

MHS Mutually Hurting Stalemate

MILF Moro Islamic Liberation Front

MILF Moro National Liberation Front

NGO Non-Governmental Organisation

NIFA National Islamic Front of Afghanistan

UCDP Uppsala Conflict Data Programme

UCDP/PRIO-ACD Uppsala Conflict Data Programme/Peace Research Institute Oslo – Armed Conflict Dataset

UPDA Uganda People’s Democratic Army

(6)

5

List of Figures

Figure 1. Veto Player Causal Mechanism………...17

Figure 2. Necessary Constitutive Conditions of a Veto Player………....18

Figure 3. Outbidding/Moderating Causal Mechanism………....23

Figure 4. Fragmentation - Peace Agreement: Conditional Causal Mechanism………...…….24

Figure 5. Case Selection………...……...28

Figure 6. Rebel Groups………..………34

Figure 7. Case Empirics……….……….……62

(7)

6

I. Introduction

Conflicts are seldom fought between only two parties. Approximately 44% of armed opposition movements since 1989 have been fragmented (Walther & Pederson, 2020: 446). These conflicts are inherently unstable, as cross-cutting rivalries and alliances form, shift, and collapse with astounding fluidity. These trends are not new, for example, in Iraq, the Patriotic Union of Kurdistan and Kurdistan Democratic Party fought as bitter rivals until the mid-1980’s when the two began cooperating to the surprise of many (Akcinaroglu, 2012:889). In 1979, the alliance between the Chadian Forces of the North and Popular Armed Forces broke down to lethal consequence (ibid.).

Such instability complicates conflicts, introducing uncertainty about actor landscapes in war.

Reflecting these trends, in the last 15 years researchers have moved beyond dyadic analyses of war towards complex multiple-actor studies. These have highlighted the significant effect of fragmentation on conflict dynamics, frequently emphasising its exacerbating effect, such as increasing inter-group violence and civilian targeting, whilst making conflicts longer and harder to resolve. However, some research suggests there is variation in the degree to which fragmentation inhibits conflict resolution, and may even encourage it (Findley & Rudloff, 2012). Given that only a quarter of civil wars end in negotiated settlements (Cunningham, 2011:27), understanding the conditions under which fragmentation influences the likelihood of an agreement is central to improving their prospects.

This thesis addresses this lacuna by asking under what conditions does fragmentation affect the likelihood of a peace agreement? In doing so, this paper aims to reveal mechanisms that affect the impact of fragmentation and assess the relevance of existing frameworks in explaining variation in the prospects of peace agreements.

Combining insights from negotiations, fragmentation, and political parties’ literature, I create a novel causal mechanism and three hypotheses. Drawing on Cunningham’s (2011) veto-player framework, a reduction in the number of veto players is hypothesized to increase the likelihood of a peace agreement. Where the number of veto-players increases, I argue that the likelihood of a peace agreement depends on the presence of outbidding. Where outbidding takes place, it is hypothesised preferences will diverge to a greater extent, further reducing the bargaining space and the likelihood

(8)

7 of a peace agreement. Where the number of veto-players increases, but outbidding does not occur, it is hypothesised that more moderate rebels will likely be co-opted into peace agreements.

To test this, a qualitative structured-focused comparative approach is used to analyse four cases, the Philippines (1990-1996), Uganda (1986-1988), Ethiopia (1873-1983), and Afghanistan (1980-1990).

This approach allows deep exploration of the theorised mechanisms. The analysis draws on primary sources, including peace agreements and rebel statements, and secondary sources, such as NGO, think-tank, and academic research.

Overall, this thesis finds mixed support. Whilst the hypotheses find limited support, with half of the cases concurring with expectations, an examination of the causal mechanisms within each case finds substantial support for both the veto-player and outbidding mechanisms, although with some qualifications. Consequently, this thesis finds that the mechanisms are credible and relevant but require further refinement.

The findings contribute to the field by augmenting our understanding of variation in the link between fragmentation and peace agreements. Furthermore, it assesses and confirms the relevance of Cunningham’s (2011) veto player framework, whilst advancing his theory with a credible outbidding mechanism. Regarding policy implications, the findings suggest that developing policies that prevent outbidding may increase the likelihood of successful negotiations.

This paper outlines the theoretical framework, drawing on negotiations, fragmentation, and political parties’ literature to posit a novel causal mechanism and hypotheses. Next, the research design is elucidated, and the empirics are presented. Then, the cases are analysed, and cross-case conclusions are drawn. Limitations, biases, and alternative explanations are then discussed. Finally, this paper concludes with a summary of key findings, implications, and areas for future research.

(9)

8

II. Theoretical Framework

Previous Research

Negotiations and Conflict Termination

Rationalist conceptions of war suggest that the choice to enter, continue, or renew conflict arises from bargaining failure. Bargaining can be tacit, pre-emptive behaviour in lieu of direct communication, or explicit, such as negotiations (Schelling, 1960). Crucially, whether bargaining is successful depends on belligerent’s abilities to find a solution that is acceptable to all parties. This solution is found in what is called the ‘bargaining space’. The size of the bargaining space rests between the value of an expected outcome and the perceived utility of hostilities for each group (Fearon, 1995). The larger the value of each, the smaller the bargaining space, and the less likely a settlement. Failing to find an acceptable bargain, the parties shall turn to war.

Accepting the assumption that both the onset and continuation of conflict arises from bargaining failure, existing approaches generally fall within four categories: economic theories, informational theories, commitment theories, and theories on indivisible issues. Economic theories of war suggest that violence is driven by opportunistic greed and ‘loot-seeking rebellions’, wherein rebels extort civilians and local resources for profit (Collier, 2000:841; Grossman, 1999). This occurs because the war economy presents warlords with a more profitable avenue than peace (Collier & Hoeffler, 2004;

Keen, 2000:22). Supporting this, the presence of valuable extractable resources, such as diamonds and oil, has been found to correlate with more intractable conflicts (Fearon, 2004). Following this logic, wars which are costlier are more likely to be resolved, as they are less profitable (Collier & Hoeffler, 2004). Consequently, increasing the costs of war, by increasing hostilities, or reducing its benefits, for instance through embargoes, can increase chances of a settlement.

