The Frontlines Have Shifted: Explaining the Persistence of Pro-State Militias after Civil War

Many ceasefire and peace agreements stipulate the disarmament and demobilization of pro-state armed groups involved in the conflict, yet few of these groups ever completely demobilize. This study seeks to explain the process of incomplete demobilization by advancing a theoretical argument that pro-state armed groups are least likely to disarm and demobilize when the monopoly on violence is frag- mented, and when there is relative balance of capabilities and interests between the government and pro-state armed groups. Under these circumstances, both governments and pro-state armed groups may favor incomplete demobilization enabling them to pursue their strategic objectives. We draw on unique interview data with pro-state paramilitaries from Ukraine to empirically illustrate our theoretical expectations. or peace accords which stipulate demobilization of pro-government armed groups. This article presents a set of theoretical expectations which seek to explain the phenomenon of incomplete demobilization, drawing on a number of empirical examples before illustrating our theoretical

Militias and other irregular armed groups are an important feature of modern conflict. In addition to rebel forces and ethnic militias, many governments have come to rely on pro-state militias (PSMs) 1 to counter both external and internal opponents. Recent years have seen a burgeoning academic discussion on the development and impact of such groups, including both empirical 2 and theoretical work. 3 This work has shed light on the consequences of PSMs for the conflict, most notably in terms of intensifying wars and worsening human rights abuses. 4 There is also evidence to show that the end of the conflict often does not mean the end of PSMs, with significant numbers persisting even after conflict termination. 5 The difficulties posed by such groups in the post-conflict period have been widely acknowledged, with difficulties demobilizing such groups potentially leading to banditry and warlordism and other forms of low-intensity violence. 6 However, there has been little theory-building related to the post-conflict demobilization of PSMs, a gap which this article sets out to address. In particular, there has been little space devoted to the question of why some states, such as Ukraine and Guatemala, struggle to demobilize their pro-state militias following conflicts. We define incomplete demobilization as a failure by governments to disarm and demobilize their PSMs in the aftermath of ceasefire or peace accords which stipulate demobilization of pro-government armed groups. This article presents a set of theoretical expectations which seek to explain the phenomenon of incomplete demobilization, drawing on a number of empirical examples before illustrating our theoretical expectations on a case study of pro-state militias from Ukraine. It should be noted that the scope of our theory relates to incomplete demobilization in cases where the ceasefire or agreement (formally) held, rather than functioning as an explanation of why peace settlements fail in general.
We begin with a discussion PSMs, with a focus on how they interact with states in the post-conflict period. In order to explain which factors enable and sustain an incomplete demobilization of pro-state armed groups following civil war, we then argue that both governments and PSMs are unlikely to pursue complete demobilization as long as the fragmentation of the means of violence persists, and as long as both actors share a balance of capabilities and interests. We illustrate these proposed causal processes using a case study built around qualitative data gathered through interviews with former paramilitary combatants and experts from Ukraine. This case effectively demonstrates the potential utility of our proposed theoretical expectations, although we stop short of testing generalizable hypotheses using a single case. We conclude with a discussion of the policy implications arising from this research.

PSMs and Incomplete Demobilisation
Pro-state armed actors come in many different forms, with much of the recent work on this subject focusing on clarifying these differences. For example, the Pro-Government Militias Database (PGMD) distinguishes between informal groups, which are arms-length from the government and largely clandestine, and semi-formal groups which are more closely and openly tied to the government while still distinct from the regular armed forces. 7 Aliyev 8 instead proposes an alternative typology based primarily on the relative strength of militias vis-à-vis the state, distinguishing between "state-manipulated" and "state-parallel" militias. State-manipulated groups are used as auxiliaries in order to counter-specific threats, while state-parallel militias emerge to fill the gaps left by weak states. While both of these typologies are useful, there is still space for a great deal of variation within these categories.
In some cases pro-state groups can actually come into conflict with the government or specific parts of the state, further subverting easy categorization. For example, Loyalist groups in Northern Ireland were pro-state in the sense that the used violence in perceived defence of the British state. However, despite evidence of some collusion between British military intelligence and Loyalist groups, they were liable to be killed or imprisoned in confrontations with the British armed forces and police. 9 In Guatemala the government funded numerous paramilitaries during the conflict, but following the peace agreement and the election of a more moderate government there was significant difficulty in reigning these groups. 10 While such groups are broadly pro-state, it is important to recognize that they are not necessarily always pro-government. This hints at the complexity of state-aligned groups in conflict. For the purposes of this article the focus is less on different types of PSMs, and more on how the relationships between these groups, the state and other political actors affects the prospects for successful demobilization and demilitarization.
PSMs bring various benefits to governments during wartime, including extra manpower, local intelligence and plausible deniability. 11 To some extent, these benefits continue into peacetime, with the potential to use extra-state organizations to repress internal opposition. However, there are also various factors which push governments to demobilize these groups when they are no longer required. Firstly, the demobilization of these groups is often a key part of the peace agreement which ended the war. Foreign and domestic adversaries are likely to insist on this demobilization as a condition of peace. Secondly, the international community and organizations have developed norms against the use of such groups, with allies and development agencies often pushing for their demobilization. More fundamentally, the existence of even sympathetic non-state actors challenges the state's monopoly on violence, undermining the state's Weberian sovereignty. All else being equal, we can generally assume that a government would prefer to have a strong, centrally controlled armed forces rather than a network of extra-state groups with ambiguous loyalties, with demobilization or integration of arms-length groups acting as a form of coup-proofing. However, despite these powerful motivations there are numerous cases of broadly pro-government extra-state groups remaining active following civil wars.
