How do rebellions begin? Despite a large literature on civil war onset, until recently there has been limited systematic study of the initial start of nonstate organized violence, especially in states’ rural peripheries. Yet about two-thirds of large-scale civil wars since 1944 begin with peripheral challengers (Siberdt Reference Siberdt2024). The lack of state, media, and infrastructure presence that makes rural peripheries hard to study also makes them likely sites for illicit nonstate armed group formation.
This paper probes two patterns of armed group formation in rural peripheries that common approaches to conceptualizing armed conflict onset have often overlooked and cannot easily explain. First, as we show below, evidence from new datasets on nonstate armed group formation indicates that such groups often start slowly, committing low levels of violence for months or years before, and if, they escalate into large-scale civil war. Yet even small-scale violence is a visible and thus relatively risky act for a vulnerable, nascent armed group. Why not wait to commit violence until one’s group is ready to inflict large-scale violence and thus have a higher likelihood of gaining concessions from a state?
Second, even anti-state armed groups—“rebels” or “insurgents”—that later become known for their ideological or political commitments are often initially difficult for states, journalists, and researchers to distinguish from violent “bandits” or “criminals,” labels that connote private, economic motivations.Footnote 1 For example, the ongoing Islamist insurgency in northern Mozambique began with months of small-scale attacks for which the insurgents did not claim credit, and which the government initially dismissed as criminality (Bekoe, Burchard, and Daly Reference Bekoe, Burchard and Daly2020, 7–12). Similarly, in the years leading up to El Salvador’s civil war, the predecessor groups of the Farabundo Martí National Liberation Front (FMLN) committed mostly small-scale violent acts for several years after they first formed (Kruijt Reference Kruijt2008, 84). Observers struggled to tell whether the acts were committed by “guerrillas, common criminals, or moonlighting military officers” (quoted in Crandall Reference Crandall2016, 53). This type of confusion was also evident, for example, in 1980s Aceh in the lead-up to the Free Aceh Movement (Gerakan Aceh Merdeka, GAM) insurgency (Barter Reference Barter2014, 45–46), the Mexican Revolution (Brunk Reference Brunk1996, 332–34), peasant uprisings in colonial India (Guha Reference Guha1999, 79–81, 98–101), the Russian Civil War (Landis Reference Landis2008, 41–59), the Arab revolt in British Mandatory Palestine (Hughes Reference Hughes2019, 129–30), the Indonesian War of Independence (Cribb Reference Cribb1992), the Greek Civil War (Gerolymatos Reference Gerolymatos2004, 206), the Algerian War of Independence (Horne [Reference Horne1977] 2006, 78, 88–89), the Shifta rebellions in Kenya (Whittaker Reference Whittaker2015, 50–52) and Eritrea (Ertola Reference Ertola2023, 84–85), as well as Uganda in the late 1980s and 1990s (Lewis Reference Lewis2020, 107–23). We demonstrate the difficulty outside actors have faced in discerning the motives of nascent armed groups in 15 additional cases—three discussed in this paper and 12 in the Supplementary Material.Footnote 2 The ubiquity of these patterns in case study evidence suggests that it is not rare.
We provide an explanation for why this small-scale violence occurs in rural peripheries before anti-state armed groups escalate to large-scale political violence, why it is often difficult for outsiders to distinguish it from “criminality,” and what these patterns suggest about promising future directions for research. Our logic contrasts the initial uses of armed group violence with the later phases that are more typically studied. Building on the state formation literature as well as ethnographic and historical accounts, we argue that in rural peripheries with minimal state monitoring—where barriers to armed group formation are low—nonstate armed group formation of all kinds is more common than is usually appreciated in the social science literature on civil conflict onset. Under these conditions, violence that is small in scale and anonymous is useful for a variety of newly formed armed groups: it allows them an opportunity to gain needed weapons and a chance to test the state’s local coercive and intelligence capacity, while maintaining civilian acquiescence to local nonstate violence.
We argue that because these benefits and mitigation of risk are relevant for nonstate armed groups with a broad range of motivations, the purpose of the violence is ambiguous from the standpoint of outsiders like the state. This ambiguity thus presents a significant opportunity for groups aspiring to rebellion, since states cannot discern the threat they pose. Also, in rural peripheries, the common proximity or recent legacy of nonstate armed groups that do not seek to extract political concessions—“bandits”—makes the skill set of organizing small-scale, anonymous violence more readily available to nascent rebels.
We argue further that there is an observable divergence in forms of violence usually labeled “criminality” or “banditry” from that of “rebellion” if and after an armed group opts to formalize, train, and build their forces to challenge the state more directly and offensively. Doing so is necessary to challenge the political status quo in newly contested territory; this competitive state building is the sine qua non of rebellion (Kalyvas Reference Kalyvas2006, 218). In contrast, it is unusual for criminal groups, even violent cartels, to engage in substantial anti-state violence. Most large-scale criminal violence occurs between cartels, and when it is anti-state, it is typically defensive (Lessing Reference Lessing2017, 3, 28–29, 45–47). Crucially, this divergence indicates why scholars often overlook the ambiguity of violence and the role of “criminality” in the initial phases of rebellion: by the time armed groups commit sufficiently large-scale violence to gain substantial attention from the state and media and cross thresholds for inclusion in standard conflict datasets, the ambiguity and perceived criminality usually no longer exists. Case studies of rebellion also do not typically focus on hazy early events that were not explicitly tied to anti-state rebellion.
To demonstrate the empirical promise of our explanation for why violence committed by groups later recognized as rebels often initially takes an ambiguous form attributed to “criminality”—before later shifting to more overtly political violence that is typically larger in scale—we closely examine nonstate armed group formation and the outset of violence in three major episodes of civil war in the twentieth century: the Chinese Communist Party (CCP) in Republican China (1912–49); the Việt Minh in the First Indochina War (1946–54); and the 26th of July Movement (M26) in the Cuban Revolution (1953–59). To demonstrate the broader empirical reach of our arguments, in the Supplementary Material we analyze 12 additional cases from a broad range of historical and contemporary contexts. The case studies aim to illustrate the basic claims that underlie our arguments—especially our claims about the within-case, temporal variation in armed group behavior.
This paper makes at least two contributions to knowledge about the origins of organized armed conflict. First, building on a burgeoning research program on the interrelations and distinctions among rebels and criminals (e.g., Barnes Reference Barnes2017; Kalyvas Reference Kalyvas2015; Lessing Reference Lessing2017; Trejo and Ley Reference Trejo and Ley2020), and drawing on classic works on state formation (e.g., Olson Reference Olson1993; Skaperdas Reference Skaperdas2002; Tilly Reference Tilly, Rueschemeyer, Evans and Skocpol1985), we show the fundamental role of criminal organizations or their legacy in the origins of several rebel groups forming in rural peripheries. Therefore, while there are numerous research questions for which it is justifiable to focus on only the post-divergence phase when the political goals of nonstate armed group violence are clear, for research focused on the origins of organized anti-state armed conflict, the potential importance of groups known as, or events attributed to, “criminals” should be seriously considered—both conceptually and empirically.
Second, this paper shows that low-level violence initially thought to be criminality can be the precursor of large-scale violent rebellion, including civil wars for which the rebels have a clear political or ideological motive. While one could dismiss these small early acts of violence as unimportant, ignoring them risks getting basic descriptive elements wrong about what happens before large-scale conflict in numerous cases. Mischaracterizing the landscape of armed conflict initiation in rural peripheries surely impedes efforts to prevent larger-scale violence in such contexts, especially in the early stages before armed groups become larger, more entrenched, and harder to demobilize. For instance, the failure of international peacebuilding in the Democratic Republic of the Congo was exacerbated by the international community’s assumption that continuing, simmering local violence was merely private and criminal, and thus did not warrant attention (Autesserre Reference Autesserre2010, 72–74). For the study of armed conflict onset, this demonstrates the need for great care in theorizing and measuring the start of organized political violence. For the study of larger-scale civil war dynamics, while it is beyond the scope of this paper to determine whether and how initial “criminal” rebel origins may influence subsequent conflict dynamics, this paper demonstrates the importance of such research.Footnote 3
We proceed as follows. First, we show that a standard assumption about nonviolent mobilization prior to rebellion cannot easily account for basic patterns emerging from new datasets on armed group formation. Second, we present a theoretical argument that explains why rebels are often indistinguishable from criminal groups in their nascent phases. It also helps to explain why rebels later diverge from criminals if they more openly confront the state. Third, we show the empirical relevance of our claims through case studies of the origins of three major rebellions. We conclude by discussing the paper’s implications for contexts beyond states’ rural peripheries and for promising future research directions.
