Introduction

Pro-government militias (PGMs) are a prominent feature of civil wars. In Iraq, the United States sponsored the Sunni Awakening groups to fight against insurgents. Afghanistan formed militias to battle the Taliban. Over the course of the Syrian civil war, the number of militias fighting for Bashar al-Assad has increased and includes Shia militias supported by Iran and Hizbullah from Lebanon (see, e.g., Leenders & Giustozzi, 2019). While religious cleavages shape the supply of irregular armed groups in the Middle East, these groups can be found in counterinsurgency (COIN) campaigns across the globe. In eastern Ukraine, pro-Ukraine militias fight against pro-Russia armed separatists; in northern Nigeria the Civilian Joint Task Force battled Boko Haram; and in Colombia rural defense groups assisted the government against FARC (the Revolutionary Armed Forces of Colombia, or, in Spanish, Fuerzas Armadas Revolucionarias de Colombia). While gaining increasing attention in the scholarly literature, using irregular forces against insurgents has a long history and was a common element of colonial security sectors (Anderson, 2005; Janowitz, 1977; Thompson, 1966b). The British used Assyrian militias in Iraq to assert authority under the auspices of the League of Nations mandate for the territory (Ahram, 2011, p. 62). They also employed auxiliaries in Ireland, many of whom went on to serve in Palestine in the 1920s, and they created Home Guards in Malaya and Kenya to defeat insurgencies in the 1950s (Anderson, 2005; Branch, 2009). In fact, many elements of COIN strategies “call for the use of local militias to extend armed presence or allow locals to have a stake in their own security” (Paul et al., 2010, p. 62).

Responding to rebel threats with this sort of public–private collaboration occurs across conflict types, rebel actors, and regime types. Assuming a unitary or “Weberian” state actor masks the creativity of governments in combatting insurgents and safeguarding against elite threats (by designing security forces as a coup-proofing strategy, see De Bruin, 2020). Across various research fields, scholars are increasingly attentive to the activities of armed groups recruited or induced to side with the state (e.g., Blocq, 2014; Jentzsch, 2017; Jentzsch et al., 2015; Kowalewski, 1992; Lyall, 2010; Stanton, 2015). Attention to these organizations mirrors the wider academic interest in nonstate actors and in “not entirely state” and “not entirely private” (Ostrom, 1990) solutions to policy problems. It is an analytically rich seam of questions and controversies.

The article reviews the research on the role of irregular pro-government armed groups in civil wars (for a recent review of the civil war literature, see Cederman & Vogt, 2017; for civil–military relations, see Brooks, 2019). After defining PGMs, it examines why governments use militias in COIN campaigns, despite the risks of outsourcing violence. To understand the use of these groups, both the demand and supply sides require attention. Using descriptive statistics, the article overlays PGM presence on various categories of civil war, building upon the literature on classifying civil wars. It notes that governments use militias both inside and outside civil wars (see also Raleigh & Kishi, 2020). The discussion then focuses on the use of militias in armed conflicts and concludes with some challenges for future research.

Pro-Government Militias in the Context of Armed Conflict

Some scholars enlarge the scope of militias to include nonstate actors without specific links to government. In a special issue on militias, Jentzsch et al. (2015) used the term militia for an armed nonstate actor that is “anti-rebel.” Militias sometimes also include rebel groups (e.g., Bates, 2008; Raleigh, 2016). Instead of focusing on the link to government, other classifications claim that the link to political actors is analytically more useful, as groups may fight for the opposition as well (Schuberth, 2015, p. 306). With changes of government, groups linked to political parties, for example, may no longer be pro-government (Otto et al., 2020). But the puzzle remains of why governments, with regular forces available, continue to rely on often quite ill-equipped nonstate armed groups. The relationship between the government and militias is diverse and very dynamic, ranging from incorporation to suppression (Staniland, 2015). This review of PGMs fits best Staniland’s typological category of “collusion.” The review is limited to armed groups that contribute to COIN campaigns and that have a link to the government beyond sharing an enemy. Private armed groups of landlords or criminal gangs are excluded, unless there is a link to either subnational or national government under civil war conditions.

