Introduction

Tilly’s (1990, p. 208) observation that civil wars hollow out states needs little justification given their well-known intensity and intractability (Licklider 1993, p. 8). That Tilly (1990, p. 20) also observed that ‘administration diverts [warfighters] from war’ makes the fact that many stateless rebel groups conduct administrative actions as a facsimile of stateness during civil wars (Mampilly 2011) particularly strange. Stranger still is the notion that such administration might lead to postwar democratic transition, especially given the apparent tendency for former rebel political parties to display clientelism which Ishiyama and Basnet (2022) suggest is counterproductive to democratization.

Despite this, rebel governance has caught the eye of researchers who seek to elucidate why some countries emerge from civil wars in the midst of democratization and some do not. Some of the statebuilding literature suggests that democracy can only emerge when the postwar state has established certain institutional competencies (Cliffe and Manning 2008, pp. 172–176; Paris 2004).Others have argued that democracy comes from rebel efforts to cultivate a certain sort of civic culture during the war (Knight 2018). Still others have argued that the answer lies with whether rebels can afford to repress claims made by expectant populations (Huang 2016). This paper argues that the hypothesis emerging from the statebuilding literature is incorrect, that Knight’s hypothesis is conceptually flawed, and that Huang’s hypothesis is correct but partial. Instead, rebels can generate expectations that political goods are to be provided postwar by providing them during wartime, regardless of the degree to which rebels must rely on material contributions from civilians. Assuming the rebels persist after the conflict, this raises the costs associated with not providing them.

Put another way, while wartime institutional capital from rebel governors doesn’t necessarily presage postwar democratization, rebel governors can encourage democratization by making repressing dissent harder (or easier, as the case may be). Rebel governors who rely on civilian aid in the absence of foreign patrons have little choice but to respond to civilian claims. Similarly, rebel governors who establish a means of consultation also encourage democratization, not because it helps to illuminate civilians’ ‘true’ interests, but because it makes the prospect of consulting with one’s governor feasible to expect and, in doing so, difficult to deny should governors change their minds. Finally, rebel governors who establish proficiencies that make repressing dissent easier might make democratic transition more remote.

Rebel governance

From the outside, the fact that rebel groups would seek to develop governing apparatuses at all can seem curious. After all, why would a rebel army choose to spend scarce resources on something that does not directly aid their war effort, especially given the fact that if that effort fails, the group may well cease to exist? The answer appears to lie in the utility of appropriating state responsibilities and functions (Mampilly 2011, pp. 11–13). Revolutionaries and counterrevolutionaries alike have emphasized how important it is for rebels to maintain ties to civilians (Guevara 1998, pp. 10–12; Galula 2006, pp. 62–67), and one way to do so is for rebels to establish their own independent system of governance. Many 20th and 21st Century rebel groups took this advice seriously. Because of this, Mampilly argues that rebels stand to make substantial gains by establishing their own independent system of governance. By providing goods and services normally associated with the state or even by adopting the symbolic elements of statehood, rebel governors can project the legitimacy of their cause or undermine that of the incumbent state (Mampilly 2015).

In order to govern, the rebels must establish some degree of control over territory with civilians in it (Kasfir 2015, pp. 27–30). This is not an especially high threshold (guerilla war efforts also require some sort of “base area”) and Kalyvas (2006, pp. 132–138) notes that incumbents and rebels tend to have natural bases of support in cities and the countryside, respectively (at least initially) as the countryside tends to experience less consistent intervention by the central government. This represents a clear opportunity for enterprising rebels to displace or contest incumbent authority from the center.

Substantial ink has been spilt regarding what this effort would look like. Mampilly (2007, pp. 75–76) finds that rebel governance tends to mirror the country’s experience of governance previously—that is, if the state intervened in civilians’ lives, so should their new rebel governors. Arjona observes that this is a preferable state of affairs for rebels regardless of previous experience, but notes further that this can be limited by how well-consolidated local institutions are, regardless of their connection to the central government. Kaplan (2017) observes that these local institutions can provide any of a broad repertoire of means, violent and otherwise, to resist or constrain rebel would-be governors. In short, robust local institutions mean that rebels have to settle for occasional interventions into civilians’ lives. This is still preferable to anarchic disorder, but less attractive than what she describes as more robust ‘rebelocracy’ (Arjona 2016, pp. 26–29).