From Public, Nonviolent Mobilization to Rebel Violence? Evidence from New Datasets
Contrary to the main claims in this paper, several influential works assume that organized armed conflict onset within states is preceded by a phase of public, nonviolent contestation and mobilization of the population in the area where rebels emerge. This assumption underpins much of the literature that emphasizes the importance of political grievances and group-level inequalities in propelling civil conflict (e.g., Cederman, Gleditsch, and Buhaug Reference Cederman, Gleditsch and Buhaug2013; Cederman, Weidmann, and Gleditsch Reference Cederman, Weidmann and Gleditsch2011; F. Stewart Reference Stewart2002), as well as more recent work that aims to disentangle a claim-making, nonviolent conflict phase from that of a later militarized phase (e.g., Bartusevičius and Gleditsch Reference Bartusevičius and Gleditsch2019; Cunningham et al. Reference Cunningham, Gleditsch, González, Vidović and White2017; Germann and Sambanis Reference Germann and Sambanis2021). For example, Bartusevičius and Gleditsch (Reference Bartusevičius and Gleditsch2019, 225) open by stating, “Armed conflicts rarely erupt out of nowhere. The armed stage of a conflict is typically preceded by a formative nonarmed stage, where parties articulate incompatibilities and become aware of their opposing positions.” Germann and Sambanis (Reference Germann and Sambanis2021, 178) posit that “[m]ost civil wars are preceded by nonviolent forms of conflict,” and thus focus their analysis on why only some nonviolent self-determination movements become violent.
By this general account, disputes first emerge nonviolently and publicly. Nonviolent organizations plan, prepare, and mobilize the local population where the nonviolent contestation occurs, since a natural constituency for rebels exists there. This sequence implies that the motives of nonstate armed actors emerging from this contestation would be clear, since anti-state aims already would have been publicly articulated. Indeed, many civil wars, including recent ones in Syria and Ethiopia, largely support this approach; both the Syrian and Ethiopian civil wars were preceded by months of public protest and mostly nonviolent mobilization before escalating into large-scale civil war. Recent research suggests that this pathway is especially common in urban settings—especially for rebel groups that form in cities out of opposition parties, social interest organizations, or militaries (Uzonyi and Koren Reference Uzonyi and Koren2023).
Yet the relative ease of documenting urban politics and warfare has long been known to create potential for urban bias in conflict scholarship (Kalyvas Reference Kalyvas2004). We argue below that cases where rebellion instead begins with small-scale, ambiguous violence and the mere acquiescence of civilians are likely concentrated in rural peripheries of low-capacity states—where low state and media presence means this violence is especially unlikely to be clearly or fully documented. In the conclusion, we propose why a sequence of mobilized civilians proactively supporting early rebellion may be more common in highly politicized or stronger-state contexts.
How common is it for rebellions to emerge in peripheries, out of something other than nonviolent movements? Recent research indicates that a majority (66%) of large-scale civil wars since 1944 begin with peripheral challengers and only a small minority (7%) begin with urban protests (Siberdt Reference Siberdt2024, 12). Moreover, three recent actor-focused data collection efforts indicate that nonviolent movements are not a common precursor for anti-state violence. Cunningham, Dahl, and Frugé’s (Reference Cunningham, Dahl and Frugé2020, 488) global data on organizations in self-determination movements (1960–2005) indicate that nonviolent tactics rarely preceded violence. Malone’s (Reference Malone2022) global data on nonstate armed group formation (1970-2012) indicate that only about 27% of nonstate armed groups had initial membership from nonviolent political organizations or parties, religious or labor organizations, or student groups. Finally, using different data that focus exclusively on anti-state armed group formation in Africa (1997–2015), Lewis’s (Reference Lewis2020; Reference Lewis2023) data indicate that among the rebel groups that formed, most (over 80%) did so in rural areas and only a few (about 11%) did so within one year following a nearby protest (Lewis Reference Lewis2023, 8).Footnote 4
Additionally, the Malone and Lewis datasets show that a period of small-scale violence frequently occurs once armed groups form. Specifically, Malone’s (Reference Malone2019, 39) data show that among the subset of armed groups that go on to engage in large-scale civil war, about 42% start with low-intensity violence, culminating in fewer than 25 battle-related deaths at least for the first year. Lewis (Reference Lewis2023) indicates that most African rebel groups start with small-scale violence: less than one-quarter of African rebel groups in the dataset committed an attack that resulted in at least 25 recorded battle-related deaths in the first year after they formed (Lewis Reference Lewis2023, 8).Footnote 5
These basic descriptive findings show that many armed rebellions do not emerge from public mobilization, at least in the form of nonviolent organizations or protest. They also indicate that many, if not most, anti-state armed groups start with small-scale violence, including numerous groups that go on to instigate a major civil war. Approaches that assume that nonviolent movements or organizations precede violence cannot readily account for these patterns.
Of course, early small-scale violence could have a trivial explanation: it could simply be the case that new armed groups have weak military capacity, so the violence they are able to produce is minimal. But if weak military capacity is the key constraint, then why would aspiring challengers of states not simply refrain from all anti-state violence—a highly visible, and thus risky act—until they had covertly gained greater coercive capacity? Our answer is that certain forms of small-scale violence are useful to the goal of building toward full-fledged, anti-state rebellion—and can outweigh its dangers especially if its broader purpose is ambiguous. We develop this argument below.
Ambiguous Violence at the Outset of Armed Group Formation
We present here a theoretical argument about the nature of nascent nonstate armed groups on the periphery and the ambiguous violence they commit until (and if) they diverge into more observably different types of groups. By nascent nonstate armed groups, we mean newly formed armed groups that have yet to reveal their longer-term ambitions. Where we refer to aspiring rebels below, we mean a subset of those nonstate armed groups who secretly possess the ambition of opposing the state in future open rebellion. Our account is agnostic about whether aspiring rebels hold such an ambition from the start of armed group formation, or after formation but before the behavioral divergence we sketch below; both are consistent with what follows.Footnote 6 We argue that only after divergence—which occurs only for groups that survive the initial stage and chose to grow in size and ambition—does it become clear for outside observers which armed groups are primarily political challengers to states (“rebels”) and which seek less confrontational relationships to protect their channels of private gain (“criminals”). We leave for future work cross-group comparisons, including examining why some armed groups go on to instigate major rebellions while others do not.Footnote 7 Here, we focus on temporal variation in behavior within groups.