In many cases this continued mobilization occurs because the state deliberately cultivates and protects these groups even into the post-conflict period. Often this is because those groups are personally loyal to a particular leader, such as Nguesso's Cobra militia in the Republic of the Congo. In particularly weak, personalistic and isolated states the benefits of a political leader maintaining their own private militia may outweigh the negatives arising from fragmented sovereignty or international condemnation. However, other cases see a more ambiguous relationship between the government and pro-state militias, in which the government fails to exert control over increasingly independent armed actors. 12 For example, in Guatemala and El Salvador, the democratically elected governments struggled to prevent autonomous factions within the military from pursuing their own campaign of clandestine violence through paramilitary organizations. The demobilization of these groups was formally attempted and partially occurred, but ultimately remained incomplete.
Existing studies on PSMs have tended to focus on concept-building and typologies, as discussed above, rather than theorizing. Some studies have touched on the question of why PSMs may be motivated to remain active following peace agreements, including Phillips' 13 work on inter-group rivalry and Bateson's 14 study on militia socialization in Guatemala. However, there has been little attention paid to the process of demobilization specifically. There has also been relatively little theorizing about how state-militia-society relations affect demobilization, with existing work often focusing mainly on PSMs. In this study, we propose a set of theoretical expectations explaining why and how incomplete demobilization occurs, and use in-depth interviews with members of Ukrainian pro-state militias to empirically illustrate this.

Theorizing Incomplete Demobilization
This section will draw on a number of empirical examples to analyze the conditions which allow incomplete demobilization to occur and persist. While a significant amount of attention has been paid to why demobilization efforts fail there has thus far been no attempt to address the specific outcome of incomplete demobilization, particularly as relates to pro-state groups. We propose a set of meso-level theoretical expectations in the sense that our theoretical claims focus on organizational actors within a wider conflict and post-conflict context. It is also fundamentally an attempt at relational theorizing, as it focuses on how the relationship between the state and specific armed groups shapes the outcome. It is the interests, capabilities and connections between different actors which create the particular relationship which results in incomplete demobilization. We use the term theoretical expectations rather than hypotheses as we are not attempting to provide a generalizable empirical test at this stage, given our single-case approach. Instead, we are attempting to demonstrate a set of plausible causal processes which can form the basis of future cross-case research.
This theory does not attempt to provide necessary and sufficient conditions for the emergence of incomplete demobilization. The failure to successfully demobilize a particular group may have multiple antecedent causes across different contexts. Instead, we present a sequential causal process that explains why and how incomplete demobilization of PSMs persists into the post-conflict period. The first step in this process, the fragmented monopoly on violence, is a necessary condition of civil conflict and therefore of incomplete demobilization following a civil conflict. The following two steps, the balance of capabilities and balance of interests address how incomplete demobilization persists and becomes entrenched. This covers the key process and key factors underlying the persistence of incomplete demobilization, but does not seek to explain every relevant factor across every case. However, this theory explains the case of Ukraine well, and we believe it can be applied fruitfully to other cases

Fragmented Monopoly on Violence
The key background condition which allows incomplete demobilization to occur is the fragmented monopoly on violence, or the means and authority to engage in organized violence being wielded by an increasing array of state and non-state actors. This is a core feature of civil war, given that by definition a civil war involves some challenge to the Weberian monopoly on violence by at least one rebel group. 15 However, the fragmentation often goes further than this, with multiple pro-state and extra-state groups involved. Furthermore, it can also involve the fragmentation of state security institutions as different agencies pursue their own strategic goals, a phenomenon which has been identified in cases such as Nigeria. 16 States across the world have frequently devolved the use of violence to different actors, some of which may be tightly controlled by the state's military intelligence apparatus and others which may be almost entirely independent. In the most complex cases this can involve a dizzying array of small and large groups, all with differing capabilities and relations to the state. For example, the ongoing conflict in Syria has seen the participation of the government's conventional armed forces, state-controlled militias, foreign volunteers, foreign-backed militias and religious armed groups on the pro-government side alone, notwithstanding the dozens of opposition groups. These groups are not incidental; by 2015 the weakness of the Assad regime meant that he was essentially reliant on this diverse alliance of extra-state groups. 17 A comprehensive peace settlement aimed at ending a conflict will generally try to repair this fragmentation of the authority and the means to use violence, ideally reestablishing a governmental monopoly on the legitimate use of violence. 18 This generally involves some mixture of demobilization and reintegration into civilian life, for example through a Disarmament, Demobilization and Reintegration (DDR) program, and related Security Sector Reforms (SSR) such as the integration of paramilitaries into the formal security apparatus. However, there are numerous challenges to these efforts. Firstly, there may be factions or groups which actively resist demobilization. Pro-state groups might resent the government's compromise with their opponents, or may fear the consequences of post-conflict transitional justice efforts. For example, elite-backed paramilitaries in Guatemala remained mobilized following the peace agreement, frequently targeting transitional justice campaigners in order to maintain their position of impunity while undermining the government and the peace agreement. 19 Furthermore, armed groups may resist demobilization because they have economic interests which require their continued use of violence. For example, Colombian paramilitary groups re-mobilized due to their continued involvement in the drugs trade 20 and desire to prevent land reform, 21 while they also targeted witnesses and transitional justice activists. 22 Even when pro-state paramilitaries are not actively opposed to the peace agreement they may simply be left out of DDR efforts, with the focus on large-scale rebel demobilization meaning that informal militias may be ignored as was the case in Afghanistan.