Mampilly (2007, pp. 18–20) observes patterns in the manifestations of governance in rebel-administered territory. Rebels seek to establish systems of policing and juridical functions, and infrastructures of public goods provision (often in that order), considering rebel governors to be effective if they successfully provide both. He also observes that rebel governors sometimes attempt to provide consultative instruments (or, as Mampilly describes them, feedback mechanisms). This third mode concerns the reactivity (if not necessarily the representativeness) of the institutions of rebel governance, as well as their provision of a means by which a population may interact and consult with its rebel rulers. These consultative instruments can manifest themselves as local elections or referenda, but they need not involve voting at all. Consultative bodies or procedures can also constitute this mode of governance (Table 1).

Table 1 Incidence of rebel governance
Full size table

Existing theories: statebuilding and social mobilization

The statebuilding hypothesis stems from the notion that rebels have incentives to build on their wartime institutional achievements toward ‘social contractual’ forms of state-civilian relationships (Olson 1993; Arjona 2014). In other words, prospective rebel governors have incentives to provide for their would-be citizens so as to reinforce their status as governors and this provision then ends up supporting the transition to democracy. As such, so long as a prospective rebel governor prevails in the conflict (Huang 2016, pp. 43–49), successful wartime administration would presage democratic transition.

On its face, this theory makes intuitive sense. Postwar statebuilding already requires immense institutional capital, especially when such a project has democratization as its goal (Cliffe and Manning 2008, pp. 172–176). Additionally, Paris (2004, pp. 187–207) argues that robust institutionalization is necessary, especially in post-conflict societies, for elections to not serve as catalysts for renewed conflict. It makes sense then that rebels who begin building their states before the war ends might give their postwar state something of a head start here.

However, the rebel statebuilding theory runs into several problems. First, the hypothesis rests on a necessary link between capacity and democratization which is tenuous at best. Capacity may be useful for an extant democracy (Tilly 2007, pp. 16–20; Linz and Stepan 1996, 3–15), capacity does not necessary beget democracy (Organski 1980, p. 72). Moreover, capacity can be just as useful (if not more so) for an autocracy as for a democracy (Huang 2016, pp. 107–109; Way 2002; Lyons 2016, pp. 172–176). Not only can effective and efficient governance assuage concerns regarding an autocrat’s ability to govern, but it can also be useful in suppressing criticism and dissent (Davenport 2007, pp. 488–491; 2010, pp. 77–79).

As such, it is perhaps unsurprising that recent research has not found much empirical support for the statebuilding hypothesis (Huang 2016; Knight 2018; Ishiyama and Widmeier 2020). Reyko Huang posits an alternative—dubbed the ‘civil mobilization’ theory—which also builds upon Olson, but to a different end. Rather than the institutions themselves (which Huang argues do not contribute to democratization), the critical element of rebel governance’s potential democratizing power stems from the rebels’ reliance upon the civilian population for funding and materiel. If rebels have no other choice than to rely upon the local population, then democratization is likely. If rebels rely primarily on foreign patrons or natural resources, democratization is unlikely (Ibid., pp. 10–11, 60–61). This stems from Olson’s notion of the stationary bandit—in order to maintain rents from a civilian population, the stationary bandits must prove their usefulness by providing good governance (Olson 1993).

Specifically, Huang’s theory argues that such rebel groups must make their case to civilians. If they get involved, even if it is not voluntary, Huang argues that the very act of supporting the rebels generates awareness, engagement, and expectations (Huang 2016, pp. 31–40). Awareness comes in the form of being aware of one’s political agency (or even one’s rights), as well as the possibility of alternative conceptions of the political space to what the incumbent government has decreed. Engagement means that supporters can no longer be detached from the political as well as the martial struggle that is the civil war. These processes also produce opportunities to make claims regarding how they should be governed, which generates expectations. If the rebels prevail (if they are victorious or if they secure a negotiated settlement to end the conflict), they will find repressing the civilians upon whom they rely too costly, which may in turn encourage democratization. To this end, it appears as though relying on civilian aid does correspond with increased rates of participation in postwar elections (Ishiyama and Widmeier 2020, pp. 58–60).