A second cause of bargaining failure is imperfect information (Fearon, 1995; Filson & Werner, 2002;

Slantchev, 2003; Smith & Stam, 2004). This theory proposes that if belligerents had complete information regarding the utility of war, they would always choose to settle. Imperfect information, resulting in miscalculations regarding relative capabilities and opposition resolve, precede breakdowns in bargaining, as they lead to over-optimism regarding chances of victory (Fearon, 1995). This is compounded by incentives to withhold or misrepresent information in order to manipulate rivals’

(10)

9 calculations, to compel concessions or deter challengers (Fearon, 1995; Lawler & Ford, 1995).

Consequently, even the most rational leaders might unwittingly miscalculate the utility of war and avoid settlement in favour of hostilities. The revealing of information, through fighting or negotiations, allows belligerents to reach agreements in future (Filson & Werner, 2002)

Third, commitment problems hinder agreements as belligerents may be unable to credibly commit to carrying out its terms (Walter, 2002). Commitment problems are linked to the security dilemma, as adhering to peace agreements can relegate belligerents to a more vulnerable position than if they had remained hostile (ibid.). Peace agreements usually include addendums for demobilisation, and rebels are often required to relinquish control of territories, thus reducing their strength (Walter, 2009). Once hostilities are paused, it is possible that belligerents might exploit the ceasefire to remobilise troops or build resources for renewed conflict. Essentially, belligerents cannot be certain that their opponents will not renege on a peace agreement (Kirschner, 2010). Without credible guarantees that each party will follow-through on commitments, they will not reach a settlement.

A final bargaining problem is indivisible issues. According to theory, these include religious and territorial incompatibilities, such as secessionist conflicts (Toft, 2002; 2003; Fearon, 2004) or conflicts over ‘sacred spaces’, which have symbolic importance for multiple groups and for which there are no acceptable alternatives or pay-offs (Hassner, 2003; 2006; Goddard, 2006). However, recent analyses of Islamist conflicts have downplayed the significance of issue indivisibility, proposing that the transnational nature of the conflicts (Nilsson & Svensson, 2020; Söderberg Kovacs, 2020) and the influence of spoilers and public backlash (Matesan, 2020) are more important.

Fragmentation and Conflict Resolution

Research on fragmentation generally observes an exacerbating effect on conflicts, such as intensifying violence (Bakke, Cunningham & Seymour, 2012; 2015; Fjelde & Nilsson, 2012; Staniland, 2012;

Bloom, 2005; Wood & Kathman, 2015), encouraging defections (Pearlman & Cunningham, 2012;

Kalyvas, 2008), increasing intractability (Cunningham, K. 2016), creating spoilers (Stedman, 1997;

Nilsson, 2008; Nilsson & Söderberg Kovacs, 2011), and hastening its renewal (Rudloff & Findley, 2016). Regarding how fragmentation affects the likelihood of a peace agreement, the existing literature posits three core mechanisms: creating spoilers, exacerbating commitment and information problems, and reducing bargaining space.

(11)

10 It is well-established that spoilers can emerge out of fragmenting parties to derail peace processes (Stedman, 1997; Nilsson, 2008; Nilsson & Söderberg Kovacs, 2011). Put simply, leaders for whom continued conflict has utility are likely to violently reject peace processes (Stedman, 1997:5). This occurs because peace processes exacerbate pre-existing divisions within rebel movements and groups, causing disagreements over preferred policy, leadership, and strategies, particularly between

‘moderates’ and ‘hardliners’ (Cunningham, 2006; Duursma & Fliervoet, 2020; Lounsbery & Cook, 2011). Alternatively, Pearlman’s (2008:83) internal contestation theory suggests rebel spoilers emerge to improve their power relative to their rebel rivals. Competition and leadership contests within rebel movements may cause a group to reject deals which align with their preferences to undermine a rival group's strength (ibid.).

Second, fragmentation exacerbates information and commitment problems. For instance, Kathleen Cunningham (2013) argues that fragmentation reduces the credibility of rebel groups and increases uncertainty regarding the chances of settlements effectively resolving disputes. According to this approach, this occurs through two mechanisms. First, as the number of rebel factions increases, the ability of a group to credibly commit to agreements on behalf of others and enforce their compliance becomes increasingly difficult to gauge (Cunningham, K. 2013; Blattman & Miguel, 2010:16).

Furthermore, factionalism increases the likelihood of leadership contests and change, potentially affecting their ‘internal continuity’ and their ability to make credible commitments for the longer term.

Second, fragmentation makes assessing rebel strength and resolve more difficult, increasing the likelihood of miscalculations of the utility of war (ibid.). These information and credible commitment problems increase the likelihood of bargaining failure. However, whilst these arguments make sense when considering internally fragmented rebel groups, it does not account for the fact that independent rebel groups seldom claim to be able to commit on behalf of others. Consequently, whilst it is compelling in explaining why internal fragmentation limits negotiation attempts, it does not adequately explain how movement fragmentation influences the likelihood of a settlement.

A third mechanism elucidated within the literature argues that fragmentation reduces the likelihood of a peace agreement by reducing the bargaining space (Cunningham, 2006; 2011). As rebel fragmentation increases the number of actors, the number of preferences that must be addressed also increases, thus limiting the number of settlements which would be acceptable to all parties. This

(12)

11 increases the likelihood of an actor exercising a ‘veto’. Notably, veto players differ from spoilers, as spoilers are only identified ex-post. Veto players, on the other hand, can be identified ex-ante, as whilst they have the potential to spoil peace processes, they do not necessarily do so. This paper builds on this segment of the literature, assessing the relevance of Cunningham’s (2011) veto-player framework, whilst augmenting it with insights from political parties’ literature and outbidding mechanisms, in adjusting the extent to which preferences diverge.

However, contrary to the expectations of these theories, it has been observed that the duration of war does not always increase with fragmentation and such wars can and have ended in negotiated agreements, at times quicker than non-fragmented conflicts (Findley & Rudloff, 2012). Utilising a game-theoretic model of war and several examples, Findley & Rudloff (2012) find that as the probability of fragmentation increases, the duration of time until a negotiated settlement decreases.

They argue that this is because fragmentation weakens rebel groups, creating conditions in which negotiations and settlements are a more favourable option (ibid.). Furthermore, as discussed, fragmentation often arises from disagreements over strategy. In particular, schisms are likely to arise between moderates and extremists over issues such as the use of violence, priority of demands, and participation in peace processes (Duursma & Fliervoet, 2020; Kydd & Walter, 2006; Stedman, 1997;

Zartman, 1995). In other words, splinters can occur precisely because some factions wish to pursue a settlement. According to Findley & Rudloff (2012:20), the Democratic Republic of Congo is just one example where fragmentation resulted “in new configurations of actors more conducive to joint cooperation, following which peace agreements were signed”.