The fragmentation of the monopoly on violence is not something which occurs purely outside the state but is instead a process which is closely linked to divisions and rivalry between different state actors and governing institutions. States are not truly unitary actors, and different extra-state groups may have connections with specific actors or branches of the government. As discussed above, the democratically elected government of Guatemala was not always able to effectively control some parts of its own intelligence and security services, which maintained close links to extra-state paramilitaries. 23 State capacity is certainly an important issue here, as weaker states will struggle to contain and coerce extra-state groups effectively in the post-conflict period thereby enabling their continued mobilization. However, we suggest that fragmentation is at least as important as overall state strength and capacity. Even a strong and well-developed state can struggle to deal with a plurality of extra-state groups, particularly if the state itself is divided. Ukraine is a relatively well-developed state by international standards, but nevertheless struggled to effectively assert control during and after the conflict due to high levels of division within the political system. 24 Coalition governments and personalistic politics have contributed to rivalries between the president and minister of internal affairs as well between different aspects of the security apparatus such as the National Guard and the Security Service (SBU). Such institutional rivalry has a long history in Ukraine and other post-Soviet states, with governance often being undermined by factionalism and corruption. 25 When examining a state's attempts to deliver demobilization it is important to think in terms of centralization and unity as well as overall capacity, as traditionally conceptualized. In other words, it is important to think about the functioning of a state, rather than simply its capacity. It is our expectation that states with high levels of internal division are likely to have a higher chance of incomplete demobilization.

Relative Balance of Capabilities
If the weakness and disunity of the state is matched by relatively strong extra-state actors, creating the fragmented monopoly on violence discussed above, then a relative balance of capabilities may emerge, thereby facilitating an incomplete demobilization. Extra-state groups will often agree to demobilize under pressure from the government and other actors, but in practice they frequently maintain their independent capacity for violence. They may do this in a number of ways. For example, they could reemerge as new groups with new names and new organizational structures, just as the Colombian AUC devolved into a more decentralized network of "criminal bands." 26 Groups could demobilize but then reappear over time, as with Guatemalan Civil Defence Patrols which demobilized following the end of the war but then reappeared to fill the vacuum of state power in rural areas. 27 Finally, extra-state groups can partially demobilize by handing over some of their equipment, but covertly maintain their organizational and military capacity, as occurred with Serbian paramilitaries between the Bosnian and Kosovo wars. 28 This might involve using veterans' associations to maintain command structures, as well as hiding weapons caches. In this scenario they will be formally demobilized, but covertly prepared to fight.
The crucial causal relationship here is that the aforementioned fragmentation of the means of violence helps to create an effective balance of capabilities and influence between the state and the extra-state group(s). The state is strong enough to force some level of formal demobilization, but not strong enough to fully coerce or co-opt the group. The PSM is not strong enough to openly challenge the state, but is strong enough to maintain its capacity to act. This does not mean that there is absolute military parity between the state and PSM(s). While the state is generally stronger than individual paramilitary groups, the risks of escalation may be too much for the government to handle, particularly as there is evidence that state repression of such groups can lead to a serious escalation in violence. 29 It should also be noted here that the mechanism here is capability, including diverse forms of influence and power, rather than purely military strength. The capabilities of both the state and extra-state groups lie not only in the number of weapons and combatants, but also in its degree of political influence, economic resources and public legitimacy. If paramilitary volunteers are viewed as war heroes by a significant section of the general public, as they were in Ukraine, 30 then publicly challenging them may carry significant risks for any democratically accountable politician because this group will be capable of mobilizing public support. Similarly, a group with links to economic and political elites may be more able to resist efforts to force its demobilization. A group can be relatively small in size but can wield outsized political influence due to high levels of public support or powerful social connections. Conversely, a paramilitary group which can be successfully recast as common criminals or as threats to public security may be easier to counteract. 31 It is therefore our expectation that a more symmetrical balance of capabilities between the state and an extra-state group will be more likely to lead to incomplete demobilization.

Balance of Interests
If this relative balance of capabilities develops and persists in the post-conflict period it will potentially lead to a marriage of convenience, as the state and the extra-state group enter into an uneasy alliance. This relationship will tend to become entrenched over time as the interests of all the actors become more aligned. PSMs clearly already have some shared interests with the government, which is why they supported them during the conflict. This might be a shared interest in maintaining a broad political order, such as with anti-communist groups in Latin America, or resisting foreign interference, such as anti-Rwandan groups in the Democratic Republic of the Congo. Even in cases where the government would prefer to demobilize these groups and regain greater control over the security situation, their broader shared interest are unlikely to disappear simply because the conflict has formally ended. For example, the demobilization agreement between the Colombian government and the AUC did not remove the threat of the leftist FARC guerrillas, while the Minsk Protocol in Ukraine did not remove the threat of further Russian interference. While the formal end of the conflict frequently changes the incentive structure for the state in favor of demobilization, the shared interests and goals of the government and PSMs can persist and reemerge.
However, in addition to this PSMs will develop specific interests in maintaining their capacity to mobilize, while elements within the state may find it within their own interest to maintain positive relationships with these groups. For example, extra-state groups may develop economic relationships with the political and business community, while their capacity for using force can be of use to the country's elite. At the individual level, involvement in a continued war economy may be preferable to living a life of civilian poverty. The opportunity for legitimate economic activity is often limited in post-conflict states, leading to a pool of trained combatants in need of employment. As such, both the leaders and the ordinary members of pro-state paramilitary groups may have an interest in continuing their activities and therefore in perpetuating a situation in which their mobilization is justified. The political elite can also benefit from developing a base of support among "'war heroes'", while paramilitary groups benefit from their own privileged social position. As Acemoglu et al 32 describe in the case of Colombia, this can evolve into; "… a symbiotic relationship with specific politicians holding power: paramilitaries deliver votes to politicians with preferences relatively close to theirs, while politicians they helped elect leave them alone and possibly, implicitly or explicitly, support laws and policies that they prefer".