We define militias as armed groups linked to the government and separate from the regular forces but limit the focus here to those operating within the context of COIN campaigns and civil wars (Carey et al., 2013, p. 250). The term paramilitary may include these groups but is also often inclusive of regular professional units such as police forces and border guards (Böhmelt & Clayton, 2018; Dowdle, 2007). Like Jentzsch et al. (2015), we exclude private security firms due to their status as commercial contractors. To further distinguish private security firms from militias, Jentzsch et al. added territorial conditions, noting that in contrast to military contractors, militias tend to operate at home rather than abroad. We relax the territorial condition and consider militias operating both within and across borders in the context of conflict. But the incentives for governments to contract with these firms (Avant, 2005) may be similar to those that motivate alignment with militias. Our definition also includes surrogate forces, which Hughes and Tripodi (2009) divided into home guards, militias, counter-gangs, and pseudo-gangs. According to their classification, pseudo-gangs are government-sponsored, supposedly independent groups that target rebels or terrorists, while counter-gangs (a term for co-ethnic armed groups used in defeating the Mau Mau in Kenya) are lightly armed, mobile gangs, often consisting of former rebels, fighting insurgencies. Militias are identified as larger mobile forces aligned with the government, while they define home guards in terms of the “static defence of villages” (Hughes & Tripodi, 2009, p. 9). Similarly, Clayton and Thomson (2014) and Peic (2014) identified civil defense forces as armed groups that recruit civilians or former rebels, operate locally, and are defensive rather than offensive. This work highlights the significance of ties to the civilian population and whether these militias are local or community organizations (Daly, 2016; Jentzsch, 2022).

Why do governments create or align with irregular forces when fighting a rebel group? Why do governments not merely strengthen regular forces, which are likely to be better equipped, better trained, and more disciplined? Hughes and Tripodi (2009) point out the disadvantages of militias. For example, when confronted with a strong rebellion, the government might depend on the militia, thus affording them greater bargaining power and potentially greater independent resources (Bolte, 2021), making it harder for the government to contain them in the future. Militias are often characterized by internal strife and may lack discipline, accountability, and reliability. Their agendas or ideological preferences may differ from those of the government (Hughes & Tripodi, 2009). Yet despite these disadvantages, between 1981 and 2014 over 80% of country-years with armed conflict generating at least 25 battle-related deaths included PGMs.¹ They are not a phenomenon confined to the post–Cold War era or restricted to certain continents. We review the arguments offered to explain the use of PGMs in civil wars.

Why Use Militias to Fight Insurgents?

Types of Civil Wars and PGMs

Governmental and Territorial Conflicts

The UCDP/PRIO Armed Conflict Dataset version 20.1 (Gleditsch et al., 2002; Pettersson & Öberg, 2020) distinguishes between two types of civil wars depending on what drives the conflict. Governmental conflicts are about the government, for example, the political system or the replacement of the central government; territorial conflicts are about the insurgents’ demands for territory. Figure 1 shows the percentages of governmental conflicts per year that had at least one PGM, using information from the Pro-Government Militias Database (PGMD) version 2.0 ((Carey & Mitchell, 2017). The graph shows percentages of governmental conflicts with any form of PGM and distinguishes between those with informal PGMs, which have only a loose link to the government, and semi-official PGMs, which are officially recognized and often formally included within the structure of the security forces, while being separate from regular military or police (Carey et al., 2013).

Open in new tab

Figure 1. Pro-government militias in governmental conflicts, 1981–2014.

The proportion of governmental conflicts with any kind of PGMs fluctuates around 80%, with the smallest share of 63% in 2009 and the highest in 1988 with 94%. The share of governmental conflicts with informal PGMs has usually been lower compared with those with semi-official PGMs, though this pattern has been changing since 2010, with informal PGMs becoming more widespread within such conflicts, while the use of semi-official PGMs seems to have flattened at a much lower level.