This disagreement regarding how and when fragmentation influences the likelihood of a peace agreement reveals a research gap. Crucially, we do not yet know for certain under what conditions fragmentation does or does not have certain theoretically expected outcomes. Consequently, this study addresses this lacuna. In doing so, it also aims to assess the relevance of Cunningham's (2011) veto- player framework and advance it with an outbidding mechanism. As will be explained, whilst Cunningham’s (2011) framework is compelling, I argue that his condition of ‘preference autonomy’ is insufficient in explaining variation in bargaining space. Noting deficiencies in existing theory, it is important to refine them so that they better capture empirical realities. Furthermore, as only a quarter of civil wars end in negotiated settlements (Cunningham, 2011:27), understanding the conditions under which fragmentation influences the likelihood of an agreement is central to improving their

(13)

12 prospects. Consequently, this paper seeks to answer: under what conditions does fragmentation influence the likelihood of a peace agreement?

Theory

The theorised causal mechanism relies on two separate but related logics, the logic of veto players and the logic of outbidding. First, key concepts are defined. Second, Cunningham’s (2011) framework is outlined, arguing that how fragmentation affects the likelihood of a peace agreement is dependent on changes in the number of veto players. Next, the insufficiency of Cunningham’s condition of

‘preference autonomy’ is elucidated, and ‘preference divergence’ is presented as a compelling alternative. The veto-player framework is subsequently augmented with an outbidding mechanism, arguing that the extent to which preference divergence occurs is conditional on whether we see escalatory outbidding.

Concepts and Definitions

This section defines the independent and dependent variables of interest: rebel movement fragmentation and peace agreements. Two broad concepts linked to the outbidding mechanism are also defined: ethnicity/co-ethnics, and preference divergence/convergence (radicalisation/moderation). These definitions inform the operationalisation of these variables, discussed later in the ‘Research Design’.

IV: Rebel Movement Fragmentation

Drawing from the literature, ‘fragmentation’ generally falls within two categories. Group, or internal, fragmentation refers to levels of internal cohesion or factionalisation within rebel groups (e.g.

Staniland, 2010; Woldemariam, 2016). Movement fragmentation, on the other hand, relates to the level of cohesion and diversity of groups within a particular movement, for instance, a secessionist movement (e.g. Cunningham, K. 2013; Fjelde & Nilsson, 2018; Salverda, 2017). This paper is concerned with the latter.

To understand if a movement is fragmented, one must first define a rebel movement. According to Bakke, Cunningham & Seymour (2012:266) movements “mobilize around a collective identity in pursuit of particular interests related to this identity in a fundamental way”. Consequently, where

“rebel groups share a collective identity around which their activities are organised”, they are from the

(14)

13 same movement (Mosinger, 2017:15). For instance, groups which purport to represent the aspirations of any given ethnic group, whilst varying in their specific preferences, such as seeking autonomy vs secession, are considered part of the same movement.

To be fragmented, there must exist within this collective identity a number of independent rebel groups.

This paper adopts Mosinger’s (2017:15) definition of independence, arguing that groups within a movement are considered independent provided that they do not share a “hierarchical superior”.

One approach to fragmentation suggests that it is driven by the splintering of existing rebel groups into two or more constituent parts (Rudloff & Findley, 2016). This view has credence as many rebel groups emerge out of schisms with an extant parent group. However, this does not acknowledge that many rebel groups arise concurrently and do not share a parent group. Despite differing origins, the proliferation of groups effectively fractures the movement into more constituent parts. Consequently, understanding fragmentation in terms of actor proliferation is useful, and is particularly suitable for capturing the theorised mechanism which concerns the proliferation of veto players.

DV: Peace Agreements

Whilst there is no legal definition for a peace agreement, consensus dictates that it is a formalised agreement between two or more belligerent parties aimed at ending an armed conflict. According to Bell (2008: 305), “peace agreements are documents produced after discussion with some or all of the conflict's protagonists, that address militarily violent conflict with a view to ending it”. Adopting this definition, peace settlements are consensual contracts between various actors in war that emerges from prior and/or ongoing negotiations. Notably, peace agreements will not necessarily succeed at ending a conflict. As the Uppsala Data Conflict Programme (UCDP) observes, a peace agreement “addresses the disputed incompatibility, either by settling all or part of it, or by clearly outlining a process for how the warring parties plan to regulate the incompatibility” (Pettersson & Öberg, 2020).

This paper adopts a definition based on both Bell (2008) and the UCDP, defining a peace agreement as the formal document produced during/after consensual discussions between two or more belligerent parties that addresses some or all the incompatibilities of a conflict. As such, peace agreements which aim for dyadic peace, as well as full peace, are considered. Conversely, ceasefires which only halt violence, without addressing core incompatibilities, are not considered peace

(15)

14 agreements. Similarly, unilateral documents that address some or all the incompatibilities but do not involve the consensual participation of any other belligerent parties are also not considered peace agreements.

Ethnicity and Co-Ethnic Constituencies

To understand ethnic outbidding, and why it occurs, the concept of ethnicity and co-ethnicity must be explored. Despite its common reference in social sciences, there is no consensus for the definition of ethnicity. Generally, ethnicity refers to the cultural and social societal cleavages from which individuals identify belonging from shared origins, background, culture, traditions, identity, language, or religion (Aygemang et al.,2005; Cederman et al., 2010). Any combination of these, or all, may delineate an ethnic group's boundaries, depending on the specific context from which it is borne.

Lacking systematic boundaries, Gabbert’s (2006:1) definition of ethnicity as a “specific form of social differentiation whereby actors use cultural or phenotypic markers to distinguish themselves from others” is most useful. Following this logic, we can conceptualise ethnicity as, primarily, the product of self-identification. Consequently, that an individual perceives their belonging to a group is more important than identifying the markers that constitute that group’s boundaries. Put simply, ethnicity can be identified through an individual's self-selection into a specific group or through the perception of others of their belonging to that group.

Once ethnicities are identified, identifying co-ethnic constituents is straightforward. A government’s co-ethnic constituency is identified by assessing which ethnic group occupies the dominant government positions (Fjelde & Hultman, 2014:1238). A rebel group's co-ethnic constituency can be identified by assessing which ethnic group they recruit from most and/or which ethnic group they profess to represent the aspirations of (ibid.).