The postwar period can therefore see the development of a balance of interests, as state and non-state actors find ways to benefit from the uneasy alliance which they are in. Even though the state as a whole may prefer to undermine and control the extra-state groups, individual parts of the state will find mutual gains from their continued relationship. At a number of levels, it will be in the interests of both state actors and extra-state group to maintain the situation of incomplete demobilization. Crucially, these connections are only likely to become closer and more entrenched over time, as networks deepen and interests increasingly align. It is therefore our expectation that connections between the state and extra-state groups will become more entrenched over time in cases of incomplete demobilization. These connections will, in turn, make it more difficult for the government to forcibly dismantle these armed groups.

Case Study: The Ukrainian Volunteer Army
To empirically illustrate the process of incomplete demobilization, we employ a case study of the Ukrainian Volunteer Army-Українська добровольча армія (UDA), and its incomplete demobilization following the Minsk I ceasefire agreement in Ukraine. The UDA renders itself as an intriguing case study of incomplete demobilization not only because the Ukraine's demobilization has not been previously studied, but also due to its potential to explain incomplete demobilization processes in similar fragmented states with multiple para-or extra-state actors. As it will be discussed in further sections of this article, the Ukrainian government was neither able nor willing to demobilize the UDA. The group has functioned as a typical pro-state paramilitary organization engaged in a civil war to defend interests of the state, but throughout its history, the UDA also remained highly critical and even antagonistic to successive Ukrainian governments. Both in its pro-state stance and its criticism of the regime, the UDA is similar to a universe of other pro-state groups in other conflict-affected states. With the proviso that this is a single case study with all the limitations that brings, the micro-dynamics of Ukraine's incomplete demobilization should be transferable to other cases. Finally, the fact that the war in Ukraine remains unresolved makes this a timely and important empirical study.
The UDA's predecessor, Ukraine Volunteer Corps (DUK) was created in July 2014 by Dmitro Yarosh, the then Member of Ukrainian Parliament and the head of Ukraine's ultranationalist political party "Praviy Sektor" (Right Sector). 33 Since its creation, the DUK functioned as a military wing of "Pravyi Sektor, " tasked with fighting pro-Russian separatists during the Donbas War. The group was reorganized by Yarosh into the Ukrainian Volunteer Army (UDA) in December 2015. From February 2016 to September 2018, UDA actively participated in counterinsurgency operations against DNR/LNR (Donetsk and Luhansk People's Republics) separatists in Donbas region, primarily engaged in reconnaissance and sabotage missions in Donetsk area. 34 During the early months of Donbas War, the DUK/UDA could boast as many as 10,000 armed members. However, hundreds of UDA members chose to sign contracts with the Army or move to other (state-affiliated) paramilitary organizations after the end of active combat phase from September 2014 to April 2015. Following the incorporation of all Ukrainian paramilitary volunteer battalions into the Ministry of Interior troops, SBU (Security Service of Ukraine) and VSU (Armed Forces of Ukraine), the UDA was left as the only paramilitary organization without any form of official affiliation with the state. Despite frequent rumors about incorporation of UDA into the VSU, the organization remained outside of state control until its demobilization in 2019. Numerous UDA combatants, attracted by financial and social benefits offered by the VSU left the group in order to join the Army. 35 By May 2018, it is estimated that there were only several dozen active-duty UDA militants left in Donbas. 36 In accordance with the Minsk II Protocol, volunteer battalions which were not controlled by the Ukrainian state were required to disarm, disband and leave Donbas by the mid-2015. 37 However, no efforts were implemented to demobilize UDA during president Petro Poroshenko's term in office. As part of his electoral promises to resume talks on the Minsk agreements, the newly elected Ukrainian president Volodymyr Zelensky ordered disarmament and demobilization of the remaining UDA units prior to the December 2019 "Normandy format" meeting of the heads of Ukraine, Russia, France and Germany. On September 12th 2019, the UDA's 8th battalion "Aratta, " along with its affiliated Chechen volunteer "Sheikh Mansur" and the OUN (Organization of Ukrainian Nationalists) battalions, were disarmed and their members ordered to leave Donbas region. 38

Methods and Data
Our choice of qualitative case study as form of enquiry is predicated by the fact that incomplete demobilization is by nature a clandestine process that is hard to observe empirically, let alone to quantify. Processes which we expect to theoretically affect the incidence of incomplete demobilization, such as fragmentation of violence and the balance of interests and power, are likely to elude precise measurement and quantification. Beyond this, understanding how these processes evolve at a micro intra-group level will allow us to explain why pro-state groups seek incomplete demobilization, as well as allowing us to investigate the factors that prevent governments from disbanding these armed groups. Case studies are well suited to the study of such causal processes, 39 and can serve as an effective illustration of our theory. 40 However, the small-n nature of this approach means that this does not serve as a generalizable test of our expectations.
To understand why the UDA is hard to demobilize and why two successive Ukrainian governments have chosen to accept the groups existence, irrespective of their obligations on the Minsk Protocol, we seek to examine how the group understands its capacity vis-à-vis the government, its ideological stance, and its mobilization potential. Examining the group's micro-level dynamics will also enable us to answer the other part of our theoretical puzzle, or the question of how the incomplete demobilization occurs.
To achieve these empirical objectives, semi-structured in-depth interviews with former members of UDA were conducted from May 2019 to February 2020. Open-ended qualitative interviews were selected as the main form of data collection due to the challenges of accessing a sufficient pool of informants necessary to carry out a survey. The main sample consists of 23 former members of the UDA, who remained with the group after the Minsk I Protocol agreement was signed on September 5 th 2014, and until the last 8 th battalion ("Aratta") of the group was disarmed and demobilized by the mid-September 2019.