Figure 2 plots the proportion of territorial conflicts with PGMs between 1981 and 2014. For these types of conflicts, the share with informal PGMs has been on a constant upward trend and has caught up with the share of territorial conflicts in which governments use semi-official PGMs. In the past, governments seem to have been less willing to rely on irregular forces as they pose the most severe delegation and control problems. Additionally, semi-official PGMs allow the government to publicly demonstrate a local or ethnic component to COIN that can be useful in tackling the secessionist struggle.

Open in new tab

Figure 2. Pro-government militias in territorial conflicts, 1981–2014.

Counterinsurgencies

Instead of using what the government and rebel groups fight over, armed conflicts can also be categorized by the distribution of power between the two warring parties (Kalyvas & Balcells, 2010). Counterinsurgencies are characterized by unequal distribution of military capabilities, where a much weaker rebel group faces a militarily superior state (Kalyvas & Balcells, 2010). In such conflicts, access to local information and the ability to blend in with civilians is crucial. Militias with local ties to the communities should deliver the government an important advantage in the context of COIN wars.

Lyall and Wilson (2009) explain the sharply decreasing success of 20th-century governments in fighting COINs with the introduction of the tank in World War I. Mechanized armies became isolated from the local population in contrast to the forager armies of the 19th century. These forager armies were “capable of identifying potential collaborators and of skillfully dividing insurgent ranks: witness … the widespread practice of creating proxy forces staffed by locals” (Lyall & Wilson, 2009, pp. 73–74). This argument is based on the assumption that armies, or armed groups, with access to local networks and knowledge give the government an important edge to defeat insurgents.

We use the data from Lyall and Wilson (2009) to identify counterinsurgencies as civil wars in which small, mobile insurgent groups use guerrilla attacks and try to win over parts of the civilian population to defeat the government.² These data are then supplemented with information from the PGMD (Carey et al., 2013) to identify whether the government collaborated with an armed nonstate actor. As the PGMD only goes back to 1981, additional data was collected on the presence of such PGMs back to 1945. We identify whether the militia was recruited and employed locally and whether the irregular armed group had an informal or semi-official link to the government.

Figure 3 displays the proportion of COIN wars (1950–2005) with PGMs participating.³ These new data show that proxies, including local armed groups that support the government, survived mechanization and remained a standard feature of COIN throughout the post–World War II period. If the widespread practice of using local proxies survived mechanization, what could explain the declining effectiveness of government in COIN? Maybe governments create or align with these irregular forces in particularly difficult to resolve wars, after their military has failed to defeat the insurgents.

Open in new tab

Figure 3. Pro-government militias in counterinsurgency wars, 1950–2005.

While a systematic analysis of a potential selection bias of the COIN wars in which militias appear is beyond the scope of this article, Figure 4 shows at what point in the conflict these forces first appear. In the vast majority of COIN wars, PGMs are present already at the very beginning of the conflict. This contradicts earlier data that suggest that militias are usually not present at the beginning of the conflict but on average appear four years into the conflict (Peic, 2014, p. 175). What explains these patterns? Are governments that already collaborate with irregular forces more likely to initiate armed conflicts or are they more likely to be targeted by insurgents? Do governments establish these connections once a conflict has broken out but before it has escalated into a major war with more than 1,000 battle-related deaths? Future research has important questions to tackle on the life cycle of PGMs and the dynamics of armed conflicts.

Open in new tab

Figure 4. The timing of PGM occurrence during counterinsurgency wars, 1945–2005.

How Effective Are Militias, and What Price Do Civilians Pay?

Many conflicts feature PGMs, but do they provide an advantage to governments and prevent an outright loss to the rebels, or can they turn a conflict into a swift government victory? The collaboration between the Awakening militias in Iraq and U.S. forces provides an example where highly professional forces benefit from the knowledge and the skills of local irregular forces. In their analysis of the success of the Iraq surge, Biddle et al. (2012) argue that neither the irregular forces nor the American troop surge on its own was sufficient to reduce the violence in Iraq in 2007. They identify the intelligence on the insurgency collected from the Awakening groups, the fact that they were no longer fighting against the coalition, and the consequences this alliance had for Shiite militias as their principal contributions to the success of the surge. But these militias were also linked with rising levels of violence against civilians (Clayton & Thomson, 2014). Biddle et al. (2012) contrast the forces in Iraq with those in Afghanistan. They argue that the militia forces created by the coalition, the Afghan local police, were unable to recruit former rebels—which prevented a similarly successful outcome. Algeria under the French, Kenya under the British, or more recently Kashmir, Sri Lanka, and Chechnya, all saw former rebels organized to fight alongside regular forces to reverse rebel gains. In examining the use of militias in South Asia and elsewhere, Biberman (2019) presents a mixed picture of their effectiveness.