Huang’s theory represents a pathbreaking contribution to the subject. She argues that rebels who rely on civilians for material support and who lack foreign patrons allow civilian leaders to develop a proficiency in making claims of their rebel governors, ultimately convincing these governors that political goods are less expensive than material ones. While some rebel groups who rely heavily upon the civilian population do so while paying mere lip service to the greater good these gains may serve (Sabates-Wheeler and Verwimp 2014; Mampilly 2007, p. 100; 2011, p. 217; Fink 2001, pp. 46–47), civilians appear able to alter or resist rebel action by a variety of means (Arjona 2016; Kaplan 2017). Indeed, while Gowrinthan and Mampilly (2019) note that the norms and narratives rebel governors promulgate impacts civilian dissent, enterprising dissenters can likewise alter and reframe these norms and narratives to their purposes. As such, the primary fault with Huang’s theory is not in its correctness, but in its completeness. Reliance on civilian aid in the absence of foreign patrons does indeed make repression more expensive, but so does consultation.

The argument: consultation

Consultation makes the possibilities and impossibilities within a given political space mutable. Even if carefully curtailed by rebel governors, providing a means of consultation makes subsequent efforts to control a population or convince it that its political agency is unnecessary much more difficult. If rebels establish instruments by which civilians can exercise some degree of Tillyan consultation (2007, pp. 13–15), then it will be difficult (although not impossible) to put the rabbit back into the hat. The reason for this is not that the instruments themselves are resilient enough to survive official sanction during or after the conflict, but rather that their operation can cultivate the idea that consultation is possible. This in turn means that those pathways by which consultation was allowed (and the philosophical justification for their operation) become useful tools to enterprising would-be democrats. The claim-making that is facilitated here does not come from the provision of material goods, but rather the fact that such a pathway was opened in the first place.

Knight (2018) makes a related argument, positing that by eliciting feedback from civilians, rebels can cultivate a consciousness of their potential role in government. The argument here is distinct for two reasons. First, Knight (2018, pp. 32–33) argues that a rebel group’s consciousness-building activities’ effects can be felt after a civil war regardless of whether or not the group itself persists. Having had their consciousness raised, the argument goes, they will expect input into how they are governed regardless of whether or not the group doing the consciousness-raising still exists. The problem with this pertains to the expectations rebel governance cultivate, specifically that they pertain to the rebels themselves. While several authors have observed the strategic use of rulers’ narratives by civilians as a means of pursuing their interests (Scott 1990, pp. 96–103; Gowrinthan and Mampilly 2019), if the group weaving such a narrative ceases to exist, one would expect the narrative itself to lose its potency. In other words, civilians may want a say in an incumbent’s political order, but they are not likely to expect as much if it did not already exist. Further, the possibility of a victorious incumbent committing reprisals or otherwise cracking down on dissent (Toft 2010, pp. 46–49), this seems particularly unlikely.

Second, while not explicitly relying on the rightly derided concept of ‘false consciousness,’ (Scott 1990, pp. 96–103; Wickham-Crowley 1991, pp. 104–110) it nevertheless has civilians playing a somewhat passive role in the process. In Knight’s (2018, p. 6) theory, rebels take responsibility for cultivating what Almond and Verba (1963, pp. 14–26) would call a ‘participant’ culture in wartime. Huang (2016, pp. 16–49) is correct to argue that rebel governance’s role in democratization is not strictly top-down, but while Knight may have oversold rebel governors’ role in the process, Huang may have undersold it.

By providing a means of consultation, rebel governors do not illuminate ‘truer’ interests (e.g. Lukács 1971, p. 259; Freire 2005, pp. 47, 53–54). Rather, they make the pursuit of those interests a distinct possibility. Specifically, they create a means by which the boundaries of what is possible or impossible may be altered within a given political space. Hayward’s conception of de-faced power and freedom is particularly useful here. Rather than viewing power as a means of constraining free or authentic action, she argues that power is best described as ‘social boundaries that, together, define fields of action for all actors’ (Hayward 1998, p. 12). As such, she argues that inequity regarding power can be described as asymmetries in the ability to define or alter those boundaries. Consultative instruments, by introducing a de jure means to participate in governing, increases Haywardian freedom. That is, it increases the ability of civilians to alter those social limits which constitute power (Ibid., p. 21). Thus, consultation’s effect is less about raising consciousness (as Knight suggests) and more about providing a feasible means of strategically pursuing interests which predate a given would-be supporter’s conversion to the cause.

Why would a rebel governor opt to introduce such instruments? After all, the ballot box has ‘proven to be the coffin of revolutionary movements’ (Goodwin and Skocpol 1989, 495–496). Given the choice, rebels prefer to administer their territory in a top-down manner on their own terms rather than doing so contingent upon civilian consent (Arjona 2016, 26–29). Also, consultative instruments do not contribute to well-administered territory (Mampilly 2007, pp. 18, 37, 89–91). Sure, some revolutionary theories and ideologies explicitly call for civilian input as an essential piece of building a base of support (Mampilly 2007, pp. 11–13), but that is no guarantee that rebels will actively seek such input—particularly if that input is critical of their would-be rebel governors.