All informants were males aged between 19 and 48 years old. The sample includes 5 individuals from western regions of Ukraine, 12 from eastern, 1 from southern, and 5 from the capital Kyiv. Informants were recruited through contacts acquired from the Ukrainian journalists, the "Pravyi Sektor" press center, veteran services and former members of volunteer recruitment centers. The main selection criterion was the informants' membership in UDA for a minimum of 30 consecutive days after September 2014. The UDA did not practice the use of military ranks; all informants described their rank as "volunteers" and neither served as kombat (battalion commander), or platoon/division commanders. The informants were recruited to include members of different battalions and squads. Due to security concerns, we are unable to reveal names of battalions associated with informants as it is likely to compromise the informants' anonymity. All informants self-identified as ethnic Ukrainians: 8 interview participants described themselves as Russian-speakers, 11 as dual Ukrainian-and Russian-speakers, while 4 as only Ukrainian-speakers. During interviews, informants were offered a choice between Ukrainian and Russian languages and, with the exception of 4 interviewees, all opted to be interviewed in Russian.
We believe that our sample is representative for two main reasons. Firstly, the number of UDA militants left operating in Donbas after the Minsk II agreement is estimated to be between 300 and 500. 41 Our sample covers approximately 8% of the lowest total estimate (300). Secondly, all UDA frontline personnel, in two battalions and three squads, were members of the same social media groups on Facebook, Telegram and WhatsApp. Nearly all informants confirmed that they shared news and communicated on a daily basis with their comrades from other divisions. Due to the existence of rotation system that functioned across the organization, each combatant was personally familiar with at least 200-300 other organization members, not only amongst the Donbas-based personnel, but also in other parts of the country. Bearing in mind that UDA practiced "senior amongst equals" approach to group hierarchy, the battalion and platoon commanders were known to share all relevant information amongst other group members. 42 A former member of a Ukrainian paramilitary battalion (not UDA) was hired as research assistant to facilitate recruitment of participants and to conduct interviews. Although the principal investigator is from the former Soviet Union (not Ukraine), and is a native speaker of Russian and has a command of Ukrainian language, the presence of a Ukrainian paramilitary veteran was instrumental to overcome trust issues with UDA informants.
In-depth interviews lasted between 30 min and 1,5 h. Informants were provided with open-ended questions and topics enabling them to reflect on their experiences as part of UDA, their current status, and their insider understanding of the group's dynamics and objectives. Since the interviews focused on the group's current goals, pressing safety concerns required relying on fieldnotes as the only form of recording. To alleviate ethical risks and to further secure informants' identity, a de-identified sample approach was employed, with the informant names replaced by a code, enabling researchers to securely destroy all personal data (names, contact details and other identifiers) immediately after each interview. For the purpose of this study, we replace codes with pseudonyms to refer to individual informants. Bearing in mind that the bulk of interview data consists of the informants' stories and narratives, we used narrative analysis as the principal method of data coding.
In addition, open-ended interviews were conducted with 6 former employees of the DUK/UDA volunteer recruitment centers in Kyiv and Dnipro. This group of informants was recruited due to their experience of communication with hundreds of paramilitary volunteers. Lastly, this study also draws its empirical insights from in-depth interviews with two Ukrainian investigative journalists who chose to remain anonymous, and who had worked with the DUK/UDA extensively since 2014.

Demobilized… Not Quite!
In the second week of September 2019, "Aratta" combatants were disarmed by the Ministry of Interior's Special Forces "Kord" squad, and heavily armed members of the SBU. "We knew what time they are arriving, … and we moved everything [useful] from the base, weeks in advance. There were TV crews shooting how we 'surrender' our weapons. We left them a lot of 82-mm mortar shells, anti-tank mines, a few Shmels [portable disposable flamethrower] and some other Soviet era weapons that we never managed to use" -confessed a witness of the disarmament event. 43 The government sources recorded that 10 tons of weapons were surrendered by the "Aratta" and "Sheikh Mansur" battalions on the disarmament day. 44 The former members of UDA were eager to hint that the surrendered arsenal does not include modern models of AK assault rifles, RPGs (rocket propelled grenades), night vision equipment, military drones, or any other weaponry that the group was known to use in their operations in Donbas.
Notwithstanding official televised statements by senior government officials that no uncontrolled armed volunteers are left in Ukraine, the UDA's 8 th battalion "Aratta" was moved to Yur'yevka village on the outskirts of Mariupol (Donbas region), and the 5 th battalion was relocated to Velikomykhaylowska village of Dnipro region. Although both battalions maintain limited personnel (only a dozen staff members each), they continue military training and recruitment of new volunteers. The demobilized volunteers formed territorial defence squads, which according to Yarosh 45 are designed to establish popular "resistance movement" across Ukraine. Regular boot camps are conducted by the organization's regional offices across Ukraine for DUK/UDA veterans and new volunteers to practice target shooting, martial arts, and to study in political ideology classes.