Initial Conditions
We focus in this paper on peripheral areas of low-capacity states, where states have minimal presence and thus minimal monitoring capacity.Footnote 8 While much of the new wave of social science research on criminal politics focuses on urban crime in Latin America (e.g., Barnes Reference Barnes2017; Berg and Carranza Reference Berg and Carranza2018; Duran-Martinez Reference Duran-Martinez2015; Lessing and Willis Reference Lessing and Willis2019; Magaloni, Franco-Vivanco, and Melo Reference Magaloni, Franco-Vivanco and Melo2020), our arguments build on historical and anthropological research on the prevalence of illicit groups often known as “bandits” in rural peripheries across the world (Hobsbawm Reference Hobsbawm2000). Relevant works tend to focus on the peripheries of specific regions, including Latin America (Ballvé Reference Ballvé2020; Civico Reference Civico2015; Idler Reference Idler2019; Joseph Reference Joseph1990; Knight Reference Knight2022), Europe (Blok Reference Blok1988; Gallant Reference Gallant and Heyman1999; Koliopoulos Reference Koliopoulos1987), East Asia (Billingsley Reference Billingsley1988; Lary Reference Lary2010), South and Southeast Asia (Davis Reference Davis2017; Guha Reference Guha1999; Rafael Reference Rafael1999), the Middle East (Barkey Reference Barkey1994; Kalman Reference Kalman2024; Swedenburg Reference Swedenburg2003), and Sub-Saharan Africa (Crummey Reference Crummey and Crummey1986; Debos Reference Debos2016; Korf and Raeymaekers Reference Korf and Raeymaekers2013; MacEachern Reference MacEachern2018).
Echoing these works and others, we posit that in these contexts, barriers to entry for armed group formation of all kinds are low because even armed groups without any guaranteed external assistance or substantial coercive capacity can form and grow where state monitoring is low (Fearon and Laitin Reference Fearon and Laitin2003; Idler Reference Idler2019, 8; Lewis Reference Lewis2020, 35–39; Weinstein Reference Weinstein2007, 13, 341). In addition, in such contexts difficult terrain (Scott Reference Scott2009, 43), the ability of armed groups to exploit borders to evade authorities (Braun and Kienitz Reference Braun and Kienitz2022, 311; Korf and Raeymaekers Reference Korf and Raeymaekers2013, 7), and tight budget constraints can further undermine state monitoring and response. These conditions allow for a wide range of armed coercion (Bates, Greif, and Singh Reference Bates, Greif and Singh2002)—including criminal organizations (also known as “bandits,” “thugs,” “cartels,” and “warlords”), which primarily seek economic enrichment (Knight Reference Knight2022; Marten Reference Marten2012) or militias, sanctioned or unsanctioned by the state, organized to protect communities from criminal organizations or other predatory armed groups (Ballvé Reference Ballvé2020; Jentzsch, Kalyvas, and Schubiger Reference Jentzsch, Kalyvas and Schubiger2015; Perry Reference Perry1980).Footnote 9 We assume that all nascent, illicit nonstate armed groups seek to survive long enough to build basic coercive capacity because this enables extraction of rents and improves their chances of deterring or repelling future threats (Barnes Reference Barnes2022; Gambetta Reference Gambetta1996; Kalyvas Reference Kalyvas2006; Marten Reference Marten2012). Of course, many nascent armed groups will be eliminated or decide that the risks of further growth outweigh the potential additional gains, but we posit that many armed organizations in the early stages of formation will attempt to grow, at least minimally, if they can.
We also assume that even weak states typically prefer not to allow new illicit, nonstate armed groups to form on their territory and can end them if they can identify and locate them in this early, weak phase.Footnote 10 This is especially true of armed groups that the state suspects will go on to engage in competitive state building in opposition to the state (“rebels”), since those groups pose a greater, sometimes existential, threat to the state than criminal groups, militias, or warlords that operate outside but not in opposition to the state (Civico Reference Civico2015; Lessing Reference Lessing2017; Marten Reference Marten2012; Staniland Reference Staniland2021). For states, therefore, the challenge is to gain information that can identify and locate nascent armed groups and sort them based on the severity of the threat they pose.
The Benefits of Anonymous, Small-Scale Violence
Under these conditions, all nascent, illicit armed groups initially face a similar strategic challenge: ensuring they evade the attention of the state while beginning to build their coercive capacity and understand their operational environment. Committing violent attacks—especially against state targets—can be dangerous because it can attract the scrutiny of the state during this especially vulnerable phase. Violence can also attract the attention of nearby civilians who, if they obtain useful information, can denounce the new armed group to the state. However, anonymous, small-scale violence can be useful—allowing armed groups to gain the benefits we explain here while mitigating the risks—since it often does not appear sufficiently threatening to induce a poorly funded state to invest the resources needed to identify and apprehend those who committed it. For reasons of scope, and because it is most provocative to states, we focus here on these groups’ repertoire of anti-state violence, such as assassinations of local politicians or raids on a small detachment of police officers.
Why not simply refrain from violence altogether? We posit that violence can be beneficial to nonstate armed groups—both those contemplating rebellion and those planning to pursue solely criminality—for two reasons. First, nascent armed groups with myriad motives use violence to obtain material resources that nascent groups often lack but are needed to ultimately build basic coercive capacity, and making such violence anonymous and small in scale mitigates the risk that such acts entail. For example, extortion or attacking a small police outpost at night probably will not result in substantial physical harm or a major battle, and minimizes the risk of discovery and capture, but offers a chance of gaining financing (in the case of extortion) and arms or supplies (in the case of the police outpost attack). Small, unclaimed acts of violence are also less easy for the state to attribute to specific individuals than larger claimed ones, and thus offer an easier exit option—leaving the area or hiding among the civilian population while ceasing violence—if the response from state actors is stronger or more effective than expected. Probing with larger, claimed anti-state acts of violence risks attracting a more intensive state response because states calibrate their response to the perceived threat.
Second, small-scale, anonymous violence allows nascent armed groups to test their operational environment, probing the responses of the state and local population. For example, by killing a local official in a remote area (and not publicly claiming responsibility or political intent), an armed group not only hinders the central state’s ability to collect information about the group but also probes state responsiveness to violence against its local apparatus. It also helps the armed group to learn where other local officials look away or dig deeper, and whether such officials have the intelligence capacity to learn about their identity and location. Gaining this knowledge about their operational environment enables armed groups to limit proximate threats and better select areas most conducive to expanding the organization—without sending a clear signal of the future broader threat the group poses to the state.
Civilians are an important part of this operational environment. They often learn about local, incipient armed groups, and their denunciations to the state can hasten the end of those groups. We assume that most civilians near armed groups’ early operations do not already hold strong preferences in favor of or against a local armed group, and in fact lean toward remaining silent about a new group if the violence does not affect them directly or they perceive that they may ultimately benefit from it.Footnote 11 These assumptions are well supported by seminal studies of civil war that draw on qualitative evidence from historical accounts of numerous cases (see especially Kalyvas Reference Kalyvas2006, chap. 5) and analyses based on in-depth fieldwork in rural, low state-capacity contexts (e.g., Debos Reference Debos2016; Idler Reference Idler2019; Lewis Reference Lewis2020). They are also consistent with the findings of relatively infrequent nonviolent mobilization prior to rebel formation that are noted in the previous section. Further, in such peripheral contexts, there is often a recent legacy of nonstate armed actors providing protection for local populations, which can render small-scale, anonymous violence comprehensible and even legitimate to local communities (Knight Reference Knight2022, 33–38; Perry Reference Perry1980, 73).Footnote 12
Therefore, to induce civilians to stay silent, what nonstate armed groups in peripheries need from civilians in this early phase is not active support and deep belief in their cause, but rather a more minimal acquiescence or ambivalence. This can be relatively easy to achieve if armed groups commit small-scale violence and remain anonymous—especially if violence occurs beyond the armed group’s proximate community, so that danger is not perceived as immediate. Small-scale violence is also less likely to induce operational errors that trigger local narratives questioning the group’s competence (Lewis Reference Lewis2020). Larger-scale violence is riskier, more public, and potentially directly harmful to civilians or their kin.Footnote 13 As a result, it can lead sizable numbers of civilians to become wary and consider denouncing the nascent armed group, and is therefore less appealing to the armed group at the outset.