It seems that the political ‘good’ of consultation may be quite effective in marshalling support, even independently of the provision of other manifestations of rebel governance. The FMLN rebels in El Salvador’s civil war provide an instructive example. The group produced ample consultative instruments that attracted substantial popular support as it sought to ‘accommodate civilian feedback’ (Mampilly 2007, p. 103). While these means made FMLN ideology attractive and helped to deal with day-to-day issues in FMLN-controlled and contested territory, these instruments were also designed to provide political and military support for the war effort (Negroponte 2012, pp. 34–37).

Participation in El Salvador appears to have been a two-way street, though, as evidenced by Wood’s postwar fieldwork (Wood 2003, pp. 21–27, 83). Supporters did not merely express moral conviction that their cause was just and defiance of an authority they resented by their support, but they also expressed what Wood called ‘pleasure in agency.’ In other words, supporters felt that they contributed to the transformation of a social and political order which would ultimately produce a more just one. Even though each supporter’s individual contributions to the FMLN’s likelihood of prevailing were ‘vanishingly small’ (Ibid., p. 236), Wood reports that campesino supporters claimed ownership of the group prevailing in its war effort during these interviews. Finally, for many of these supporters, their role in the rebellion manifested as participating in insurgent cooperatives (Ibid., pp. 236–237).

These cooperatives represent a means by which wartime consultation introduces not just expectations of consultation, but also a degree of uncertainty, both of which philosophers of democracy have argued are essential (Dewey 1929, pp. 851–852; Barbu 2013, pp. 12–22). The cooperatives were the successors of Poder Popular Locales, which had provided both a means of marshalling support and of enlisting locals to participate in governing their locality (Todd 2010, p. 81; McClintock 1998, p. 60; Alvarez 1988). Government offensives spelled the end of the PPLs, but the insurgent collectives which replaced them functioned similarly, only with its participants taking even more responsibility for their operation (Wood 2003, pp. 174–192; Hammond 1998, pp. 110–119). One consequence of these cooperatives’ success was the formation of campesino political organizations, many of whom continued to covertly support the FMLN’s war effort despite being ostensibly independent of the group (Wood 2003, pp. 84–86, 160–177).

This development demonstrates two things. First, despite FMLN support of (and coordination with) these collectives, their operation means that one need not wait for the war to end before the expectations brought on by wartime rebel governance practices to manifest themselves. Second, it demonstrates that rebels can play an essential role in raising popular expectations of consultation, not by disabusing people of their ‘false consciousness’ but by making consultation demonstrably possible in a Haywardian sense and therefore, reasonable to expect.

Having attracted support with these political goods, rebels have raised expectations regarding their continued provision. This, as Huang (2016, p. 36) argues, makes failing to meet these expectations expensive. It also may allow rebels the opportunity to develop sophistication in political organizing in much the same way as she argues happens among civilians (Ibid. p. 35). This seems to have also been the case with El Salvador’s FMLN rebels. In addition to adapting after the Salvadoran government targeted the PPLs, the group appeared ready to transition to acting as a political party while the conflict was still happening. In 1989, the FMLN offered to participate in the 1989 presidential election, requesting a 6-month postponement to allow them to prepare for it (Wade 2008, p. 37). Shortly afterward, rebel commander Joaquin Villabos made the case that the group was in favor of democratic transition (Goodwin and Skocpol 1989). Indeed, despite internal divisions (Manning 2008, pp. 116–120, 134–137), the group held its first party convention the year after 1992’s Chapultapec Accords and competing in elections in March the year after (Wade 2008, pp. 38–40). While rebel political parties are not guarantors of postwar democratization (Ishiyama and Basnet 2022), it would seem that the FMLN was prepared for the transition to electoral competition.

An alternative: the law of the instrument

On the other hand, rebels may also seek to create the means of controlling a socio-political order within their controlled territory using policing and juridical functions. Indeed, Mampilly (2007, p. 20) observes that establishing a police force and legal system is the most common of the three goods of governance he describes. Using these policing and juridical functions, the governor of such a political order would have the wherewithal to prevent (or punish) activities deemed unacceptable within this order. Davenport’s work on state repression and the costs associated therewith makes this connection clearer—if you can administer an order, you have the tools to repress within that order. This not only reduces the costs of repressing dissent, it also makes dissent look like a problem whose solution is repression (Davenport 2007, pp. 488–491; 2010, pp. 77–79). Put in laymen’s terms, his application of the ‘Law of the Instrument’ can be summed up as ‘when one is a hammer, many problems appear as nails.’