Preference Divergence and Convergence: Radicalisation and Moderation

The outbidding mechanism is concerned with the divergence of preferences. Understanding what preferences are, and what constitutes divergence or convergence, is therefore critical for understanding the theory. Broadly, an actors’ ‘preferences’ refers to their values and their relative order. Whilst preferences and ideology are not synonymous, ideology is a useful proxy for understanding preferences, particularly when discussing institutionalised preferences belonging to a

(16)

15 rebel group. Gade et al. (2019:2075) offers the most useful conception, arguing that group ideology can be disaggregated along three axes: conflict framing, conception of the ideal polity, and territorial aspiration. Consequently, this thesis adopts this definition, and considers preference changes along these axes.

The process of moderation and radicalisation connect to the concept of polarisation, defined by Tilly

& Tarrow (2006:217) as the “increasing ideological distance between political actors or coalitions''. In other words, moderation and radicalisation refer to diametrically opposed changes in demand, ideology, or stance along the three axes. Assuming political stances to exist on a spectrum, moderation and radicalisation can be conceptualised as centripetal and centrifugal movements from the center, respectively. Therefore, moderation from one group can be understood as preference convergence, as the distance between the groups' ideals reduces. Radicalisation from one group, on the other hand, increases the level of preference divergence as the distance between ideals increases.

Notably, both moderation and radicalisation are referent concepts measured by the changing distance between ideals (Kundnani, 2012). Consequently, discussions regarding moderation and radicalisation with reference to changing stance should not be misinterpreted as a judgement of the ideas themselves (Sprenkels, 2019).

Causal Mechanisms

Having established definitions, this section develops the causal mechanism. First, the logic of veto players and its impact on the likelihood of peace agreements is elucidated (Cunningham, 2011). Here, the limitation of Cunningham’s (2011) ‘preference autonomy’ is discussed, with preference divergence posited as a more useful concept. Next, the logic of outbidding is explained, outlining the mechanism through which competition for civilian support can result in preference divergence. Lastly, the veto player and outbidding mechanisms are merged to create a novel causal mechanism, positing three hypotheses regarding the conditions under which fragmentation influences the likelihood of peace agreements.

The logic of ‘Veto Players’ and their Impact on the Outcome of Peace Processes

According to Cunningham (2011:15), there are four core mechanisms through which an increase in armed groups in conflict inhibits the likelihood of a peace settlement. First, as the number of groups

(17)

16 increases, the bargaining space, within which a mutually acceptable agreement exists, decreases or diminishes entirely (ibid.). Second, the likelihood of bargaining failure because of information asymmetries is higher in multi-actor conflicts, as gauging the likelihood of victory or defeat vis-a-vis the state and each other becomes more complex (ibid.). As a result, armed groups may miscalculate the costs and benefits of the utility of continued conflict, with overly optimistic calculations extending conflict duration. Third, armed groups have incentives to delay the signing of peace agreements to manipulate greater concessions from the state (ibid.). This means parties are more likely to leave, or refuse participation in, peace negotiations (ibid.:134). Fourth, multiple groups will ally and fight with one another across numerous issues in different ways, leading to ever shifting and fragile alliances.

This reduces the formation of solid negotiating ‘blocs’, which would effectively reduce the number of

‘actors’ at the table, mitigating the three prior problems (ibid.:15).

But whilst increasing the total number of groups certainly decreases the number of resolutions that are acceptable for all parties, not all parties are equally important for the reaching of a peace agreement.

The logic of veto players dictates that only groups who can effectively ‘veto’ a settlement they perceive as unacceptable are able to spoil a peace process (Cunningham, 2011). Conversely, groups without the capacity to exercise a ‘veto’ have little influence on the outcome of negotiations. Consequently, the more veto players that exist, the less likely a peace agreement will be reached.

Applying this logic to fragmentation, fragmentation will decrease the likelihood of peace settlements under the condition that fragmentation increases the number of veto players. If the nascent rebel group is not a veto player, whilst the extant groups are, overall, the number of veto players in conflict has not increased and therefore the projected likelihood of settlement remains the same. If the nascent group is a new veto player, but the group from which it split loses its status, similarly the number of vetoes remains constant. Any changes here in the likelihood of a peace agreement, will depend on whether there was a divergence of preferences. If the number of veto players decreases, for instance if the splintering of a group significantly weakens the viability of both the splinter and parent groups, then the likelihood of settlement increases as the potential bargaining space is increased.

(18)

17 Figure 1. Veto Player Causal Mechanism

Internal Cohesion, Viability, and Preference Autonomy

According to Cunningham (2011), whether a group is considered a veto player rests on three conditions, internal cohesion, viability, and preference autonomy. First, the groups must be sufficiently cohesive to overcome collective action problems associated with exercising a veto. Groups which cannot present cohesive demands and foster compliance amongst their members to follow through on any threats made in pursuit of those demands, do not have veto capacity (ibid.:36).

Second, the group must be able to continue the conflict unilaterally, even if all other combatants cease hostilities (ibid.). In other words, they must be strong enough to pose a sufficient threat of sustained hostilities against the state, even in the absence of all external support. Measuring a group's viability encompasses several factors, including their military capabilities, geographical conditions, and access to resources (ibid.). The impact of military capabilities on a group's strength is axiomatic. Larger forces, more training, and better weaponry will naturally increase the ability of a group to sustain hostilities.

Conversely, groups which lack such capacity will be too weak to fight the state and thus cannot exercise a veto. Geographical conditions may also impact capabilities, as deeply rural conditions such as forestry or mountainous regions may aid rebel groups in insurgent warfare. Lastly, access to resources is imperative to fund and supply sustained hostilities (ibid.:37). Resources can be acquired through both coercion and consent, such as looting, taxation, donations, and external sponsors.

(19)

18 However, if the group is dependent on an external state sponsor, they are not a veto-player as they cannot continue the conflict unilaterally (ibid).

Lastly, for a group to be a veto player its preferences must diverge from other groups in the conflict.