Ambiguity as Opportunity for Aspiring Rebels
This logic yields two key implications for nascent groups that aspire to rebellion. First, the use of a similar repertoire of violence by all nonstate, nascent armed groups on the rural periphery obscures its purpose for outside observers, making it ambiguous whether the group committing the violence is a likely future threat to the state.Footnote 14 This ambiguity therefore benefits the nascent groups that opt to rebel. By using anonymous, small-scale violence, aspiring rebels appear to the state as one of the usual suspects, “mere bandits,” lowering the state’s willingness to spend limited resources to end the group quickly. Therefore, states often fail to respond effectively to nascent rebels not because of a complete lack of violence but instead because of a misperception about small-scale violence that obscures group type.
Second, for aspiring rebels seeking to build the skills of using ambiguous violence, the existence or remnants of other small armed groups on the rural periphery offers an opportunity.Footnote 15 Armed groups with varied motives on the periphery can learn from one another contemporaneously or via another group’s legacy—for example, in the crafts of committing small-scale, anonymous violence or accessing illicit markets.Footnote 16 This tactical learning among armed groups further obstructs observers (like state actors and journalists) from understanding and predicting the future ambitions of emerging armed groups.
Even beyond the potential for learning, where numerous nonstate armed groups are present in a region, the shared need to organize and use a similar form of violence can create a sizable pool of violence entrepreneurs with skills and experience in doing so (Barkey Reference Barkey1994; Blok Reference Blok1988; Debos Reference Debos2016; Finkel Reference Finkel2017; Gallant Reference Gallant and Heyman1999; Idler Reference Idler2019; Lary Reference Lary2010; Perry Reference Perry1980). One group can therefore benefit directly from another via recruitment, regardless of the group’s long-term ambitions. In sum, the existence or legacy of criminals on the rural periphery provides aspiring rebels with both the cover and skill set to survive the nascent stage of formation.
Divergence
We argue that the phase of ambiguity ends and nonstate armed groups’ violent behavior observably diverges if they survive past this nascent stage, with those opting to try to seize full political and territorial control of some (for secessionist rebels) or all (for center-seeking rebels) of a state’s territory more likely than those who do not to engage in sustained, larger-scale offensive violence against the state.Footnote 17 Armed groups that become rebels, adopting a primary goal of conquering the state (or a sizable portion of its territory), must inflict costs on state forces and penetrate more territory, moving closer to the capital. This is the case even if they simultaneously continue criminal activities (see, e.g., Asal, Rethemeyer, and Schoon Reference Asal, Rethemeyer and Schoon2019) or engage in turf battles with local bandit groups as they signal their aims of dominance in an expanding area. To coerce the state into bargaining, rebels must at some point signal their capability.
By seeking to supplant the state (and sometimes other nonstate armed challengers), rebels must invest in formalizing, training, and transforming former “bandits” into soldiers willing and capable of offensive strikes and sustained combat initiated against state forces, instead of the low-risk raiding typical of ambiguous violence or the defensive anti-state violence that more commonly characterizes larger criminal groups (Lessing Reference Lessing2017). For instance, although many types of armed groups use terrorism to pressure a state, “all rebel groups attack military targets,” as Page Fortna (Reference Fortna2015, 524) notes. Few armed groups other than rebels pursue goals that require sustained offensive attacks on state forces, such as regular ambushes to wear down state forces or assaults on state security outposts. These attacks do not necessarily take the form of conventional warfare, but at a minimum rebel groups seek to increase the size and frequency of their attacks on state forces and attempt to push deeper into territory controlled by the state.
The initiation of sustained, larger-scale offensive violence signals an effort and a set of practices that are qualitatively different from, and riskier than, banditry. But larger-scale violence can enhance a group’s bargaining power with the state in three important ways. First, larger-scale offensive violence against state forces compels the central state to update its beliefs about the group’s primary goal and its capabilities, and thus the severity of the threat it poses. Second, it can help rebels to demonstrate their capabilities to international audiences, which (as states know) can convince external foes of the state to invest in them (Huang Reference Huang2016; Salehyan, Gleditsch, and Cunningham Reference Salehyan, Gleditsch and Cunningham2011). Third, it can help to demonstrate the rebels’ capabilities to local civilians. More active civilian support often follows the side perceived to be stronger or more likely to prevail (Kalyvas Reference Kalyvas2006, 126–28). Although aspiring rebels may initially only need the ambivalence of civilians, larger-scale violence against the state can assist in mobilizing civilian support that enhances their available resources or bargaining power.
Armed groups that persist as criminal groups or warlords do sometimes also grow in size and ambition, becoming “stationary bandits” that share some behaviors in common with rebels, such as state building by eliminating armed rivals and providing governance (Blattman et al. Reference Blattman, Lessing, Tobon and Duncan2021; Olson Reference Olson1993; Marten Reference Marten2012; Tilly Reference Tilly, Rueschemeyer, Evans and Skocpol1985). However, rebels’ violent campaigns to seize total political control over large swaths of a state’s territory are different than the more limited local territorial control that is more common for criminals (Barnes Reference Barnes2022; Gambetta Reference Gambetta1996), warlords (Malejacq Reference Malejacq2019; Marten Reference Marten2012), and paramilitaries (Civico Reference Civico2015), all of whom seek to establish and maintain a protection racket (Gambetta Reference Gambetta1996). These groups also sometimes provide governance in areas they control, and even gain legitimacy among civilians (Ballvé Reference Ballvé2020; Blattman et al. Reference Blattman, Lessing, Tobon and Duncan2021; Lessing Reference Lessing2021; Marten Reference Marten2012).Footnote 18 But because of the high costs of fully supplanting the state, this form of territorial control is often not complete, and relies on the collusion (rather than the elimination or displacement) of local state security actors.
Crucially, even militarily strong, ambitious criminal organizations—such as those of contemporary Latin America or Republican China—lack the incentives that nascent rebels face to bear the costs of initiating sustained large-scale offensive violence against the state. The production and disruption costs of large-scale violence are high and unnecessary for consolidating a protection racket (Lessing Reference Lessing2017; Reference Lessing2021; Skaperdas Reference Skaperdas2002). Few such groups are therefore willing to reinvest rents into the high costs of establishing the capabilities to train, replace, and supply military forces that can sustain large-scale offensive violence against the state, instead of keeping it for private gain. Such groups only need enough military capacity to compete over turf with rival criminal groups or to defensively constrain state forces until a new arrangement is reached (Barkey Reference Barkey1994; Lary Reference Lary2010; Lessing Reference Lessing2017). Further military investment would merely exhaust the rents a group seeks to defend.
Instead, as prior research shows, when nonrebel groups like cartels confront the state, they typically do so in more limited (and less violent) ways with the aim of influencing state repressive policies and maintaining their access to and control over illicit markets (Barnes Reference Barnes2017, 974; Lessing Reference Lessing2017, chap. 3). For instance, many studies on recent anti-state violence in Latin America emphasize violent acts such as the assassination of specific state officials (Barnes Reference Barnes2017; Berg and Carranza Reference Berg and Carranza2018; Trejo and Ley Reference Trejo and Ley2019), defensive measures against state incursions (Magaloni, Franco-Vivanco, and Melo Reference Magaloni, Franco-Vivanco and Melo2020), or terrorism to coerce the state to abandon repressive policies (Lessing Reference Lessing2017), but rarely the regular military attacks typical of rebels.
Indeed, while large-scale organized criminal violence has emerged in several Latin American countries, the vast majority of it is intercartel violence (Lessing Reference Lessing2017, 46). In the rare cases when substantial state–cartel violence occurs, it is often initiated by the state. For example, Civico (Reference Civico2015, 208) argues that in Colombia, “a direct confrontation between the state and organized crime, such as the paramilitaries and the related drug-trafficking cartels, has been the exception—as when the state declared a war against Pablo Escobar and the Medellín cartel—but the norm has been either alliance or negotiation between the state and the illegal groups.” In sum, although a diverse array of armed groups may engage in what Charles Tilly (Reference Tilly, Rueschemeyer, Evans and Skocpol1985, 181) characterizes in his classic work as “state making,” armed groups engaging in rebellion additionally engage in what he calls “war making” to combat the state “outside the territories in which they have clear and continuous priority as wielders of force”—a costly signal of their intentions toward the state.