This runs directly counter to the uncertainty which democracy requires. Dewey describes democracy as substituting ‘fixed subordination’ with ‘mutual consultation and voluntary agreement,’ both of which introduce the possibility of dissent and an element of pliability to the political and social order (Dewey 1967, pp. 674–677). On the other hand, an arrangement of fixed subordination is one in which the subordinated have no meaningful path to altering their state of affairs, at least within the realm of accepted protocols and behaviors.

From a practical perspective, applying capable policing and juridical functions to repressive ends may stifle mechanisms of democratization. As Kasfir writes, ‘[c]oercion greatly limits the extent to which civilians can modify rebel governance for their own interests’ (Kasfir 2015, 33). This would appear to short-circuit both the consultation process and the claim-making actions Huang (2016) describes as essential to democratization.

Moreover, the control that effective policing and juridical functions can provide rebels a justification for not extending political agency. For example, UNITA had a well-organized apparatus of policing and juridical functions—and impressive public goods provision in its controlled territory to boot. Even aside from UNITA leader Jonas Savimbi’s brutal application of his instruments of control against those he suspected of disloyalty (Pearce 2015, pp. 114–115, 146–147) or rivalry, regardless of the reasonableness of the accusation (Paul et al. 2014, pp. 168–170). His relentless assault on dissent (Minter 1991, pp. 141–143) may have been informed by their success in providing the goods of governance and the belief that this should have been sufficient to earn civilians’ unwavering support. To this point, Bakonvi and Stuvøy (2005, pp. 368–372) note that Savimbi’s use of violence in the service of security and stability within UNITA’s controlled territory did not necessarily undermine his legitimacy. Pearce’s interviews reinforce this point. One interview with a priest who lived in UNITA-controlled Jamba revealed that not only were atrocities committed by UNITA’s forces hand-waved away as being the inevitable result of war, but that ‘freedom of opinion in war is a utopia’ (Pearce 2015, pp. 112–113).

Hypotheses

This paper argues that democratization and autocratization postwar are determined by difficulty (or lack thereof) rebel governors have in repressing dissent, including calls for political agency. Three factors affecting this cost are considered.

H1:

Provision of consultative instruments by prevailing rebels accelerates democratization.

This hypothesis should capture the degree to which rebels make repressing dissent harder for themselves by introducing consultation. Consultation by its nature makes repression more difficult as it prescribes a means of airing and resolving grievances. Further, attempts to curb or do away with the means of consultation become costlier as the rebel governors themselves had made the possibility that civilians need not be merely the objects of governance feasible.

H2:

Provision of policing and juridical functions by prevailing rebels decelerates democratization.

This hypothesis should capture the degree to which rebels make repressing dissent easier for themselves by introducing the means by which it can be achieved. While not necessarily applied to repressive ends, policing and juridical functions can be leveraged to exercise control of a political order. This control can extend to marginalizing and/or punishing dissenting voices.

H3:

Reliance on civilian resources by prevailing rebels in the absence of foreign patrons accelerates democratization.

This hypothesis should capture the degree to which rebel governors must accept the liberalization of the political sphere in order to maintain their position. Having established their role in the rebellion, civilians would be in a position to demand such liberalization, as Huang has argued.

Design

This analysis extends Knight’s statistical analysis on the subject. In lieu of the Cox proportional hazard models Knight (2018, p. 76) used, this paper will conduct an analysis using Weibull models, as they can meaningfully outperform Cox models when the shape of the parameter is known (Crumer 2011). In this case, one would expect that the likelihood of transition to diminish over time. Civil wars and their resolution are periods of intense uncertainty in which long-accepted truths of a political order may find themselves upturned. As time passes, one would imagine that whatever order emerges from such a conflict would solidify absent another crisis.

