Dangerous lessons

Abstract

Contemporary research on the international diffusion of civil conflict privileges physical, proximity-based variables like conflict spillover and refugee flows. I argue that the international diffusion of civil strife can occur through a learning mechanism, and that this phenomenon may occur on a worldwide basis. Individuals most likely to rebel, or proto-rebels, may learn about the utility of rebellion before mobilizing and violently challenging the state. Such international learning therefore occurs during the pre-conflict process. International sources by which proto-rebels learn include active, ongoing civil wars, as well as revolutionary governments that have been founded by rebels victorious in past wars. Revolutionary governments radically shock the international system, teaching proto-rebels about the possible benefits to be gained from violently challenging the state. In order to test these assertions, I undertake empirical analyses using militant organization data that capture the year in which rebel movements first emerge, during the period 1968–2001. I then explore the spatial and temporal relationships between rebel movement emergence, civil conflict, and revolutionary regimes, using the country-year as my unit of analysis. I further examine how these relationships are attenuated by cultural and regime-type similarity. I find, in line with the literature, that active civil conflicts generally inspire rebel mobilization only in directly neighboring states, while revolutionary regimes established after rebel victories are associated with mobilization on a global basis. I conclude that proto-rebels learn and take inspiration from some global sources of information, and that significant analytical utility is to be gained by focusing on revolutionary regimes established as a result of rebel victories, as well as mobilization in the pre-conflict process.

Keywords

civil war diffusion learning militant group emergence

Introduction

In 1996, the Communist Party of Nepal (Maoist) launched a rural-based insurgency against the monarchical government of that country. Evidence would later show that this group conscientiously emulated the mobilization strategies and warfare tactics of several contemporary Maoist groups, including the Naxalites of India, the Khmer Rouge of Cambodia, and, most interestingly, the Sendero Luminoso of Peru. Although Peru and Nepal are separated by thousands of miles and differ in many key respects, insurgents in both cases learned techniques of mobilization from one another and both employed the rhetoric and doctrines originally developed during the Chinese Revolution (Marks & Palmer, 2005).

This kind of learned mobilization is not limited to Maoism or even to active insurgencies. The overthrow of regimes, the secession of new states, and the survival and persistence of revolutionary regimes has historically exerted similar effects, inspiring militants to acts of violence throughout the world. A prominent example is in the mobilization of dissidents across Latin America after the success of the Cuban Revolution, contributing to the creation of militant groups like the Uruguayan Tupamaros and the Nicaraguan Sandinistas (McSherry, 2005).

Together, these anecdotes suggest a puzzle; namely, that civil conflicts and revolutions appear to inspire mobilization in dissidents contemplating rebellion, or proto-rebels as I term them, even in countries that are otherwise unconnected. The existing conflict literature offers little insight into these puzzling linkages, instead focusing on those physical conditions responsible for the spread of conflict among geographic neighbors (Buhaug & Gleditsch, 2008; Salehyan, 2009). While conflict is indeed contagious among neighboring states, it may also diffuse among non-neighboring states, as in the recent Arab Spring or in the observed connection between Peru and Nepal.

In order to gain leverage over this puzzle, I argue that proto-rebels learn from internationally available information on conflict. This learning mechanism is framed according to the logic of collective action. Although proto-rebel entrepreneurs may employ dozens of possible solutions to the collective action problem, the state possesses a decisive advantage in terms of policing power and military force, rendering militant mobilization and the pursuit of violent strategies uncertain (Lichbach, 1995). Proto-rebel entrepreneurs may thus look internationally for information on the expected utility of rebellion, and then provide it to their followers.

Although the existing literature has engaged in a search for evidence that proto-rebels learn from global sources of information (Buhaug & Gleditsch, 2008; Danneman & Ritter, 2014), it has been unable locate evidence of such, instead consistently finding that diffusion is the result of direct spillover from neighboring states. By contrast, I argue that proto-rebels may learn from two sources: first, from ongoing civil wars, which demonstrate how and when the collective action problem might be overcome; second, from regimes successfully established by rebels, which demonstrate the benefits of rebellion. If rebels successfully displace a ruling government or form their own state through secession, a powerful example is created that may then be learned from by proto-rebels globally. In other words, a key missing element in the puzzle of global learning is the revolutionary regime.