Finally, it bears noting that if this divergence occurs, it is not always permanent. Some substantial criminal enterprises later attempt rebellion, as was the case for Pancho Villa’s and Emiliano Zapata’s groups during the Mexican Revolution (1910–20). This phenomenon appears to be rare, however, since even larger armed groups that established themselves as economically driven criminal organizations face difficulties in reordering their organization for the project of sustained combat against state forces (Perry Reference Perry1980, 140–48). Many previous members of mature criminal organizations may desert or resist transitioning from bandits to soldiers (Averill Reference Averill2006, 173; Billingsley Reference Billingsley1988, 259), as was the case for Villa’s and Zapata’s organizations (Brunk Reference Brunk1996; Vanderwood [Reference Vanderwood1981] 1992). Likewise, rebellions may devolve back into more ambiguous, “criminal” violence—but typically after group fragmentation, disbanding due to defeat, or demobilization during a peace process that releases a large pool of violence entrepreneurs (e.g., Daly Reference Daly2016; MacEachern Reference MacEachern2018).
Illustrative Cases
In this section we demonstrate the empirical relevance of our arguments by analyzing three cases of nonstate armed group origins in the lead-up to major anti-state armed conflict: the CCP in China during the Republic period (1912–49), the Việt Minh’s formation prior to the French Indochina War (1946–54), and the M26 during the Cuban Revolution (1953–59). We selected these three cases of insurgency because they are hard cases for our argument. They are known today for their highly ideological nature and success at building large social movements, whereas it would be unsurprising if small rebel groups known for their “greed” or at least minimal political support started in a way that appeared criminal. Our theoretical arguments were inspired by our knowledge of cases beyond this paper’s three case studies and 12 Supplementary Material cases, especially Uganda and other cases from Africa (see Lewis Reference Lewis2020, 107–24), as well as historical works on banditry (e.g., Hobsbawm Reference Hobsbawm2000). We developed our core arguments prior to deeply researching armed group origins for the cases examined here and in the Supplementary Material.
All three cases of revolutionary insurgencies examined here are known for their ideological origins, for inspiring legions of new armed groups around the world (M. Stewart Reference Stewart2021), and for exemplifying long-enduring regimes born of social revolution (Lachapelle et al. Reference Lachapelle, Levitsky, Way and Casey2020; Levitsky and Way Reference Levitsky and Way2013). By scrutinizing these insurgencies’ earliest phases as nonstate armed groups, these case studies demonstrate that these groups were in fact born out of ambiguous violence that—at least at the time—was deemed criminality.
Twelve additional examples that illustrate our theoretical framework can be found in the Supplementary Material; we list those cases in Table 1. In addition to the historiographic considerations noted below, we selected these cases with the aim of demonstrating that our arguments extend across low state-monitoring capacity contexts in a wide range of times and places.
Additional Cases in the Supplementary Material
Notes: Conflict years are provided as reference for identifying the conflicts. They are based on conventional dating of the conflicts from secondary sources, not the years of armed group formation.
We take a qualitative and primarily historical approach because we aim to illustrate the mechanisms leading to pre-rebellion periods of low-level rebel violence that appears (to outsiders) criminal in nature. Also, we aim to show support for our assertions that these processes are often intentionally hidden by (or may even remain murky among) key actors, and misunderstood by outside observers, in the years during and immediately following their occurrence. Our three main cases occurred sufficiently long ago that they allow us to show how historiography has evolved from uncritically accepting their roots as ideological to showing the more complex reality, and feature well-developed historical debates that use a wide range of sources (Balcells and Sullivan Reference Balcells and Sullivan2018; Kalyvas Reference Kalyvas2001).
Problems of incomplete or politicized sources can loom especially large for the study of a conflict’s origins because this phase often occurs in secrecy, and because events relevant to “who started it” often become contentious, politicized issues—especially in the years during and immediately following the events. For instance, official histories of the conflicts emerging in the immediate aftermath of war tend to emphasize revolutionary leaders’ and others’ romanticized views of early base areas at the expense of rebels’ cooperation with criminals or the minimal political motivations of many early recruits (Billingsley Reference Billingsley1988). Indeed, the political and criminal groups named in our cases are typically remembered respectively for their political or their criminal acts. Our cases draw from the best available historiography on the dynamics of rural crime and the early behavior of the rebels traced through each case. Although these prominent cases of rebellion have vast secondary-source literatures, only a small portion extensively detail the early behavior of armed groups that became rebels.
In what follows, each case identifies the repertoire of anonymous, small-scale violent acts used by nonstate armed groups early on—including those that were later clearly identifiable as criminal or rebel groups. It also shows that its purpose was ambiguous to outside actors, who struggled to distinguish nascent rebellion from criminality. In addition, these cases emphasize the importance of armed groups learning from one another to hone the skill set of organizing ambiguous violence, and they show the divergence in the behavior of these groups during later phases of the conflict.
Revolutionaries or Bandits? Armed Groups in Republican China (1912–49)
Mao’s CCP is known as a clear case of an ideologically motivated insurgent group with sophisticated organizational control over its units and the capacity for conventional warfare. However, this is not how the group started; in fact, this image sharply contrasts with the widespread small, localized bandit groups plaguing China from which the rural CCP insurgency emerged. During the early periods of the CCP’s formation, the small armed groups that eventually merged into the Red Army used ambiguous violence, making them difficult to distinguish from the welter of other small armed groups in rural areas of Hunan and Jiangxi. In China, until relatively recently, scholars deemphasized the role of bandits in the early CCP due to political sensitivities surrounding the group’s origins (Youwei and Billingsley Reference Youwei and Billingsley2018, 246).
In the 1920s, Republican China struggled to extend state authority to remote rural areas of the country, especially in the periphery of north China where banditry was a pervasive problem (Perry Reference Perry1980, 38, 63; Shan Reference Shan2006, 25). Provinces faced many small bandit groups, usually ranging from a few dozen to a few hundred fighters. Bandits attacked local militias to capture arms and deter them from operating against bandit activity (Averill Reference Averill2006, 86; Billingsley Reference Billingsley1988, 57; Shan Reference Shan2006, 32). To avoid large state forces and to sustain themselves, bandit gangs avoided excessive violence, which could turn victims toward the state and undermine their security and supplies (Billingsley Reference Billingsley1988, 178). As Tiedemann (Reference Tiedemann1982, 410–11) notes, “[t]raditional bandits took great care in the selection of targets, because indiscriminate attacks on the rural population would have antagonized the masses on whose support, or at least indifference, the bandits relied.” Communities rarely supported bandits out of sympathy or admiration. Instead, “their feelings were much more ambivalent, reflecting simultaneously the need for protection in a violent world and the knowledge that at any time the protectors could become aggressors” (Billingsley Reference Billingsley1988, 186).
In October 1927, when Mao and the initial CCP arrived in the border region of Hunan and Jiangxi provinces, and continuing through roughly mid-1928, the early repertoire of violence of the rural CCP cadres was difficult to distinguish from the behavior of the local bandits. Like the bandits, the groups that coalesced into guerrillas sporadically attacked vulnerable state forces or local militias, particularly to capture arms, while avoiding larger forces. On many occasions, the nascent CCP and the bandits used similar methods to extort supplies and money, killing or kidnapping “enemies of the people” such as local government officials, police chiefs, and landlords (Billingsley Reference Billingsley1988, 257; Opper Reference Opper2019, 76–77). Similar to banditry, small anti-state attacks not only aimed to obstruct the local defenses of the state, but also demonstrate their competency to local communities (Averill Reference Averill2006, 168, 170). Both bandits and the early proto-CCP cadres relied on nearby villagers for information and to avoid detection by larger government forces. They even located their camps in traditional bandit sites for similar safety reasons. As the historian Phil Billingsley (Reference Billingsley1988, 255) notes, “[t]o those observing this behavior it was virtually impossible to distinguish the ‘guerrillas’ from any other bandit gang.” Local government officials dismissed early Communist cadres as “bandit chiefs” (251).