Using micro-level data from the Philippines between 2001 and 2004, Felter (2009) compares the military effectiveness of elite forces, regular forces, and civil defense forces. He evaluates effectiveness along four dimensions: striking targets effectively with minimum collateral damage, central monitoring with decentralized execution, access to local information, and signaling credibility in victory. He shows that the civilian defense forces, the Civilian Armed Forces Geographical Units, performed worse than the other types of forces across all categories. Local knowledge appears to lead to feuds and vendettas and makes them targets for rebel violence. As a result, they suffer much higher casualties than regular or elite forces, even controlling for their smaller size and possible differences in deployment strategies.

Additionally, Felter (2009) shows that the proportion of firearms that were lost is much higher among indigenous forces compared with regular and elite forces. This finding highlights the risks of yielding a monopoly of violence. Felter explains these failures as the result of poor leadership and supervision. When elite forces lead the indigenous forces, they become far more effective than either the elite forces or the regular forces (Felter, 2009, p. 26). Similarly, in Algeria, using local forces to find the guerrillas and French paratroopers to kill them was an effective approach: “within months the ALN had lost half of its soldiers killed, captured, or converted to the French cause” (Evans, 2012, p. 245). These findings suggest that the qualities that irregular forces bring to COIN can be more fully exploited when problems of delegation are reduced by leading them with highly trained and specialized forces.

Analyzing 250 civil war episodes between 1991 and 2015, Aliyev (2020a) confirms the negative impact of delegation problems on the ability to end the fighting and suggests that PGMs are associated with longer civil wars. Acting as veto players (Cunningham, 2006), they complicate the resolution of the conflict and therefore drag out its conclusion. Our data on COIN wars between 1945 and 2005 supports this finding. Figure 5 plots Kaplan–Meier survival functions for COIN wars with and without local PGMs.⁴ Locally recruited and active militias are supposed to be particularly advantageous for the government. Yet they not only fail to bring a swift conclusion of the war, but they are systematically linked to periods of fighting. PGMs can exploit an ongoing war to loot for self-enrichment without being constrained by the government, and they may be able to pursue their own violent tendencies or strong convictions that drive the conflict (Steinert et al., 2019) and therefore be less willing to accept its conclusion.

Open in new tab

Figure 5. Counterinsurgency war duration with and without local PGMs, 1945–2005.

Analyzing irregular civil wars between 1944 and 2006, Peic’s (2014, p. 171) analyses suggest that using civil defense forces substantially improves governments’ chances of defeating an insurgency. Analyzing 30 counterinsurgencies, Paul et al. (2010, p. 63) come to a less optimistic conclusion. They show that 25 of them used militias or “community policing.” While 19 resulted in COIN losses, the authors caution that the use of militias alone cannot account for the losses. In 15 of these 25 COIN operations, militias were at cross-purposes, and 13 of those 15 ended in COIN losses (Paul et al., 2010, p. 63).

Figure 6 uses our data on COIN wars between 1945 and 2005 to show the share of such wars with and without (local) PGMs that ended in a loss, draw, or win for the government. The percentage of wars that end in a government loss is almost identical for COIN wars with and without any PGM or with local PGMs. Irregular forces that fight alongside the government, even if they possess local knowledge, do not seem any more able to avoid a defeat by insurgents. The chances of a government victory seem to be somewhat lower with than without PGMs. The share of wars that end in a draw is somewhat higher when local PGMs, or any type of PGM, is involved in the fighting. Local or community-based militias may not increase the winning percentage for governments, but they make draws somewhat more likely.