This article is structured as follows. First, I motivate the study with a look at the international spread of civil conflict and the effect that revolutionary regimes have upon it. Second, I develop a theory in which proto-rebels learn from and mobilize in response to international events. Third, I describe a research design using several unique variables. Fourth, I engage in a quantitative analysis. Finally, I offer a concluding discussion.

Learning and the diffusion of civil conflict

Interdependence and transnationalism are among the defining features of the international environment – yet, these phenomena are studied primarily with reference to positive developments like international liberalism (Simmons & Elkins, 2004). By contrast, the dark side of transnationalism has received attention only recently. Efforts in this area typically focus on international terrorism (Brandt & Sandler, 2010; Braithwaite & Li, 2007; Enders & Sandler, 2006; Findley, Piazza & Young, 2012; Krieger & Meierrieks, 2011; Midlarsky, Crenshaw & Yoshida, 1980; Neumayer & Plümper, 2009, 2010; Piazza, 2008), and so a detailed exploration of the various causal pathways operating at the transnational level is a welcome development. These pathways include such phenomena as the international spread of civil conflict and repression (Beissinger, 2002; Black, 2013; Braithwaite, 2010; Buhaug & Gleditsch, 2008; Danneman & Ritter, 2014; Forsberg, 2013; Kuran, 1998; Lake & Rothchild, 1998; Maves & Braithwaite, 2013; Gleditsch, 2007; Salehyan & Gleditsch, 2006; Salehyan, 2009), and the global transmission of tactics (Horowitz, 2010), ideologies (Kalyvas & Balcells, 2010), the movement of foreign fighters (Hegghammer, 2013), demonstration effects (Beissinger, 2002; Kuran, 1998), and even the possibility that rebels emulate others and adopt their rhetoric in order to draw upon transnational support (Bakke, 2013).

At the theoretical core of civil conflict’s international spread is a phenomenon called ‘diffusion’, which is defined as a process in which an event occurring in one place alters the probability of its occurrence in another (Elkins & Simmons, 2005; Gilardi, 2012; Strang, 1991). The basic social science literature on the topic identifies more than 30 potential mechanisms of diffusion (Elkins & Simmons, 2005). The most relevant of these for the study of conflict are migration, emulation, and learning (Wood, 2013).

Migration is the most studied mechanism. Here, political violence in one country creates cross-border externalities that impact the internal environment of another country; thus, conflict spreads in an almost disease-like fashion. Civil conflict has been shown to drive refugee flows into neighboring states, disrupting them and rendering conflict in the recipient state more likely (Salehyan & Gleditsch, 2006). Moreover, civil violence has been shown to transmit from one state to another via cross-border ethnic ties (Buhaug & Gleditsch, 2008).

While migration is well explored, learning has been neglected. Learning is defined as a process in which policymakers use the experience of other countries to estimate the likely consequences of a policy innovation (Gilardi, 2012). Decisionmakers resort to this search for information because the outcome of any particular action is uncertain. Observing others vicariously allows actors to evaluate possible courses of action. A rich line of literature explores the nature of this search, labeling such phenomena ‘demonstration effects’ (Beissinger, 2002; Black, 2013; Danneman & Ritter, 2014; Forsberg, 2014; Kuran, 1998; Lake & Rothchild, 1998; Maves & Braithwaite, 2013). In the simplest version of this mechanism, decisionmakers have prior beliefs but then update them based upon the new data (Elkins & Simmons, 2005).

The learning mechanism is also subtly different from emulation. Emulation is a mechanism in which actors adopt a policy not because they believe it is the best option, as learning implies, but because actors attempt to signal conformity in a social system. Emulation is therefore a mechanism in which diffusion and interdependence are socially constructed, rather than arising from a rational calculus (Wood, 2013).

With the scope of learning now defined, I examine the international sources that proto-rebels observe. The literature has focused on active sources of civil conflict as a source of learned information (Black, 2013; Maves & Braithwaite, 2013; Danneman & Ritter, 2014). One overlooked source of information is that of the revolutionary regime. Indeed, if rebels are successful in their efforts, violently overthrowing their government or seceding into a new state, a powerful global example may be provided to other proto-rebels. To the extent that this question has been studied, it is limited to the precedent-setting effect of ethnic secession, with the evidence suggesting a very limited impact (Forsberg, 2014; Saideman, 2012). Yet, history is replete with examples of the above phenomena that go beyond ethnic conflict and the bounds of geographic contiguity. Consider the link between the American and French Revolutions in the late 18th century (Dunn, 2000). It is commonly argued that the American victory over British forces positively reinforced dissidents in Europe, encouraging them to think that success against a major monarchical power was possible.