The rebel governance data for this project comes from two sources. One is from Huang’s (2016) analysis. The other is from an extension (and correction) of Knight’s (2020) dataset. Knight’s rebel governance dataset, which selects civil wars from Hartzell’s (2014) Military Integration Dataset, attempts to capture both the presence and—to the extent that such a thing is possible using documentary evidence, performance—of rebel institutions with values ranging from 0 to 3. At the low end of the scale, the good of governance is not being provided by the rebel group in question. At the high end, the rebel governance apparatus is exceptionally well-consolidated, functioning as a ‘well-administered state’—unsurprisingly, Knight (2018, p. 48) found only one instance of such a manifestation. More common are scores of 2, which represent substantial provision that is not geographically or temporally isolated and conducted largely by way of rebel activity itself. The score of 1 denotes the presence of some apparatus of provision, but it is either incomplete (e.g. lacking health or education provision (but not both) in the case of providing public goods), geographically or temporally isolated, or provided largely by way of outside forces with the support or consent of rebels.

While this represents a noteworthy attempt to give a more nuanced picture of the conflicts themselves, doing so places an enormous burden on the primary and secondary sources regarding the conflicts in question, to say nothing of the work involved in interpreting these pieces. This is why, in addition to extending the dataset from its original 79 cases (ending between 1980 and 2006) to 100 (to include wars which ended in the 1970s), some of the scores for conflicts which appear to be issued in error have been corrected (Knight 2020).

For example, the original dataset noted that Sendero Luminoso was singularly unsuccessful at providing policing and juridical functions and only partially successful at providing public goods—namely, education—during the Peruvian Civil War (Knight 2018, pp. 227–228). In the case of the policing and juridical functions, Sendero’s trademark brutality and the fact that policing functions were ad hoc (Mampilly 2007, pp. 104–105) seem to have contributed to this score. This overlooks the fact that juridical functions and processes were fairly well established (McClintock 1998, p. 72; Strong 1992, pp. 120–123), handling both infractions regarding perceived infractions against the movement as well as grievances among civilians (Strong 1992, p. 87). Health provision also seems to have been overlooked. While decentralized and ad hoc (Ibid. pp. 87–88), Sendero’s efforts on this account outshone those of other rebel groups like El Salvador’s FMLN (McClintock 1998, p. 292). As such, the corrected dataset notes adjusted the policing and juridical score from 0 to 2 and the public goods provision from 1 to 2 (2020).

Where measures of democracy are concerned, this paper replicates Knight’s use of the Polity IV set, including separate analyses for polity2, Plümber and Neumayer’s (2010) alternative means of imputing missing data, and Vreeland’s (2008) x-polity ‘fix’ for civil wars. Democratization events are treated as the transition from a Polity score of 5 or below to 6 or above. I conduct an additional analysis for transition to open anocracy, treating such events as the transition from a Polity score of 0 or below to a score of 1 or above. This paper also uses Boix et al (2013, 2022) dichotomous measure of democracy, with transition events treated as going from a score of 0 to 1.

The use of different sources and interpretations of the democracy data is merited here as it allows for multiple different thresholds for transition without violating the assumption that the dependent variable in logistic regressions be binary (Rawat 2017). This is not to say that additional thresholds could not be devised using these sources—one could imagine, for instance using the positive change of a certain magnitude in Polity score as a threshold. This is not done in this case for two reasons. First, the transition from consolidated autocratic rule—a score of − 10—to ‘anarchy’ (Vreeland 2008, pp. 406–407)—a score of 0—would likely meet such a threshold for democratization while substantively not constituting such a development. Second, the Polity set’s protocols for imputing missing data is particularly susceptible to these kinds of errors, treating interregnum periods as deserving a 0 for their polity2 scores regardless of the scores for what came before or after this interruption (Plümper and Neumayer 2010).

Although there is some variance in the number of ‘failure’ events—that is transitions—depending on the data, transition threshold, or particular interpretation thereof, the range of outcomes greatly restricts the number of variables one can effectively use per model (Austin and Steyerberg 2015). While the classic ‘one in ten’ rule has been called into question, evidence suggests that bias becomes problematic once the ratio of events to variables drops below 5:1 (Vittinghoff and McCulloch 2006). As each data source, threshold, and interpretation report at least 30 transitions, the models will only include at most six variables per model.

At the same time, there is no limit to the number of models one can use to try and describe relationships within a set of data. As such, I have constructed several models which include different combinations of the following: policing and juridical functions, public goods provision, consultative instruments, a variable representing the presence of civilian aid and the absence of foreign aid, a dummy variable for negotiated settlement—as it appears to coincide with postwar democratization (Nilsson 2012) independent of rebel governance (Knight 2018, pp. 89–95)—a variable measuring the log of the number of months a war goes on, and a variable for the GDP/pc one year after the war’s end. While far from a perfect or elegant solution, this allows for a far more detailed picture of the relationships within the data to emerge.