Open in new tab

Figure 6. The outcome of counterinsurgency wars with and without PGMs, 1945–2005.

These insights point at the core problems associated with delegation. Multiple agents complicate monitoring and likely increase discretion. Discretion can be an advantage to the principal in allowing agents to respond effectively to local conditions—what Donahue and Zeckhauser (2011) term “production discretion” (p. 50)—but it becomes a disadvantage when discretion permits opportunism and shirking. Militias are likely more prone than regular forces to pursue private interests within the governments’ COIN operations, and they may not benefit from a clear resolution of the conflict.

What price do civilians have to pay when PGMs get involved and pursue their own interests?

The literature provides a mixed picture of PGM violence against civilians. Analyzing the conflict in Chechnya, Lyall (2010) shows the tactical effectiveness of using co-ethnic, locally informed militias in comparison with regular Russian troops. His findings suggest that these groups contributed to regime security. It is less clear how to assess the comparative harm done to the local population by the different types of forces. While Lyall (2010, p. 14) finds Russian forces killed and disappeared more individuals, the militias kidnapped greater numbers. He quotes a local nongovernmental organization representative: “The kadyrovtsy are much more dangerous for local residents in terms of persecuting entire families or kidnapping individual relatives … The federal [Russian] troops simply don’t have such complete information about the local residents” (Lyall, 2010, p. 15). Šmíd and Mareš’s (2015, p. 15) qualitative case study of the Chechnya conflict also finds that Putin’s policy of “Chechenizing” the conflict was successful. Russia enlisted Chechen clans as paramilitary forces but, according to their account, at the price of violent excesses.

With some nuance depending on the type of militia involved, a body of research suggests that militia forces put civilians at risk (Aliyev, 2020b). In defeating the Mau Mau insurgency in Kenya, these local forces inflicted significant harm on civilians: “[t]he Home Guard thus represented state-sanctioned and sponsored terror intended to bring the population of Central Kenya under the control of the incumbent regime” (Branch, 2009, p. 87). The British government also recruited “counter-gangs” or “pseudo-gangs” of former insurgents to operate alongside the Home Guard and regular forces. “Security Forces had better weapons … but they had firmly fastened one of their hands behind their back with a cord of legal difficulties which was not the case with the irregular forces” (Kitson, 1960, pp. 44–46). In Mexico in the 1990s, ruling party-based militias are reported to have been responsible for 15,000 murders in the wake of the Zapatista rebellion (Mazzei, 2009, p. 25). With cross-national conflict data, Cohen and Nordås (2015) suggest that militias that recruit children are linked to higher levels of sexual violence. Stanton (2015) argues that if militias and insurgents come from the same constituency, then militias are less likely to attack civilians. Comparing different types of militias in East Timor, Barter (2013) concludes that militias that were created by the government were more likely to attack civilians, while those that developed in self-defense against powerful rebels were more likely to behave defensively. Magid and Schon (2018) and Schon and Magid (2021) collected data on the ethnic ties of PGMs in Africa to examine the influence on the well-being of civilians.

Using irregular forces thus carries significant risks. Across a large range of specific indicators, Kowalewski (1992) finds support for what he terms the “critical view of paramilitaries” (pp. 79–80). Other work suggests that PGMs increase the likelihood of “agent-centered violations” such as extrajudicial killings, torture, and disappearances. These are violations that fall within the capacity of individual agents and from which there may be a plausible private benefit, be it revenge, extortion, or some gratification from the violence itself (Mitchell et al., 2014). Irregular armed groups are likely to be less well equipped to train recruits and inculcate codes of conduct. Furthermore, governments may be tempted to use these groups to shift blame and alleviate legal difficulties for controversial violence (Carey et al., 2015). In contrast, some research challenges the usefulness of a principal–agent approach for understanding the violence committed by militias, arguing that social cohesion and diffusion (Cohen & Nordås, 2015) and strategic decisions (Stanton, 2015) provide better explanations. Park and Sim’s (2021) analysis of Indonesian counterinsurgencies finds that militias were responsible for much of the sexual violence in East Timor and Aceh but less so in West Papua. As empirical measures improve, a clearer picture should emerge of levels of “agent-centered” violence attributable to different types of armed group organizations and how well these fit with the implications of a principal–agent approach.