In Cunningham’s latter works, ‘autonomous preferences’ (2011:33) replaces preference divergence (2006). The crux of this condition is that groups which have identical preferences “do not present two different sets of interests that need to be satisfied” (Cunningham, 2011: 33). Consequently, the two groups are only one veto player - provided the two other conditions are fulfilled. In most cases, armed groups will differ in at least some respects, and thus have autonomous preferences. However, this is not universal. First, rebel groups may form coalitions if they realise that their chances of victory are poor and would be improved if they aligned with other groups. In these cases, if they have effectively ceased operating independently then they are a unitary veto player. Second, when actors enter the conflict with the sole aim of supporting an existing player, it can be presumed that any settlement which would appease the original actor would also appease its supporter, and therefore they are a unitary veto player (ibid.).

Figure 2. Necessary Conditions of a Veto Player

Where ‘autonomous preferences’ and ‘preference divergence’ differ, is that the latter accepts the salience of the extent to which centrifugal or centripetal movements in stance affect the likelihood of reaching a peace agreement. Essentially, in Cunningham’s (2011) basic veto-player framework, internal

Internal Cohesion

Veto Player Viability

Preference Autonomy

(20)

19 cohesion, viability, and autonomous preferences can be considered dummy variables. Theoretically, any group which is sufficient in all three conditions will be able to exercise a veto. However, the extent to which actors' preferences diverge from one another, by Cunningham’s (2011) own admission, will have a significant effect on the bargaining space, and therefore the likelihood of settlement. In other words, variation in this condition beyond meeting the veto player threshold will, theoretically, have a greater effect on the likelihood of reaching an agreement. The greater this divergence between veto player preferences, the smaller the bargaining space and the less likely a peace agreement.

But that divergence will occur is not axiomatic. Whilst Tokdemir et al. (2020) do suggest that ideology and demands frequently change following fragmentation, they also highlight that such changes can be the moderation of stance. Moderation of preferences, whilst not the complete overlap of preferences and therefore maintaining veto player status, does represent centripetal movement in preferences.

That is to say, we can understand moderation as preference convergence. Recalling the models of bargaining, the closer the preferences of the actors in war, the larger the prospective bargaining space within which a settlement might be found. Furthermore, as Pruitt (2007) argues, lowering aspirations are pivotal for fostering readiness for negotiations amongst belligerents.

For this reason, delving deeper into the causes of preference divergence/convergence is central to understanding how fragmentation affects the likelihood of a peace agreement. Cunningham (2011) posits that variations in the level of preference divergence is caused by whether the group is a splinter or an ‘original group’. As splinter groups generally arise out of altercations between factions over specific policies, leadership, or strategies (Cunningham, 2006), splinter groups can be expected to have similar preferences as they share members and their original raison d’etre. In contrast, ‘original’

groups which arise separate to extant groups are likely to have more diverging preferences as they do not share origins.

I argue that this is insufficient in explaining preference divergence, as it assumes that group preferences following splits are essentially static. Crucially, it fails to take into account the competitive dynamics through which preference divergence can be exacerbated as a direct result of fragmentation. The following section, outlining the logic of outbidding, explores the conditions under which fragmentation leads to a greater preference divergence, thus having a significant effect on the veto- player framework predictions.

(21)

20 The Logic of Outbidding and its Impact on Preference Divergence

Applying the logic of outbidding to fragmentation in conflict, this paper suggest that increased intergroup competition between nascent and extant co-ethnic rebel groups, arising from rebel fragmentation, can incentivise radicalisation of ideology and demands (preference divergence) thus decreasing the likelihood of reaching a peace agreement. As will be discussed, however, escalatory outbidding is not inevitable. Therefore, whether outbidding occurs or not will significantly alter the level of preference divergence, and consequently, the likelihood of a peace agreement.

That the proliferation of armed groups results in increased competition is well supported. First, the arrival of nascent challengers reduces the expected share of potential gains extant actors might gain from settlements (Wood & Kathman, 2015:169). In other words, as the number of rebel groups increases, the number of settlements that are acceptable to all parties decreases, thus increasing competition between them to achieve a worthwhile settlement. Furthermore, whilst opposition groups may share their desire to topple the incumbent power, their preferences and values will differ, and thus any potential spoils from victory must be shared.

The entry of new groups also generates significant competition over civilian support and local resources (Fjelde and Nilsson, 2012; Wood & Kathman, 2015). Civilian support is essential for armed groups as they provide material goods, logistical support, and recruitment pools which are vital for the maintenance of hostilities (ibid.). Furthermore, civilian networks provide important information for armed groups regarding defectors and the actions of rivals (Kalyvas, 2006). Constituencies can also confer legitimacy to armed actors, providing leverage in negotiations and access to certain markets. Local populations and resources, however, can only support a finite number of armed groups, with access constituting a zero-sum game whereby gains by one rebel group correspond with losses for another. As civilian support is of longterm strategic importance, groups have significant incentives to procure and maintain access. Therefore, should access be threatened by new groups, armed actors can be expected to adapt their strategies.

In multi-actor settings, civilians have options for whom they support. Consequently, rebel groups must be uniquely appealing to attract and secure civilian support, loyalty, or compliance. The logic that drives this is that groups must differentiate their ‘brand’, to distinguish themselves amongst their rivals and gain a competitive advantage (Tokdemir et al, 2020). Rebel groups can do this by offering

(22)

21 unique material or non-material goods, such as shifting their ideology and demands towards the state (ibid.).

Prior research on political parties has paid particular attention to ‘ethnic outbidding’, the process through which competing political parties adopt increasingly extreme views to maintain political relevance and attract the support and votes of extreme constituencies. The logic that belies this, is that in societies that are ethnically stratified, groups which represent the same co-ethnics must compete for the support of their constituents. Under the original models of ethnic outbidding introduced by Rabushka & Shepsle (1972) and Horowitz (1985), ethnic groups present as separately organised groups (Group A and Group B) with diametrically opposed preferences. As the ideal preferences of Group A and Group B lie at opposite ends of a spectrum, if multiple parties that represent Group A arise, a party will attempt to outbid a more centrist party for a broader share of potential supporters by aligning their position closer to Group A’s ideal preferences at one end of the spectrum. In doing so, the group can claim to be the ‘true representative’ of that constituency and its preferences, and denigrate the more moderate group as an ethnic traitor, weak, or rival sympathiser (Chandra, 2005). Threatened by the prospect of a rival's increasing popularity amongst their co-ethnic constituency, and the backlash of accusations, the remaining party/parties will try to outbid the initial bidder, escalating the level of radicalisation amongst all rebel groups that represent that ethnicity.