The models themselves will be applied according to a particular order. First, Knight’s variables for consultative instruments and policing and juridical functions will be included alongside Huang’s civil mobilization variable. This is accompanied by controls for the natural log of war’s length in months, whether or not the conflict ended in a negotiated settlement, and Knight’s variable for public goods provision. The second model will replace the control for public goods provision with the gdp/pc of the country one year after the war. The third model cuts Huang’s variable and includes all four of the controls. The fourth, fifth, and sixth models each include all the non-rebel governance variables as controls alongside consultative instruments, then policing and juridical functions, and finally Huang’s civil mobilization variable.

Each mode of rebel governance (including the control for public goods provision) is interacted with a variable representing the rebel group’s prevailing in the conflict. Both the consultation-based theory posited here as well as Huang’s (2016, p. 31) civil mobilization theory presupposes that the rebel group prevails in its conflict if rebel governance is to contribute to postwar democratization. In turn, this ensures an accurate reproduction of Huang’s analysis.

Results

The results are arranged with coefficients reported. The variables for H1 (consultation), H2 (law of the instrument), and H3 (civil mobilization) are bolded and abbreviated. The results provide support for all three hypotheses, albeit to differing degrees. The analysis provides evidence consistently confirming H1. It also provides some evidence in support of H2, but to a much lesser degree. It also provides evidence confirming H3, albeit with an important caveat.

The results using the polity2 measure consistently confirm that consultative instruments accelerate transition to both democracy (Table 2) and open anocracy (Table 3), whether alongside other forms of rebel governance or in isolation. Policing and juridical functions also appear to decelerate transition to more democratic forms of government, although the significance drops out when the variable is treated in isolation from other rebel governance. Civilian aid in the absence of foreign aid does not perform as well, only appearing to accelerate transition to open anocracy and only in isolation from other governance-related variables.

Table 2 Polity2, democracy
Full size table
Table 3 Polity2, open anocracy
Full size table

Vreeland’s (2008) X-Polity measure (Tables 4 and 5) sees the models produce similar results where consultation is concerned, as they consistently depict consultative instruments as accelerants to open anocracy and democracy. Policing and juridical functions appear to slow these transitions down, but not consistently. When isolated, the significance of the relationship drops out. The models that aggregate multiple governance variables together perform better, but even this is inconsistent. Two such models report statistical significance where democratization is concerned and only one reports significance where the transition to open anocracy is concerned. Civilian aid appears to accelerate transitions to more democratic forms of governance, but the models only report statistical significance for transitions to open anocracy. All but one of these models report that civil mobilization accelerates such a transition.

Table 4 X-Polity, democracy
Full size table
Table 5 X-Polity, open anocracy
Full size table

Plümber and Neumayer’s (2010) fix for the polity set (Tables 6 and 7) also has the models reporting that consultative instruments accelerate transitions to more democratic forms of government, although the significance drops for the model that isolates it for the purposes of describing relationships to the transition to open anocracy. The models only report that policing and juridical functions decelerate transitions to democracy and open anocracy in the models which include consultation and civil mobilization variables but not the control for public goods provision. Civil mobilization appears to accelerate transition to more democratic forms of government, but the models only report a statistically significant relationship where the transition to open anocracy is concerned.

Table 6 Plümber and Neumayer, Democracy
Full size table
Table 7 Plümber and neumayer, open anocracy
Full size table

Finally, the model using Boix Miller and Rosato’s democracy data (Table 8) reports that consultation accelerates democratization, although the significance drops out when the variable is isolated. Policing and juridical functions appear to decelerate such transitions, but only one model reports a statistically significant relationship. At the same time, two additional models—including the one treating it in isolation from other governance variables—report a relationship that approaches significance (0.1 < p < 0.15). The models here do not report a statistically significant relationship between civilian aid in the absence of foreign aid and democratization.

Table 8 Boix, Miller, and Rosato
Full size table

In sum, consultation appears to accelerate transitions to more democratic forms of government. This appears to be fairly consistent, regardless of the model’s inclusion or exclusion of variables or the threshold of transition. Policing and juridical functions may decelerate such transitions, but these models provide far more limited evidence of this relationship. In fact, the most consistent statistically significant result comes from one particular configuration—a model that puts the variable alongside consultation and civil mobilization but excludes public goods provision. A couple of models show that public goods provision appears to delay democratic transition. These incidences are too infrequent to draw definitive conclusions, but there may be some truth to the notion that rebel governors may be able to provide material goods in lieu of political ones.