The use of excessive violence can be counterproductive in fighting insurgents (Daxecker, 2017; Lafree et al., 2009). Paul et al. (2010) caution governments to ensure that militias do not resort to “disproportionate or illegitimate uses of force” (p. xxiv). Examining civilian self-defense militias in Colombia, Estancona et al. (2019) argue that governments, sensitive to the control problems posed by militias, selectively use these groups as a means to influence the local population. The divergent literature on the impact of government militias on human rights and civilian welfare, and on COIN success or state failure, suggests that investigating the consequences of these groups is an important component of the research agenda.

>Finally, there is a developing interest in forecasting conflict and the effect on civilians. Forecasting models predicting the onset of conflict rely on demographic, economic, and educational variables (e.g., Hegre et al., 2021). But recent work has offered an actor-centric approach to predict the severity of conflict that is based on the organization and behavior of those on the rebel side who actually carry out the violence (Metternich et al., 2019). On the government side, Koren (2017) uses PGMs as a predictive indicator of mass killing by states. And van der Maat (2020, p. 794) finds that militias were present in all cases of what he terms “genocidal consolidation” from 1950 to 2004.

PGMs in the Aftermath of Conflict

What happens to PGMs once the conflict has ended? How do they affect post-conflict stability and security? Mueller (2003) argues that particularly in weak states, armed groups often engage in criminal activities during civil wars—even more so when irregular armed groups such as mercenary forces or death squads are used. Focusing on the disarmament and demobilization of militias in Afghanistan, Giustozzi (2008) points out how difficult it is to effectively disarm and delegitimize militias. PGMs may become the private armed forces of warlords (Marten, 2006). Wehrey (2012) describes the conundrum posed by militias in Libya. After the elections in June 2012, militias were used as hired guns while disarmament processes were underway. The failure to disarm these forces seriously affected the stability not just of the country, but of the whole region. Mueller (2003) suggests that “the key to controlling the remnants of war is the establishment of competent domestic military and policing forces” (p. 511). This may be particularly difficult at the end of a civil war fought by a diversified security sector.

Civil war may provide these groups with opportunities for private gain (Hughes & Tripodi, 2009), while peace limits opportunity. Furthermore, some militias have an ethnic membership and the ethnic basis of conflict is an important factor for the duration of a conflict (Wucherpfennig et al., 2012). Ethnically defined militias may polarize ethnic groups, which creates long-term mistrust that further complicates post-conflict reconciliation. We know from the literature on rebel groups, having multiple actors likely makes it harder to achieve a peace agreement and end the fighting (Bakke et al., 2012). If they benefit from ongoing conflict, militias may act as spoilers to peace agreements. In short, there are reasons to expect, and the evidence to support it, that the presence of PGMs will make a conflict last longer. In addition, their presence may produce increased levels of violence and abuse and make the post-conflict period more volatile.

Scholars have begun to investigate the impact of these groups on conflict duration and intensity and to explore a range of mechanisms. Abbs et al. (2020) find that PGMs that share ethnic ties with the ruling elite make for a more loyal force, reducing principal–agent problems but lengthening the conflict and increasing its intensity. Aliyev (2020a) argues that militias have an interest in continuing conflict and add to the number of veto players and similarly finds militias lengthen conflict duration (but see also Balta et al., 2020). Steinert et al. (2019) show that when PGMs participate in a civil war, the conflict is more likely to reescalate than if no PGMs were involved. Militias are rarely involved in peace negotiations, and they are often excluded from disarmament and integration programs. Without attractive options in peacetime, militias have ample incentives to reignite the civil war (Steinert et al., 2019). Even without a relapse into civil war, wartime militias are associated with increased state repression in the post-war period, contributing to increased insecurity and violence (Carey & González, 2021). Examining the post-war fates of militias, Bolte et al. (2021) point to the importance of ethnic ties and military capability in how governments either terminate or integrate these armed groups.