There are several reasons for the use of similar patterns of small-scale violence by the myriad nonstate armed groups of rural China. First, they faced the same dilemma of any early armed group that starts without substantial endowments: the need to gains funds, arms, and recruits while maintaining secrecy from large state forces that could easily destroy them if detected. All nonstate armed groups sought to evade state forces and avoid excessive violence against the local population upon which they relied. Bandits tried to manage the scope of their raids carefully to test their operational environment and avoid triggering a major state response with forces from outside the province. Similarly, the CCP guerrillas that later formed in the same area used the bandits’ experience to try to launch less provocative attacks than their capabilities allowed to gain weapons and undermine the state’s local infrastructure before their actions caused the state to launch a major “bandit suppression” campaign (Averill Reference Averill2006, 136–37, 171).
Second, many of the early guerrilla bands in China were “dominated by former bandits who continued to raid and kidnap even while flying the red flag of revolution” (Billingsley Reference Billingsley1988, 35). The Communists actively sought to recruit and incorporate bandits to bolster their own strength. In the late 1920s, the bulk of Mao’s guerrilla force consisted of “soldiers, bandits, robbers, beggars, and prostitutes” (Hobsbawm Reference Hobsbawm2000). Finally, the early Communist groups, including those not consisting of former bandits, often learned their tactical repertoire from bandits. Several CCP commanders learned directly from bandit groups or copied the repertoires of the local bandit groups operating in their new base areas. Mao ([Reference Zedong1937] 1961) even incorporated “bandit technique” into his writing on guerrilla tactics that would subsequently be distributed across the world (Averill Reference Averill2006, 153; Billingsley Reference Billingsley1988, 252).
Even though it was difficult for outsiders to distinguish whether violence was from bandits or guerillas early on, as the CCP grew over the course of 1928, its violence diverged from local bandit groups. Though some bandit groups grew to several thousand and established territorial control over some remote areas, sometimes even providing a “shadow government,” they never grew to a size that could pose a significant threat to the state itself (Shan Reference Shan2006). Large bandit organizations rarely served as a source of large-scale offensive violence against the state. One of the few exceptions is the Nian Rebellion (1851–68), which created significant disputes inside the participating bandit coalition over the risks involved (Perry Reference Perry1980, 140–48). Though some bandits’ horizons grew with their size, and such groups went on to serve in rebellions against the state, more commonly, “local banditry was largely tolerated by the authorities so long as it posed no overt threat to regional security” (Billingsley Reference Billingsley1988, 254). Therefore, bandits that managed to establish some territorial control over a remote area had little incentive to continue to expand lest they draw themselves into a direct confrontation with the state. Likewise, the bandit tactic of hiding in remote areas worked well for ensuring their personal survival and enrichment, but was a poor strategy if the group wanted to go on to defeat the state (232). Any armed groups seeking to challenge the state directly had to go beyond these standard “bandit tactics.”
Instead of sticking to such tactics, the CCP continued to grow and invest in military capability to challenge the Nationalist government. They particularly focused on developing their groups into a centralized organization capable of sustained offensive attacks against the state—something that few bandit groups were either capable of or willing to undertake. Bandits who were incorporated into guerrilla groups were found to have “virtually no formal military training” and were put through “daily drills” by army cadres to formalize them into useful military units (Averill Reference Averill2006, 173). In areas under their control, the centralized CCP eventually turned on their former bandit allies, sparing “no effort to eliminate bandit gangs” (Perry Reference Perry1980, 228; Shan Reference Shan2006, 44).
Nationalists or Crime Syndicates? Armed Groups in French Indochina (1946–54)
Retrospective accounts typically view the Việt Minh as among the best-organized and most ideologically committed insurgencies in history (e.g., Goscha Reference Goscha2022; M. Stewart Reference Stewart2021). This characterization of the Việt Minh is especially pronounced in early histories of the war, which focused on large-scale violence during the French War in the 1950s in the northern regions of Indochina, known as Tonkin (e.g., Fall [Reference Fall1961] 2005; O’Ballance Reference O’Ballance1964).Footnote 19 However, violent resistance to the French occurred initially in the 1930s in the southern third of Indochina, known as Cochinchina, which includes the Mekong Delta and the provinces surrounding Saigon (now Ho Chi Minh City). The northern Việt Minh insurgency, launched in 1941, operated entirely independently from the southern Việt Minh insurgency until 1945,Footnote 20 and even after that, the two groups retained considerable organizational independence from one another (Elliott Reference Elliott2003, 36, 46, 107–9).
We focus here on the various small armed groups in the South that merged into the often-overlooked southern Việt Minh (Chapman Reference Chapman2013; Elliott Reference Elliott2003). Among the armed groups in Cochinchina was a criminal syndicate, the Bình Xuyên, formed from bands of river pirates and other smaller criminal groups. The early behavior of the northern Việt Minh is omitted due to space constraints but also aligns with our argument; we present detailed evidence to support this claim in part II of the Supplementary Material.
The early Việt Minh units in Cochinchina during the 1930s were small and generally committed ambiguous violence against the French as they built their strength. In late 1940, believing French forces were sufficiently distracted in the Franco-Thai War (1940–41), the southern Việt Minh mobilized peasant discontent into a localized uprising against French authority, known as the Nam Kỳ uprising. However, even in this weak-state context, uprisings simply revealed the small Việt Minh to the repressive French, and the French crushed them in just a few weeks. After 1940, those groups that became known as the Việt Minh focused exclusively on rebuilding their strength as a small clandestine armed group due to their fear of attracting French attention to their poor armaments, which consisted of a small collection of firearms, bamboo spears, and machetes. By 1945, across Cochinchina, the Việt Minh had only about two hundred members (Elliott Reference Elliott2003, 67). They committed only sporadic, small-scale violent acts against the French.
As they rebuilt the strength of their clandestine armed bands, many leaders of the southern Việt Minh initially hid under the protection of and interacted with numerous local criminal organizations. Before 1945, one of the main criminal groups, the Bình Xuyên, was a loose coalition of river pirates that numbered between two hundred to three hundred fighters. Similar to the clandestine Việt Minh groups that formed after the Nam Kỳ uprising, these groups would flee into remote rural areas of Cochinchina to evade colonial authorities (Chapman Reference Chapman2013). Behaviorally, the criminal groups and southern Việt Minh were difficult to distinguish during the 1930s and 1940s for two reasons. First, all these groups focused on remaining clandestine and avoiding large-scale violence that would allow the French to identify and locate them, but still did not forgo the use of violence entirely. The violence they committed consisted of isolated raids against vulnerable French positions or civilians to collect supplies and arms. Second, collaboration after the Nam Kỳ uprising between the Việt Minh and local criminal groups further blurred the difference between these two kinds of armed groups, as the Việt Minh relied on the criminal groups for help evading French forces and to partially refill their depleted ranks.
In the period immediately following France’s return to Indochina in 1945, the two groups’ use of similar small-scale violence continued as the Bình Xuyên joined the Việt Minh in a loose coalition of groups that sought to resist France’s return (McHale Reference McHale2021, 107–8). Both the Bình Xuyên and the Việt Minh were poorly organized, poorly armed, and mainly focused on evading French forces. Attacks took the form of small and sporadic ambushes, sabotage, and raids to collect arms (McHale Reference McHale2021, 131; Waddell Reference Waddell2018, 66–68). As one administrator complained, shortly after France’s return, there were widespread small bands committing “acts of banditry for which they do not even try most of the time, to ascribe a political character. … They kidnap or kill notables or, above all, demand ransom from rich landowners or well-off shopkeepers” (quoted in McHale Reference McHale2021, 144). Both groups were also collecting arms to build up their organizations to secure territorial control as Cochinchina became increasingly fragmented between the various armed groups and French forces. As the two groups consolidated and grew, they diverged into more clearly identifiable political versus criminal organizations.