The evidence collected by this study 46 suggests that former UDA members, "demobilized" by the government, privately (yet with the group's blessing) retained not only small arms and portable anti-tank weapons, but also explosives. 47 There were cases of former UDA members using these weapons in attacks on political opponents and business competitors. 48 A former UDA member described that demobilization: "has given us an opportunity to get our movement to the 'next' level", adding that "the frontlines have shifted, … they are now not only in Donbas, but all across the country." 49 Another informant believed that "not only the [Donbas] war did not end, but with the new [Zelensky] government, the war is on our streets, so 'leave Donbas, but keep weapons' strategy was [long] planned." 50 Since UDA volunteers did not receive salaries, with the exception of small "donations" (around USD100 per month), most had to rely on their families, friends or personal savings for pocket money 51 while staying on the frontlines. "After we were send back home, I could get back to my old business and make some money, but I am still 100% pravosek [follower of Pravyi Sektor ideology]," 52 said a former UDA combatant (still wearing a UDA uniform) pointing at the group's red and black flag on the walls of his fast food café. Another informant was similarly positive about leaving Donbas: "it was fun living at our base with other guys, but the Army did not allow us to fight separatists on our terms, so since 2016 we weren't free to do what we wanted." 53 The informants confirmed that demobilization was nothing more than a well-planned transformation to a "new stage of struggle." As recently as in February 2021, the group's leader Dmitro Yarosh publicly threatened to re-mobilize all UDA activists if rallies by pro-Russian-speaking political parties (such as Opposition Platform for Life) continue. 54 Independent sources confirm that UDA's mobilization potential is likely to number up to 10,000 activists, of whom about half are war veterans. 55

Fragmented Monopoly on Violence -"Divided we Stand"
Similar to many other war affected states, Ukraine experienced extreme fragmentation of the means of violence since the start of armed conflict in Donbas. Due to a series of military debacles by the poorly organized, equipped and trained Ukrainian army, much of fighting from May to September 2014 was conducted by volunteer battalions. The DUK/UDA along with "Donbas", "Aidar", "Azov" and almost 40 other paramilitary battalions functioned as the main fighting force defending Ukrainian territorial integrity. 56 The Ukrainian government began hastily formalizing the battalions immediately after the Minsk I ceasefire agreements in September 2014. With the exception of DUK/ UDA, which continuously rejected all government efforts of formalizing them, all other paramilitary battalions were either incorporated into the Ministry of Interior troops (including National Guard), or the Army (VSU). The powerful minister of interior, Arseny Avakov is widely considered as patron of most paramilitary groups, including ultranationalist "Azov" regiment, and the UDA. 57 Avakov's patronage of (ex)paramilitary groups is believed to be the reason behind his ability to retain his ministerial position under Zelensky's government until July 2021.
The UDA effectively exploited this fragmentation and the lack of a governmental monopoly on violence. "In Spring-Summer 2014, we had the highest fighting potential amongst all other pro-government military units" recalled a former UDA combatant, adding that "we had the best weapons, best soldiers, and … they [government] admitted that". 58 While all other paramilitary battalions, lacking sufficient funding and political support, were eager to accept the government's offer of incorporation, the UDA in 2014-15 had no deficit in private donations and interest from the political actors. "Our 'providnyk' (leader) [Yarosh] was so feared and respected amongst politicians that many MPs were competing for his attention, " explained a former member. 59 Yarosh was indeed an efficient fund-raiser who established close relationships with powerful businessmen Gennady Butkevitch, Aleksandr Gerega, Gennady Korban, and above all with one of Ukraine's richest and most influential oligarchs, Ihor Kolomoisky (a dual Israeli-Ukrainian citizen). 60 During his service (2014-15) as the governor of eastern region of Dnipro, Kolomoisky "helped us a lot, without him we would have never been able to remain independent" confessed one of the informants. 61 Both financial and political patronage that UDA continuously received from elected and appointed state officials, as well as from government-associated business elites, have further exacerbated the government's severely fractured monopoly on violence.
While Yarosh and his organization maintained their hostile and antagonist rhetoric toward President Poroshenko 62 and some of his ministers, they also enjoyed extensive political and financial support from other members of government, including Kolomoisky and Avakov. Bearing in mind that Yarosh himself was elected as a MP in 2014, along with fellow leaders of paramilitary battalions Sem'yon Semenchenko ("Donbas" battalion) and Andrii Biletsky ("Azov" regiment) elected to Verkhovnaia Rada (Parliament), his political weight ensured his organization's survival. In the words of a former UDA recruiter, "we weren't anti-government, we were anti-president, … Poroshenko wasn't our leader" 63 he said, nodding his head toward a large poster on the wall of UDA headquarters in Dnipro, where he was interviewed, that says "YAROSH -is our president".
Relying on his MP mandate and vast political and financial support, Yarosh successfully "protected" UDA from either incorporation or demobilization well until 2018. However, Poroshenko's scandal with Kolomoisky and Avakov's fall out of president's favor resulted in increased state efforts of demobilizing UDA. "The new head of OOS [United Forces Operation], 64 [Sergei] Naev was unofficially ordered [by Poroshenko] to get rid of us, to remove us from Donbas completely, " said a former UDA member. 65 Another ex-UDA member provided more details: We were no longer allowed to operate in OOS area freely, we had to acquire permits from the military, a whole bunch of paperwork had to be completed before we could even get close to the frontlines. And then, they [Army] were sending us on minefields, in places of potential ambush. If we survived and completed the mission, they claimed all the benefits [lavry]. If our guys and girls died; they said it is because we aren't professionals and [we] shouldn't be allowed to be there [frontlines]. 66 President Poroshenko's efforts to "evict" UDA from Donbas were accompanied by a campaign of enticing UDA combatants to join the Army or the National Guard where they were offered high salaries, a package of veteran benefits and even free apartments. A former member revealed that "a lot of our people, experienced professionals, left for the Army and MVD [ministry of interior troops], … they couldn't resist the money." 67 Although hundreds of UDA's combatants "switched sides" since 2018, an interesting fact was mentioned by another informant: "they [who serve for government] still are members of our veteran groups, and they attend our monthly [boot] camps, they are our brothers for life, they all believe in our 'providnyk' [Yarosh].
Pravosek [ultranationalist] is not a job, it's your identity." 68 Notwithstanding Poroshenko's efforts, the UDA continued to function, exacerbating the fragmentation of the monopoly on violence. This, amongst other things, prevented Ukraine from abiding by the Minsk Protocol agreements. In addition to bitter rivalry between paramilitaries and the official security forces, competition between the Army (VSU), Security Services (SBU) and MVD, which all our informants lamented about, further demonstrates that the Ukrainian counterinsurgency forces were severely fractionalized from within.