Putting civilian defense under the leadership of regular forces may provide a way of harnessing the advantages of irregular armed forces while reducing the risks associated with them. But even in circumstances where militias act in concert with highly professional regular forces, they may bring substantial harm to the civilian population, as with the Israeli army’s use of the Lebanese Phalange militia in Lebanon in 1982, or the use of these forces by the Americans in Iraq or the British in Kenya.

Open Questions

Research focusing on the actors involved in conflict, not just the structural conditions they face, and abandoning the assumption of unified forces on either side provides us with a richer understanding of the dynamics of civil war. Initial insights into the role of irregular forces aligned with the government, piecing together evidence from a disparate case literature and quantitative cross-national studies, opens up new questions for future research. As in other research fields, an important first step is to sort out theoretically meaningful dimensions for classifying these organizations (Carey & Mitchell, 2017). While militias are found in conflicts around the world, they vary in their ties to governments or particular leaders; in their visible or clandestine activities; in their recruitment and connections to other societal, political, or economic organizations; in their size; and in their national or local presence. What explains these different characteristics of PGMs? And, while there has been some progress in examining the role of ethnic ties and the influence of local or community-based groups, how do these characteristics influence the effectiveness and long-term consequences of militias on civilian well-being, the severity of conflict, stability, and peace? While case studies have provided us with important insights into these topics, more research using different approaches and covering different areas across time and space is needed for a deeper understanding of the roles of PGMs in conflict and why these proxy groups persist, given their relationship with genocide and their mixed contribution to COIN success.

We need to know more about the details of the “contract” between these groups and their respective governments. For political scientists, the structure of the relationship between the state and its armed nonstate actors, and the presence or absence of monitoring mechanisms and sanctions, suggests the likelihood of “adventurous” policy and the willingness of a government to evade accountability. It is necessary to understand the lines of communication between groups and government. Is there a functional equivalent of the chain of command found in regular forces? When are militia members most likely to shirk? These questions are important for research on global accountability, the state’s responsibility to protect, and compliance and accountability issues for the international community (Grant & Keohane, 2005). Thomson (1994) describes the 1856 Paris Declaration by which the major powers agreed to end their practice of authorizing pirates to fight and loot on behalf of the state. There appears to be little progress in that direction for state authorization of irregular forces that fight on land. A ruling by the International Criminal Tribunal for the former Yugoslavia (ICTY) involving Arkan’s Tigers suggests that state officials, even if they train and arm these groups, may be freed of responsibility for their criminal conduct.

A better understanding of when these groups enter and exit a conflict is highly relevant for policymakers. Given the risks entailed, the expectation is that use of these groups is a desperate measure (Downes, 2006) or a last resort (Peic, 2014). When do governments turn to these groups? Our data on counterinsurgencies suggest that militias are not a last resort but are present from the outset of the conflict. What role do loyalty, ideology, revenge, material gain, desperation, and coercion play in their recruitment and type of membership (Staniland, 2014; Weinstein, 2007)? The “supply” of militias requires further research, including on what the exit of these groups entails. Are they reintegrated? Does their violence shift to criminality? What is their legacy? Bolte (2021) introduces a bargaining framework to examine the trade-offs facing governments and the problem of how to manage them when they have outlived their usefulness—when the rebel threat declines. An innovative subnational study of post-conflict violence in Guatemala suggests these groups persist beyond their wartime organization (Bateson, 2017, 2021). Survey and interview evidence from Colombia suggests that the local or community-based nature of these groups contributes to their longevity (Daly, 2016). Ties to a community, which may make monitoring and control easier, can help scholars understand their behavior during and after conflicts.

Our theories of the state suggest that militia groups are a thing of the past and that security is a sovereign task. But sovereign tasks are less preciously held than anticipated. On ill-defined battlegrounds around the world, these groups continue to offer services to governments. They lessen the material and political costs of conflict. While the short-term advantages to a particular government may be apparent, researchers need to gather better data and improve the theoretical and empirical understanding of how public–private collaboration in the security area contributes to core public goods.