As it happens, the civil mobilization hypothesis finds its clearest supporting evidence in the transition to open anocracy. It would seem that rebel reliance on civilian aid in the absence of foreign patrons is a fairly reliable accelerant to partial democratic transition. This analysis provides far less evidence supporting the supposition that civilian aid in the absence of foreign aid presages the full transition to democracy, however. As such, rather than refuting the civil mobilization hypothesis, these results provide fairly consistent—if contingent—support for it.

The shape parameter appears to confirm over time the likelihood of transition diminishes. The particular shape parameter varies from between 0.6 to 0.86 depending on the democracy data and variables considered. As each reported shape parameter is less than 1, each analysis suggests a decreasing failure rate as time progresses (Weibull.com 2002). With that said, the shape parameter is not consistently reported as statistically significant throughout all the analyses. Some report that the shape parameter merely approaches statistical significance (0.1 < p < 0.15) and a few do not even approach significance. The consistency otherwise suggests that the postwar window of opportunity is real, but these inconsistencies are worth noting regardless.

As a final note, there are also several (six) models that do not confirm that the variables selected provide a reliable basis of predicting transitions. Three of these are models which only include policing and juridical functions as a rebel governance variable, one includes only the consultative instruments variable, one of which only includes the civilian aid variable, and one includes both a policing and juridical functions variable and a consultative instruments variable. This suggests that any conclusion regarding policing and justice should come with a grain of salt, as when isolated from other variables it appears to not reliably predict transition (or in this case, a delay in one).

Discussion

Rebels may not always have a choice regarding whether their rebellion will produce postwar democratization. Huang (2016) made waves arguing that rebels who lacked foreign patrons eventually had to provide political goods in exchange for material ones extracted from the civilian population. However, this analysis suggests that unless those political goods were being provided by rebels during the conflict, this reliance accelerates the transition to more democratic forms of governance but not to democracy. Wartime provision of consultation, in demonstrating the possibility of self or popular government, appears to accelerate postwar democratization regardless of whether or not the rebel group intended the outcome in the first place.

That both developments correspond with accelerated transitions of some kind suggest that rebel efforts which raise the cost of repression make the postwar transition to more democratic forms of government more likely. The same can not necessarily be said regarding making dissent easier. Although it makes intuitive sense that rebels that cultivate the means by which dissent can be repressed should be better prepared to do so, these results suggest that this is not necessarily the case. The availability during wartime of the means of repressing dissent does not appear to reliably predict the application of those means postwar, at least as far as delaying democratic transitions is concerned.

With that said, there are some reasons that these results should be taken with a grain of salt. First, the civil wars examined herein are getting smaller in the world’s collective rearview mirrors. While this extension represents an increase in the dataset’s coverage, extending the dataset further backward and forward in time would make these results more robust.

Second, there are alternative datasets worth exploring for replications of this analysis. For example, while the analysis relies on multiple measures of democracy, the conceptually dense and versatile Varieties of Democracy data may be worth employing to this end. Additionally, Albert’s (2020) dataset of rebel quasi-state institutions may be worth examining as an alternative source for rebel governance data. As such, while these results provide additional clarification regarding the nature of rebel governance’s connection to postwar democratization, stronger conclusions will require a more ambitious project than what can be achieved here.

Also, it is worth noting that the analysis here is not without endogeneity concerns. For instance, robust rebel governance during a war may presage an eventual democratic transition, but only after an intervening conflict or other catalyzing event. While the fact that the shape parameter is somewhat consistent and generally statistically significant suggests that this is not a terribly common incident, it certainly has happened. Additionally, multiple authors have noted that rebel governance strategies are determined in part by their experience of governance from the incumbent government (Mampilly 2011; Arjona 2016). As such, postwar rapid transition might be a product of relatively (but not definitively) democratic governance practiced by the prewar government rather than the result of enterprising rebel governors.

Nevertheless, this analysis suggests that circumstances and governance strategies can facilitate the transition to more democratic governance based on how difficult it is to ignore civilians’ demands or suppress dissent. This is not to suggest that as democracy continues to flag globally (Boese et al 2022, pp. 9–13) internal conflicts may have a democratizing silver lining—the horrors of war make such a suggestion ghoulish at best. Instead, this research may help illuminate the answer to the question of why some postwar states democratize and others do not. It seems that even after civil wars hollow out their hosts, democratization comes down to how hard it is for leaders to ignore their people.