Relative Balance of Capabilities -"Glory to Ukraine, to Heroes Glory!!"
The above is a popular motto, rehearsed by UDA members at their meetings and numerous anti-government rallies, which they regularly organize. The UDA, alongside other (ex)paramilitary organizations invests enormous efforts in promoting its image of the "people's army", and of Donbas war veterans as "saviours of the nation." The gist of UDA's ideology was explained by an informant: "we fought for the freedom of Ukraine. We are all heroes. We represent the people. No government can harm us because we will mobilize the nation against [that government]." 69 The UDA's success in avoiding demobilization not only rests on Yarosh's larger-than-life political charisma, but also on their well-maintained image of "national heroes," which enables the group to maintain the relative balance of capabilities vis-à-vis the government.
"They [government] know very well that if they go after us [UDA], all war veterans will stand up on our defence, and they [government] will fall down disgracefully," summarized a former member. 70 He added, "we may not have tanks and fighter jets, but we have combat experience and a lot of people, all across the country. We have our people within the Army, police, MVD. We will make sure that their security apparatus falls apart if they go against us." This demonstrates that UDA's power rests not on their military strength, but on their public support, high mobilization potential and vast network.
A former combatant revealed: "some people think UDA is falling apart, because our guys left for the Army and so on, but it was Yarosh's plan to 'plant' our people everywhere, " he continued: "they [current Army servicemen] are still our comrades, and we stood shoulder-to-shoulder at the [Donetsk] airport battle, where lots of our brothers lie dead, they will be UDA for life. " 71 Other informants confirmed that they maintain close friendship ties with their former comrades-in-arms who now serve in the armed forces or security services. The UDA veterans currently serving for the government are even frequently invited as trainers to the regular UDA boot camps across the country.
Since the start of his presidency, Zelensky demonstrated disdain toward paramilitaries and threatened to imprison them if they resisted demobilization. However, the new Ukrainian president is seemingly well aware that a violent crackdown on paramilitaries is likely to weaken his already questionable authority amongst the security forces. "Zelya 72 knows that we are not some criminals to declare us illegal. We are war heroes, we are heroes of Maidan. He cannot declare Maidan [revolution] illegal, if he does, then he is pro-Putin," explained a former UDA member. 73 An independent investigative journalist who chose to remain anonymous has stated that: Under both presidents, ultranationalists, with Pravyi Sektor [DUK/UDA] in charge, are untouchable. They can kill people, kill children. 74 They are above the law, they go free no matter what they do, because if somebody touches them, Ukraine will plunge into a civil war, much worse than we had in Donbas. It [civil war] will end Ukraine as a state. 75 The disarmament and demobilization of the 8 th "Aratta" (the last UDA) battalion in September 2019, was according to the informants, an outcome of a deal brokered by the group's long-term patron Kolomoisky with president Zelensky, who is believed to be on good terms with the oligarch. Many within UDA believed that the closure of Donbas bases -officially termed as demobilization -and transition to regional reserve squads is a positive move: "There was no real war in Donbas anymore. Our people died on minefields and under sniper fire. There is no glory [in that]." 76 Evidence collected by this study suggests that there are anti-nationalist elements within the Ukrainian security services and armed forces that are keen to see UDA and their supporters within the army purged, 77 and which enable president Zelensky and his predecessor to conduct aggressive policies toward paramilitaries. However, any actual efforts to outlaw the group are likely to cause a stir amongst other former paramilitary battalions, ultranationalist political groups (such as C-14 and National Corps), a broader veteran community, as well as possibly the general public (particularly in western regions), which continues to perceive UDA members as war heroes. As stated by an informant: "to arrest Yarosh, is the same as inviting [Russian president] Putin to Kyiv and handing him over all of Ukraine. [Russian Security Services] FSB had long issued an arrest warrant on Yarosh. Nobody in Ukraine will dare to do it without becoming an agent of Kremlin. " 78 Thus, the image of the group's leader as a symbol of Ukrainian patriotism, supplemented by the self-described "hero status" of UDA members, ensure the non-state group's legitimacy vis-à-vis the government, and prevent successive Ukrainian governments from disbanding and criminalizing the paramilitary organization.

Balance of Interests -"Death to the Enemies of Ukraine!!"