This divergence between these two groups is well documented. It can be traced back to major disagreements between the groups’ leaders, Nguyễn Bình of the Việt Minh and Bảy Viễn of the Bình Xuyên, which demonstrate how emergent differences in aims led to different kinds of violent behavior for each armed group. Nguyễn Bình wanted to invest in the military capacity of the Việt Minh so the armed resistance in Cochinchina could become strong enough to launch major offensives against the French. He also wanted to centralize and formalize the loose coalition of armed groups in the south whose leaders, he believed, suffered from a “warlord mentality” that made them more focused on extracting resources from their turf than actively fighting the French (Li Reference Li2016, 145; Waddell Reference Waddell2018, 68). In contrast, Bảy Viễn and the Bình Xuyên resisted the Việt Minh’s attempts to exert control over their group and force them to invest in forming stronger military units for fighting the French. Instead of seeking to invest in military capabilities, the Bình Xuyên wanted an independent protection racket over illegal activities in Saigon; they did not join in the Việt Minh’s strategy of urban terror, which attracted significant French police attention (Li Reference Li2016, 154, 158). While the French may have had difficulty distinguishing between them early on, during this later period they leaped at the opportunity to flip the Bình Xuyên against the Việt Minh. The Bình Xuyên arrived at a deal with the French in June 1948 (Waddell Reference Waddell2018, 73–76, 112–13).
After the Bình Xuyên’s deal with the French, the two groups’ behaviors and uses of violence diverged sharply. The Việt Minh invested in increased military capability to prepare for sustained large-scale violence against French forces through a series of conventional offensives. In contrast, the French permitted the Bình Xuyên to consolidate territorial control over criminal activity in and around Saigon. The Bình Xuyên used selective violence against both French and Việt Minh agents when they encroached on the Bình Xuyên’s turf or got in the way of their “business” (Li Reference Li2016, 170–71). As historian Jessica Chapman (Reference Chapman2013, 22) notes, “Bảy Viễn and the Bình Xuyên were motivated largely by the less-than-lofty ambition of protecting and enhancing their own wealth and power.”
The Legacy of Rural Banditry: Armed Groups in Cuba (1953–59)
The inclusion of Fidel Castro’s insurgency against the Fulgencio Batista regime in this paper may initially puzzle those familiar with the case because the early behavior of Castro’s insurgent group, the M26, would seem to contradict our expectations. Unlike the early CCP in China or the Việt Minh in Cochinchina, which sought to avoid drawing the attention of state forces, Castro returned to Cuba in December 1956 with around 80 fighters and arms aboard the Granma and a few months later broadcast his presence in the Sierra Maestra mountains through interviews with journalists. Even with only limited capacity, Castro boldly declared himself to the world rather than avoid detection by the state and removed any ambiguity about his group’s identity as a political opponent of the state.
However, as we show below, after the disastrous Granma landing, the M26 shifted to behavior that aligns with our expectations of early rebel behavior. While Castro publicly downplayed the role of bandits in his insurgency,Footnote 21 the M26 relied on the aid of local bandits in the Sierra Maestra and borrowed from traditional Cuban bandit repertories while committing anonymous, small-scale violence for nearly a year until early 1958. During this time, the Batista regime struggled to gather reliable information on the group until M26 forces improved and began launching larger attacks against the state.
The difficulty of distinguishing rebels from bandits has a long history in Cuba. During Cuba’s nineteenth-century independence struggles against Spain, many ambiguous armed groups formed throughout the island. As Pérez (Reference Pérez1989, xv) notes, “banditry was approaching unmanageable proportions, it was widespread and spreading. Bandits roamed the countryside, in untold numbers, in unchecked movement.” Political insurgencies challenging Spanish rule incorporated bandits such as Manuel García’s prominent group (Schwartz Reference Schwartz1989, 10, 159). After the 1895–98 war of independence, banditry became increasingly prevalent in the Oriente region of eastern Cuba, where Fidel Castro later began his insurgency. By the 1910s, Oriente faced the highest incidents of armed crime in all of Cuba (Pérez Reference Pérez1989, 38, 146, 177).
Therefore, in December 1956, when Castro and his forces retreated to the Sierra Maestra after the failed Granma landing, “they stepped into a tradition of rebellion, however vague and ill defined. They came upon armed struggle, they did not introduce it” (Pérez Reference Pérez1989, 192). With less than 20 men after the botched landing, Castro’s forces desperately evaded state forces. The M26 relied on the expertise of one local bandit leader, Crescencio Pérez, who helped to introduce the insurgents to local peasant communities (Bonachea and San Martin Reference Bonachea and Martin1974, 89; Pérez Reference Pérez1989, 193; Thomas Reference Thomas1977, 115). After regrouping their forces, Castro’s insurgents adopted the repertoire of violence used by the earlier bandit groups. In early 1957, Castro’s forces, numbering only 18, focused on raiding elites’ property located nearby, extorting financial contributions, and burning sugar fields (Guerra Reference Guerra2018, 261; Thomas Reference Thomas1977, 149). The only early attack against state forces during the period, the raid on the remote La Plata outpost, was conducted to gain arms and ammunition and resulted in minimal casualties. Although the attack improved group morale and provided much-needed weapons, M26 did not vocally publicize the victory to avoid drawing greater state attention. Days later, M26’s silence on the attack led the Havana Post to attribute it vaguely to “rebels” and grossly misreport the actual casualties. The army only officially acknowledged the attack over a month later (Thomas Reference Thomas1977, 127–28).
Rather than use the newly captured weapons to launch even more ambitious attacks against state forces, Castro instead pursued a strategy of quietly building his forces within the remote and traditionally rebellious east of Cuba, having learned the dangers of premature large-scale attacks from the failed Moncada Barracks attack of 1953. Castro’s decision to limit M26’s violence created confusion among the Batista regime about the nature of the armed opposition and how to parse its ambiguous violence. Initially, the regime claimed that Castro had been killed in the initial landing of the Granma. M26’s use of smaller-scale violence and the repertoires of traditional bandits led one army commander to declare to journalists that “only a band of outlaws (not Castro)” remained in the Sierra Maestra (Thomas Reference Thomas1977, 152). These claims were not simply political maneuvers by the state. Special troops committed to hunting M26 after the landing were withdrawn from the Sierra Maestra, leaving only limited state forces. In May 1957, even after New York Times journalist Herbert Matthews’s interview with Castro, M26’s continued use of only small-scale violence led one historian to note that “the irritation and even bewilderment of Batista and his officers at the worldwide interest in Castro can be readily understood” (Thomas Reference Thomas1977, 155).
Over the course of 1957, while Batista struggled against urban guerrillas, “Castro had survived with a small force of guerrilla fighters in the Sierra Maestra for over a year. Isolated from the rest of the island, sustaining few causalities … Fidel was creating a well-disciplined group of guerrillas and gaining the backing of the area’s population” (Bonachea and San Martin Reference Bonachea and Martin1974, 173). In remote liberated territory, M26 began to prepare for larger-scale offensive attacks of state forces; for example, they developed rudimentary logistics, armories, and factories to keep guerrillas supplied in prolonged fights. As M26 entered this new phase of divergence from surrounding armed groups by confronting the state in bolder way, M26 forces eliminated bandits they encountered or incorporated them into the insurgency (Thomas Reference Thomas1977, 167, 212). The insurgency needed to distinguish itself from exploitive outlaws preying on the population. As Castro declared, “if we don’t keep order in our liberated zone, the people suffer. Our revolution is tarnished” (quoted in Guerra Reference Guerra2018, 253–54).