We can extrapolate this mechanism to rebel groups in conflict as competing rebel groups and political parties share a significant similarity: they both rely on civilian support. Whilst rebel groups do rely to some extent on coercive measures (violence or threats) to ensure compliance and deter defections (Kalyvas, 2006), such measures are often counter-productive (Ross & Gurr, 1989; Cronin 2009).

Consequently, like political parties, rebel groups attempt to foster consensual civilian support by engaging in actions that signal their resolve and strength as the legitimate constituency representative, through the provision of material or non-material goods. As a result, mechanisms regarding competition over popular support can be similarly applied.

In conflict settings, outbidding rebel groups can expect several benefits. By radicalising demands, rebel groups can signal the strength of their resolve to the state, their members, and their constituents (Vogt, Gleditsch & Cederman, 2021). According to previous literature, this allows rebel groups to gain

(23)

22 greater concessions, increases mobilisation and recruitment, and homogenises membership, increasing internal cohesion and discipline (ibid.).

The strategies by which rebel groups can outbid one another fall into two broad categories: deed- based outbidding and discourse-based outbidding (Lilja, 2012). The former refers to increasing use of violence and threats against one another and the state (Bloom, 2004; Cunningham et al. 2012), whilst the latter refers to the instrumentalisation of myths, cultural symbols, and historical narratives that appeal to the emotions of ethnic constituents (Lilja, 2012).

Outbidding presents significant difficulties for the likelihood of conflict settlement through several co-existing mechanisms. First, radicalisation of demands reduces the bargaining space between the state and rebel groups, reducing the number of acceptable solutions. Second, groups that attempt to outbid rivals have incentives to remain outside of the peace process, to foster an image of resolve, thus creating potential spoilers (Stedman, 1997). These groups may derail peace through violence (deed-based outbidding) or by criticising co-ethnic groups involved in ongoing peace processes (discourse-based outbidding). This increases the costs of participation, increasing the likelihood that other rebel groups may drop out (Maney et al., 2012). Lastly, outbidding can prevent future settlements as rebel group leaders who make radical claims may find their strategy choices limited as adopting radical positions raises the reputation costs for back-tracking in the future (Jenne, 2007).

However, as aptly highlighted by Mitchell et al. (2009:397), “extremist outbidding in ethnic party systems is not inevitable”. This paper does not seek to determine why outbidding does or does not occur, rather, we are interested in how the presence of escalatory outbidding, or not, affects the likelihood of a peace agreement. Where escalatory outbidding is absent, some opposition groups may be more likely to be co-opted by the government. This is because splinters often arise from disagreements between moderates and radicals over strategy or preferences (Duursma & Fliervoet, 2020; Kydd & Walter, 2002; Stedman, 1997; Zartman, 1995). In other words, splinters may occur precisely because some members prefer peace. Consequently, fragmentation may result in preference convergence between the state and that specific fragment. Where one side favours a more conciliatory approach, if undeterred by outbidding, successful negotiations are more likely. Secondly, as fragmentation often weakens rebel groups, their cost-benefit analyses of war may alter in favour of seeking alternative solutions (Findley & Rudloff, 2012). Fragmentation can also reveal information

(24)

23 about the relative capabilities of various actors, ameliorating information problems (ibid.). Due to these coexisting mechanisms, if co-ethnic rebel groups are not drawn into peace-hindering escalatory outbidding, then more moderate groups may be more likely to negotiate a deal with the government.

The outbidding causal mechanism is summarised below:

Figure 3. Outbidding Causal Mechanism

Combining Veto Player and Outbidding Mechanisms

Combining the logic of veto players and outbidding, we can assert several conditional mechanisms.

First, where there is a reduction in veto players, the likelihood of a peace agreement will increase. This is irrespective of whether rebels engage in outbidding or not. Regardless, less players will be able to exercise a veto, and thus are unable to affect the outcome of a peace settlement, increasing its likelihood.

Where fragmentation increases the number of veto players, the degree to which it influences the likelihood of reaching a settlement will depend on the extent to which preferences diverge or converge.

Consequently, should escalatory outbidding occur, preferences will diverge, reducing the likelihood of a peace agreement. Conversely, should escalatory outbidding not occur, groups with more moderate preferences may be co-opted, thus increasing the likelihood of a peace agreement. The combined causal mechanism is outlined below:

(25)

24 Figure 4. Fragmentation - Peace Agreement: Conditional Causal Mechanism

From the above causal mechanism, we can draw three hypotheses:

H1a: Where fragmentation increases the number of veto players, and escalatory outbidding occurs, the likelihood of a peace agreement decreases.

H1b: Where fragmentation increases the number of veto players, but escalatory outbidding does not occur, the likelihood of a peace agreement increases.

H2: Where fragmentation decreases the number of veto players, the likelihood of a peace agreement increases.

Scope Conditions

A significant scope condition is that the theory only applies to groups which rely on civilian support for recruits and/or resources. Groups which do not rely on civilian support, do not have incentives to outbid their co-ethnic rivals to increase their share of supporters within their constituency. For example, the Islamic State derives much of its finances from oil and a large proportion of its forces are foreign fighters. As such, the reliance of the Islamic State on its local populations is minimal, and

(26)

25 it would likely be able to continue hostilities without the broad support of its constituencies.

Consequently, we would not expect to see escalatory outbidding from the Islamic State with its co- ethnic rivals.

III. Research Design

This section outlines the research design including method, case selection, operationalisation, timeframe, and data collection. Finally, the structure of the empirics and analysis is outlined.

Methodology

This paper uses a qualitative structured-focused comparison (SFC) approach to examine four cases:

the Philippines (1990-1996), Uganda (1986-1988), Ethiopia (1983-1993), and Afghanistan (1980- 1990). Whilst it is argued that qualitative approaches are limited by their lack of clear assessment on the mutual impact between variables and outcomes (Drozdova & Gaubatz, 2014), qualitative research is commended for its ability to contribute to the discovery of causal mechanisms (Glaser & Laudel, 2019). Given that this thesis is concerned with understanding the conditions through which a causal mechanism occurs, rather than whether there is merely a correlation, a qualitative approach is appropriate.

An SFC approach is ‘structured’ as standardised questions are applied to each case, making analytical comparisons systematic and allowing cross-case conclusions to be drawn (George & Bennett, 2005:67) This overcomes concerns that qualitative research can be insufficiently systematic and therefore unable to draw credible conclusions about the population at hand. It is ‘focused’ as it only considers theoretically relevant aspects of the cases within a delimited timeframe (ibid.). Because of this, the questions are determined by the referent research interest.