This is another ultranationalist slogan, which the UDA members eagerly spray on the walls of businesses, individuals and mass media outlets, believed to sympathize with the Russian-speakers. Although UDA has had hundreds of Russian-speakers and even ethnic Russians from southern and eastern regions of the country amongst their ranks, 79 who self-describe as Ukrainian patriots, the group maintains hostile attitude toward mass media groups, businesses and public figures who choose to speak Russian language publicly. The UDA's anti-Russian stance seems to fit comfortably with the government policy of Ukrainianization, and complete eradication of Russian language (as an ethnic minority language) from the public sphere. The successive Ukrainian governments successfully turned a blind eye on the nationalists' persecution of the "enemies of Ukraine" amongst independent (and critical of the regime) mass media, businesses and social media influencers. The UDA informants were open about their cooperation with the government: "Much of info on 'Russkii mir' agents 80 we receive from them [government]. They want us to go and ravage [koshmarit'] these people. They cannot do that themselves, because, you know… Ukraine is now [supposed to be] Europe. What are the 'big, white people' 81 in Brussels will think of Zelya [Zelensky]?!" 82 Another informant hypothesized: "When we were at our bases in Donbas, they [government] couldn't use us on the streets. Now, when we are in reserve, we are available full time. So demobilization wasn't only useful for the president to jumpstart Normandy talks, but also to 'clean up' the streets." 83 Those informants who either participated in "activist work" of persecuting "enemies of Ukraine," or had friends with such an experience, mentioned that there is a reward of USD20 to USD100 per person for each "activity," and that this reward "does not come from the UDA budget." 84 An independent source 85 confirmed that unlike other ultranationalist groups (for example C-14) that are mostly staffed by teenagers, UDA members are war veterans with actual combat experience, and are highly preferred by "the government clients" willing to pay hard cash for "street activism." The recent rise of pro-Russian-speaking political parties in 2020, such as "OPZZh" (Opposition Platform for Life) and the "Party of Shariy", challenging Zelensky's government and securing high levels of popular support in Russian-speaking southern and eastern regions, have further exacerbated the government's covert efforts of deploying paramilitaries on the streets. Similarly to the official government, the UDA describe opposition parties as the "Kremlin's agents" acting against the national interests of Ukraine. 86 Notwithstanding the UDA's hostility toward Zelensky's government, the group admits that "we have common enemies. Does that make us friends? No, but perhaps, allies. " 87 For example, during the recent regional elections in 2020, the activists of "Party of Shariy" 88 were violently attacked by the former DUK/UDA members wielding metal rods and knives. 89 MPs from Zelensky' ruling party "Servants of the People" echoed the ultranationalists' call to forbid opposition parties from legally operating in the country. 90 The UDA's "marriage of interests" with the regime is not limited to the deployment of (ex)paramilitaries by the government against its political opponents, but also stems from the wider significance of the "volunteer" community for national security. In the words of an investigative journalist who explored this issue in-depth: "the [ultranationalist] volunteers are 'preserved' by the government in order to be re-deployed to Donbas in case of another full-blown conflict with Russia." 91 The informant admitted that despite "reforms in the army, … conscription numbers remain shockingly low, and 'exploiting' a nationalist theme remains the only opportunity for the regime to attract people to the army in case of war." In view of the government's coalescing political and ideological interests with UDA, the demobilization of paramilitaries resembles more of a marriage of convenience rather than genuine efforts of dismantling unwanted non-state armed groups. Fragmentation of the means of violence, which the government laments about but also seems to benefit from, is engrained into the balance of capabilities that UDA effectively maintains vis-à-vis the government, and is sealed through common existential interests. The ascent to power of the pro-Russian-speaking opposition, supported by the majority of public in southern, eastern, and to a lesser degree central regions of Ukraine, would mark the end of Zelensky's government and the political collapse of far-right parties, broadly symbolized by Dmitro Yarosh. Both the government in Kyiv and UDA are therefore keen to preserve the current status quo.

Discussion and Conclusion
This study has proposed a set of theoretical expectations which seek to explain cases of incomplete demobilization by pro-state militias following civil wars. While the state will generally seek to co-opt or dismantle non-state groups, there are many cases where this proves difficult, leading to continued, if perhaps covert, mobilization. This outcome involves three key processes. First, there is a fragmentation of the means of violence during the conflict, which persists into the post-conflict period. This includes multiple pro-state groups emerging, as well as potentially rivalries between different arms of the state. Secondly, one of several pro-state militias are able to leverage their material strength, political connections and public legitimacy to resist efforts to fully demobilize them, leading to a relative balance of capabilities. Thirdly, the pro-state militia and the state will remain bound together by a balance of economic and political interests, not least of which is the continued threat posed by their wartime opponents.
Our empirical research demonstrates that Ukraine is undoubtedly a case of incomplete demobilization. In the UDA we see an actor which, while formally demobilized, has the organizational capacity, manpower, equipment and political coherence to swiftly remobilise. It also provides a strong illustration of our theory. The prevalence and military significance of non-state groups as well as the fragmentation of the state's security apparatus was a key factor in allowing the UDA to rise to a position of strength and importance. The UDA remained militarily strong, politically well-connected and broadly popular enough to effectively resist demobilization, even after formally agreeing to give up its arms. Its members claim that they were further able to use clandestine methods to maintain its mobilization potential, including maintaining secret arms caches and infiltrating members into the state security services. Finally, the UDA has not only developed shared interests with sympathetic politicians and businessmen, but also shares the government's antagonistic position on domestic Russian-speaking political movements as well as Russia itself. This has led to government to contract out functions of violence and intimidation to the UDA, even as it seeks to undermine the group's political influence. This means that, despite being strongly critical of the current government, the UDA finds itself in a marriage of convenience based on a balance of power and interests.
In addition to providing a conceptualization of incomplete demobilization and a theoretical framework for understanding it, this article also has clear policy implications. Firstly, it emphasizes a point made throughout the literature about the potential dangers of relying on pro-state militias. 92 Beyond this, the case of Ukraine shows the importance of states maintaining internal unity, particularly between different parts of the security apparatus. The Ukrainian state suffered not only from weak capacity vis-à-vis Russia, but subsequently saw its efforts at establishing domestic control hamstrung by a lack of unity. This suggests a need to focus on the internal coherence of states, as well as capacity. Finally, our theory and the case of Ukraine point to a need to move quickly to establish control over pro-state armed groups, whether this is achieved through co-option or coercion. The longer such groups are allowed to thrive, the more embedded they will become in the social and political fabric of a country. Patrimonial politics and shared security interests means that such groups are liable to become harder to dislodge over time.
While the theoretical and policy insights developed within this article are potentially transferrable to other cases, there are limitations to both its scope and generalisability. Firstly, the focus here has been on pro-state militias, which are at arms-length from the government as a whole. Cases involving groups which are directly tied to presidents or political parties may show different dynamics. Secondly, a single case study is inherently limited in its generalisability. Nevertheless, this article has clearly set out the key causal processes underlying incomplete demobilization and can provide the basis for future empirical work.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
The collection of the empirical data used in this paper was supported by the British Academy Small Research Grants scheme SRG1819/190334.