The regime only committed significant resources to combat M26 in mid-1958, over a year after the failed December 1956 Granma landing. In May 1958, the enlarged and trained M26 insurgency proved capable of fighting off the state’s first major offensive in the Sierra Maestra. Although Batista’s greater focus on M26 was partly due to the elimination of the previously more pressing threat posed by urban guerrillas, the rural insurgency had finally developed into a clear and undeniable political threat to the regime by 1958.
Conclusion: Bringing Banditry Back into the Study of Conflict Origins
Our arguments and case evidence indicate that the indistinguishability of early rebellion from banditry, and especially the ambiguous nature and the low quantities of the violence both produce, has obscured the potential prevalence of a pathway of early insurgency that can ultimately lead to large-scale civil war. We have argued why “bandits” and “rebels” both face incentives for committing small-scale, anonymous violence in their early phases. We presented a logic of why their respective patterns of violence usually diverge only if, and after, they expand in ambition and military capacity. What are the implications for future research on armed group formation, criminal violence, and civil war?
This paper focused primarily on historical examples of rural, peripheral regions of weak states, where state capacity to detect new threats is low. We note that those conditions apply to many contemporary low-income states, especially in Sub-Saharan Africa and South Asia.Footnote 22 Our arguments and case studies suggest that it is not merely lack of state presence in these areas that make them vulnerable to civil conflict onset; the presence or legacy of other nonstate armed groups there can lead to the existence of a pool of violence entrepreneurs possessing the skill set needed by aspiring rebels. Further, prolonged state absence may contribute to a legacy of “shadow citizenship” when communities lived under armed groups rather than state rule (Idler Reference Idler2019), which may predispose communities on the periphery to find anti-state violence and economic acts, perceived by the center as illicit, as legitimate (MacEachern Reference MacEachern2018, 144). These further opportunities for aspiring rebels suggest more research on the origins of political violence should focus on the question raised by scholars studying rural peripheries: “[W]hat happens where the state ends?” (Korf and Raeymaekers Reference Korf and Raeymaekers2013, 5; emphasis in original; see also Nathan Reference Nathan2023). An observable implication of our arguments that awaits testing is that aspiring rebel groups that form proximate to criminal organizations, or in their wake, are more likely to launch viable rebellions than those that do not, all else equal, because of how the presence of criminals complicates the state’s ability to discern the severity of the threat.
An important next step is to probe the relevance of our argument in other contexts, especially stronger states. An implication of our argument is that where states have higher capacity to monitor their peripheral territory, barriers to entry will be higher for nonstate armed groups. Thus, we expect rebel group formation will be rarer but will involve better organized, pre-mobilized groups from the start—and will therefore escalate more quickly to large-scale violence. Contrasting Ethiopia’s recent outbreak of civil war in 2020 with its civil war of the 1970s supports this point: the most recent war exploded quickly in a state that had strengthened its local monitoring capacity, with a highly organized Tigrayan faction of the former national military taking on the state amid an intensely politicized national environment (Verhoeven and Woldemariam Reference Verhoeven and Woldemariam2022, 633–34). In contrast, the rebellion in the 1970s started with several years of small-scale insurgency in the same region of Tigray, but at that time the region was more remote from national politics and state presence, and had experienced decades-long “rampant banditry” (see the Supplementary Material to this paper, as well as Berhe Reference Berhe2009, 93–95; Crummey Reference Crummey and Crummey1986, 135).
The paper’s findings suggest several additional, promising avenues for future research. We propose three here. First, this paper has conceptualized the civilian response to nascent rebels in weak-state contexts as ambivalent and akin to civilian behavior toward bandits, diverging from classic assumptions of internal armed conflict emerging out of public contestation and a well-mobilized, politicized civilian population. More research that probes the relationship between violent and nonviolent contestation (e.g., Thurber Reference Thurber2021) while integrating more subtle or secretive forms of mobilization should put into focus the reasonableness of each conceptual approach under different conditions. This also indicates the importance of additional theoretical and empirical work that scrutinizes citizen preferences and behavior amid looming violence (e.g., Schubiger Reference Schubiger2021), and how citizens navigate uncertainty at the outset of potential armed conflict (e.g., Shesterinina Reference Shesterinina2016). In particular, mid-level theorizing and the marshalling of fine-grained comparative evidence promise to advance knowledge on the varieties of civilian mobilization under different conditions and stages of armed groups’ trajectories.
Second, it indicates why researchers should continue integrating the measurement and analysis of small-scale (or “low-level”) violence into the study of civil conflict onset and escalation. Although low-level violence may not initially seem as consequential as the large-scale violence that can follow, observing it can allow scholars and states to detect recently formed armed groups before they grow more violent. A fruitful step in this direction could be to extend new datasets about the nonviolent organizational foundations of rebellions, such as the Foundations of Rebel Group Emergence (FORGE) dataset (Braithwaite and Cunningham Reference Braithwaite and Cunningham2020) and the Anatomy of Resistance Campaigns (ARC) dataset (Butcher et al. Reference Butcher, Braithwaite, Pinckney, Haugseth, Bakken and Wishman2021), to capture precursors of resistance and rebellion that may appear in news media as “criminal” activity. For reasons at the heart of this paper, disentangling which specific acts of violence were committed by which nascent armed groups (and especially the “true” motives of each nascent group) is not likely possible for ongoing, contemporary conflicts. However, the paper also demonstrates that capturing early criminal and political armed group presence and interaction, especially in retrospect, is sometimes possible. Doing so systematically would allow researchers to probe how the trajectories and bargaining power of rebels rooted in criminal organizations may differ from those of groups with other organizational antecedents. Such work could also consider how criminal origins relate to the end of rebellions. Qualitative accounts of rebel trajectories are replete with mentions of remnants of mostly defunct insurgent groups that transition to banditry, suggesting a Darwinian process by which failed political specialists in violence are consigned to using their “skill set” for criminal acts (see, e.g., Daly Reference Daly2016).
Finally, this paper has implications for theories of armed group organizational practices. Seminal scholarship emphasizes the path dependence of either resource or social endowments for later insurgent behavior (Staniland Reference Staniland2014; Weinstein Reference Weinstein2007). Our findings build on this program, suggesting another area of skill endowment for nascent rebels along a dynamic process and within the murky boundaries between predatory and nonpredatory initial group “types.” In remote rural areas, criminals represent a pool of violent entrepreneurs who are skilled in the repertories needed for survival and are sometimes successfully incorporated into political groups through training and ideological appeals (Green Reference Green2018; Sanín and Wood Reference Sanín and Wood2014). This suggests the importance of continued work scrutinizing the varied pathways to conflict escalation (Lacher Reference Lacher2022; Shesterinina and Livesey Reference Shesterinina and Livesey2024), the contingent and relational features of armed group ideology (Parkinson Reference Parkinson2021), and the subtleties of rebel–civilian relations during the secretive outset of armed campaigns. Additional research could probe the ability of rebel organizations to incorporate and transform criminals (or former criminals) at the outset of violent campaigns, and how these dynamics influence downstream conflict processes.
Supplementary material
To view supplementary material for this article, please visit http://doi.org/10.1017/S1537592725102065.
Acknowledgments
For their helpful comments on this project, we thank Jessica Braithwaite, Kathleen Cunningham, Cassy Dorff, Alex Downes, Jesse Driscoll, Omar García-Ponce, Iris Malone, Aidan Milliff, Roger Petersen, and participants in the Organizational Turn in Conflict Studies workshop at the 2024 International Studies Association Annual Convention, the Multi-Method Conflict Research Consortium workshop at the 2022 Peace Science Society Annual Meeting, the Social Origins of Rebellion workshop at the 2020 Peace Science Society Annual Meeting, and the Comparative Politics workshop at George Washington University. We also thank several anonymous reviewers whose efforts improved the manuscript, especially R5, who provided five rounds of constructive critiques.
Open access