Case Selection

As this study attempts to explain variation in how fragmentation affects peace agreements, this study must adopt a cross-case comparative approach wherein variation is present. A case study approach is beneficial as it allows detailed exploration of the mechanisms, ensuring high internal validity. By testing the causal mechanism across multiple contexts, the study also generates higher external validity (Ruffa, 2020). The process of case selection is the most vital component of research design for qualitative

(27)

26 studies or studies in which random sampling is not possible (Marshall, 1996; Seawright & Gerring 2008:295).

This paper uses a most-similar case approach, otherwise known as Mill’s method of difference. This method isolates the causes of variation in the dependent variable, by holding other variables constant.

Consequently, to determine the conditions affecting how rebel movement fragmentation (IV) affects peace agreements (DV), all chosen cases have fragmentation present, but vary on the presence of a peace agreement. Other possible factors, barring the theorised conditions, are held constant to exclude them as explanations for variation. Consequently, analysing the differences between cases should identify the conditions that influence variance in how fragmentation affects the likelihood of a peace agreement.

As we are only interested in conflict cases in which fragmentation occurred, a list of all potential cases (the population) was established. Recalling its definition, a movement is fragmented when multiple independent rebel groups mobilise around a collective identity. This may occur through splintering of extant groups or via the proliferation of ‘original’ groups. To create the population, this study drew on Duursma & Fliervoet’s (2020) dataset of splinters and augmented it with cases of fragmentation drawing on the UCDP/PRIO Armed Conflict Dataset v.20.1 (‘UCDP/PRIO-ACD’)(Pettersson &

Öberg, 2020; Gleditsch et al., 2002). Where multiple actors were listed in ‘side_b’ in a Case-Year, these were included as potential cases. This list was then filtered. First, cases of fragmentation which occurred after 2011 were removed as insufficient time has elapsed to study each case for a 10-year period. Second, actors were cross-checked against the UCDP to ensure they were not mergers of former rebel groups, as this would represent a reduction in fragmentation. For instance, Duursma &

Fliervoet (2020) list the Coordination de la Résistance Armée (CRA) as a splinter of the Front de libération de l'Aïr et de l'Azawagh (FLAA). Whilst it is true that the CRA was primarily former-FLAA members, the CRA was actually a merger of several armed groups including the FLAA, Front de libération de Tamoust , the Patriotic Front for the Liberation of the Sahara, and the Armée révolutionaire de libération du Nord-Niger (UCDP, 2021b). This case, and other mergers, were removed.

To determine the presence of peace agreements, the PA-X Peace Agreements Database v2 (Bell &

Badanjak, 2019; Bell et al., 2019) was consulted. Recalling its definition, a peace agreement must

(28)

27 address some or all of the conflicting parties’ incompatibilities. Consequently, agreements which addressed ceasefires, pre-negotiations/processes, and renewals of peace agreements were not included. As PA-X only extends from 1990-2020, academic works were consulted applying the same criteria for earlier cases.

To remove cases where endogeneity (reverse causation) could not be discounted, cases where fragmentation occurred either immediately prior to or after a peace agreement were removed. For example, cases in which fragmentation was known to have occurred in the same year as a peace agreement, but the exact date or cause of fragmentation could not be determined, were removed.

Confounding Variables

Mill’s method of difference dictates that possible confounders should be controlled to isolate the causes of variation. To identify confounders, it is pertinent to return to previous research. Two rival theories can be derived from the literature regarding how fragmentation influences the likelihood of peace agreements: power imbalances and internal cohesion. The first suggests that imbalances of power within fragmented rebel movements creates spoilers (Pearlman, 2008), as weaker groups are likely to lose out from negotiations. This is because weaker groups may be left out or have less negotiating power, reducing the potential spoils they might receive. Consequently, smaller groups are likely to undertake spoiling behaviour. The second theory suggests that reduced internal cohesion reduces the credibility of rebel groups, thus reducing the chances of a settlement being reached (Cunningham, 2013). As my own theory concerns internal cohesion, it cannot be controlled for.

To control for power imbalances, co-existing rebel groups were cross-checked with the UCDP for information regarding their estimated troop numbers over the 10 years from fragmentation. In cases where a significant imbalance exists for the majority period, the cases were removed. For instance, the Congolese Rally for Democracy had 20,000-60,000 troops, whereas the coexisting Movement for the Liberation of Congo had ~6,000 troops at its zenith (UNDP, 2021m;2021o). Whilst troop numbers are an unrefined indicator for a rebel group’s overall power, given the timeframe and scope of this thesis it was not possible to determine the relative power-balances of each in more depth.

Consequently, troop numbers are a useful proxy for assessing relative strength. Ten cases were identified where the strength of rebel groups were similar.

(29)

28 Beyond controlling for rival theories, other factors may also affect the likelihood of a peace agreement.

For instance, conflict intensity affects the likelihood of a perceived mutually hurting stalemate (MHS) (Zartman, 2001). Where conflicts impose higher costs, groups are more likely to grow war-weary and seek alternative methods of settlement. Conversely, in low-intensity conflicts, groups have little incentives to seek resolution elsewhere. Consulting the UCDP/PRIO-ACD (Pettersson & Öberg, 2020; Gleditsch et al., 2002), low intensity cases, where the threshold of 1,000 cumulative battle-deaths was not reached during the period of study, were removed. This left cases wherein the cumulative conflict intensity was ‘high’ (>1,000 cumulative battle deaths).

This resulted in the selection of four cases, in the Philippines, Uganda, Ethiopia, and Afghanistan.

Due to the systematic method of case selection, common selection biases, such as choosing cases on data availability, have been avoided. As several regions (Asia, Africa, and the Middle East) are represented, the generalisability of the study to the broader population, civil conflicts with movement fragmentation, should be high. The cases are summarised below:

Cases IV Confounders Conditions DV

Country Fragmentation Size Conflict Intensity

Increase in Veto Players

Escalatory Outbidding

Peace Agreement Philippines

(1990-1996)

Yes Large* High ? ? Yes

Uganda (1986 - 1988)

Yes Medium** High ? ? Yes

Ethiopia (1973-1983)

Yes Large* High ? ? No

Afghanistan (1980-1990)

Yes Medium** High ? ? No

*Large = >10,000 troops

** Medium = 1,000-10,000 troops Figure 5. Case Selection

Operationalisation

This section operationalises the variables, fragmentation, peace agreements, veto players, and outbidding and preference divergence, based on their prior established definitions. To ensure each variable

(30)

29 is captured validly, multiple indicators have been chosen based on prior research. The SFC questions asked systematically of each case are elucidated here.

IV: Fragmentation

To assess whether a rebel movement is fragmented, measured as a dummy variable, the SFC questions are: do the rebel groups mobilise around a collective identity? And are they independent?

Drawing on existing literature, this thesis uses the following indicators to identify whether the groups mobilise around a collective identity: the composition of members, to assess if they share co-ethnic identity (Fjelde & Hultman, 2014); rebel statements, revealing whom they claim to represent (ibid.); rebel demands, based on primary and secondary sources, revealing whether their activities centre around those identities (Mosinger, 2017); and rebel group names, which research finds change to mirror groups’

identities’ (Ishiyama & Marshall, 2017). To be co-ethnics, the majority of each group’s members must come from the same ethnic group. As ethnicity is based on self-identification, ethnicity does not have to be defined by the lowest possible denominator. For example, whilst multiple ethno-linguistic groups may exist within a region, if the groups subscribe to an overarching narrative regarding a shared regional identiy, then they can be considered co-ethnics. For their activities to revolve around the same identity, some, but not all, of their statements and/or demands must explicitly mention their shared identity. Fulfilling these two conditions, the groups are considered part of the same movement. Exhibiting identity via names is not a necessary condition, but it is useful for evaluating some groups.

To assess group independence, their leadership (Mosinger’s, 2017) and activities (Cunningham, 2011), are evaluated. Provided they do not share leadership nor the majority of their activities, they shall be considered independent, as these indicate both elite level and operational level independence.

DV: Peace Agreement

To assess the presence of a peace agreement, measured as a dummy variable, this paper asks: Is it a formal agreement? Does it include two or more warring parties? Was it consensual? And does it address the incompatibilities of conflict?

To evaluate whether it was a formal agreement, that it was both published and signed is assessed, as these signal formality of proceedings. To confirm that at least two parties are included, the number of warring

(31)

30 signatories is observed. To assess consent, prior involvement in negotiations and that it was signed by warring parties is considered, as groups which did not participate or were not signatories cannot have given their consent. The conflict incompatibility is assessed by looking at rebel demands, based on primary and secondary data, including the Foundations of Rebel Group Emergence (‘FORGE’) dataset (Braithwaite et al., 2020), and incompatibility data from the UCDP/PRIO-ACD (Pettersson & Öberg, 2020; Gleditsch et al., 2002). To evaluate whether the agreement addresses, either partly or fully, the conflict’s incompatibility, the incompatability and peace agreement addendums are compared. For example, if the UCDP states the incompatibility was ‘Government’, with secondary/primary data showing rebels demanded increasing representation, and the peace agreement includes power-sharing provisions, for instance, the incompatibility shall be considered addressed. Fulfilling all these conditions, a peace agreement is considered present.

Condition A: Veto Players

To analyse the veto-player mechanism, this paper asks were the nascent and extant rebel groups independent veto players? To assess this, it asks were they internally cohesive? Were they viable? And did they have autonomous preferences? Indicators have been developed for internal cohesion, viability, and preference autonomy, which are all dummy variables.

To assess internal cohesion, borrowing from Cunningham (2011), this study considers continuous leadership for most of the studied period an indicator, as it displays a lack of leadership contests. The type of command structure, as decentralised/factional systems are less cohesive, whether they present a united front, signifying strategic and operational interdependence, and levels of infighting, signifying poor internal discipline and control, are also assessed to determine if there was sufficient cohesion. The command structure signifies cohesion if central command can exercise sufficient control over strategy. Notably, decentralised systems do not prevent overcoming collective action problems, particularly if decentralisation is by design, although it may hinder it. Operative decisions may be independent from the central command, whilst still preserving cohesion, as the reality of war often precludes consistent communication. Infighting is judged to be indicative of insufficient cohesion if it results in the splintering of the group during the studied period. As internal cohesion is a thick variable and difficult to measure, and as rebels have incentives to misrepresent capabilities, this study considers a holistic assessment of the above indicators, based on secondary research, to evaluate sufficiency.

(32)

31 Regarding viability, this paper assesses troop/membership numbers and access to resources, as larger and better resourced groups pose more formidable threats (Cunningham, 2011). Provided each group has more than 1,000 members, this indicator will be assessed as sufficient, as anecdotal evidence suggests groups of this size have been shown to be able to withstand government offences over time (UCDP, 2021), particularly in remote guerilla warfare (Buhaug et al., 2009). To assess access to resources this paper investigates whether the groups were self-reliant, as groups which are dependent on an external state/group cannot continue conflict unilaterally, and consequently are not viable (ibid.). To confirm self-reliance, the group must be engaged in domestic resource extraction as their most important source of financing, based on academic research and Government/NGO/IGO reports. Lastly, this paper assesses whether the groups were able to sustain hostilities for at least a year, based on secondary data including the UCDP/PRIO-ACD, showing that they were strong enough to withstand government forces at the point of fragmentation. One year is appropriate as it displays capacity to withstand sustained hostilities, and because changes to capacity after 1 year would likely be influenced by factors other than fragmentation. As we are only interested in the effect of fragmentation on creating veto players (influencing group viability), delimiting the necessary time frame to immediately after fragmentation avoids observing outcomes which are influenced by other causal factors. Fulfilling all three indicators, the groups are considered viable.

Regarding preference autonomy, differences along their conflict framing of in-group/out-group, their ideal polity, and/or territorial aspiration are identified, as these are the three ideological axes identified in our definition of preferences (Gade et al., 2019:2075). Notably, they must differ in only one aspect to be autonomous (Cunningham, 2011). To identify this, rebel manifestos and secondary research regarding rebel aims and demands are assessed, as these reveal ideological stances. For splinters, the cause of the splinter is also considered, as splinters often arise from disagreements over preferences (Duursma & Fliervoet, 2020).

Condition B: Outbidding

To analyse the outbidding mechanism, this study asks: Do they engage in deed-based or discourse-based outbidding behaviours? And if so, did this result in radicalisation (preference divergence)?

Most comparative research conceives deed-based outbidding and radicalisation as observed escalation in violence (e.g. Bloom, 2004; Cunningham, Bakke & Seymour 2012; Lawrence, 2010). Consequently,

ADVANCING THE VETO-PLAYER FRAMEWORK