Together, these anecdotes suggest that rebel victory in civil war can radically shock the international system. Successful attempts at revolution or secession can create a precedent, encouraging similarly aggrieved proto-rebels to attempt rebellion. Although new research suggests that learning and demonstration effects are particularly important for proto-rebels in authoritarian regimes (Maves & Braithwaite, 2013), transnational learning and its consequences remain a relatively unexplored dimension of conflict diffusion, as the literature focuses primarily upon material variables.

Theory

Upon what information do proto-rebels base their decisionmaking? To answer this question, I develop a theory grounded in expected utility, in which decisionmakers judge courses of action according to their estimated costs and benefits, and a version of the collective action problem called the Rebel’s Dilemma (Lichbach, 1995), in which proto-rebels are subjected to poor information, free-riding, and the power of the state, each of which militate against mobilization. When decisionmakers labor under such uncertainty, they must engage in a search for successful strategies (Rogers, 2003). In doing so, proto-rebels learn about the utility of rebellion. Lichbach (1995: 118–120) argues that this process occurs for a plethora of reasons: rebellion may increase expectations among proto-rebels by signaling to them that others are primed for dissent; active dissent provides ideas to proto-rebels; the success of a rebel group at time t increases another’s estimate of winning at time t + 1; and finally, successful dissidents are able to act as principals and patrons for subsequent proto-rebel groups.

A complete theory of learning contains multiple elements. First, it recognizes that rational actors engage in vicarious observation, estimating utility from the perceived success or failure of events in the past. Second, actors are bounded in their cognitive processing abilities and thus reliant on heuristics and analogies to filter learned information (Gilardi, 2012). In recognizing these elements, this article departs from a purely rational theory of learning. Rational theories of learning contend that actors observe the probabilities of success, costs, and benefits of actions undertaken by decisionmakers in the past, and then update their own estimates of utility in a fashion similar to the laws of statistics (Elkins & Simmons, 2005; Gilardi, 2012; Simmons & Elkins, 2004; Weyland, 2010, 2014). By contrast, under the bounded rationality advanced by this article, decisionmakers do indeed learn from the experiences of others, but they are limited in their cognitive ability to do so.

While bounded rationality provides the social scientific basis for learning-by-analogy, it also introduces the possibility of agent error, in which proto-rebels learn incorrectly and mobilize for violent conflict when the actual odds of success are quite low (Signorino, 2003: 322). This logic mirrors well-known studies on learning in the foreign-policy decisionmaking literature, in which leaders attempt to apply the lessons of the past, even if that analogy has only superficial relevance to the current situation (Levy, 1994). Incorrect learning poses an important puzzle for scholars, and several have attempted to build it into their theoretical and empirical models (e.g. Bas, 2012).¹

The problem of mistaken learning by the proto-rebel becomes all the clearer when one considers the reaction of the state. States are as aware of international information as are proto-rebels, and may deploy pre-emptive repression in order to forestall rebellion (Danneman & Ritter, 2014). One interesting consequence of state repression is its well-known capacity for inflaming the situation and pushing neutral civilians into the arms of the rebellion (Mason & Krane, 1989). Nor is it the case that proto-rebels mobilize only when victory is certain. Indeed, conflict is known to result from uncertainty over power and capabilities (Fearon, 1995). By one account, only 28% of all civil wars are won by rebels (Mason & Fett, 1996: 256). Rebellion might therefore seem an insurmountable puzzle, yet proto-rebels might be convinced that a positive result is possible if evidence of such is available. I therefore argue that although the possibility of agent error and state repression stand in the way of proto-rebel learning, they do not counteract the possibility.

I turn now to the task of positing hypotheses. If proto-rebels are boundedly rational actors seeking out information from the international environment, then they will extract it from events occurring in states whose characteristics can be analogized to their own. In other words, proto-rebels operate as though events in places similar to their own are representative of their situation. Similarity is therefore the foundational element of bounded learning, describing those conditions under which proto-rebels are likely to process information from abroad and then mobilize.

Similarity-based bounded learning may occur globally, without respect for the geographic distance between actors. Indeed, the literature argues that similarity-based global learning may occur within lingual groups, ethnic diasporas, religious creeds, or even ideological communities (Hegghammer, 2013; Horowitz, 2010; Kalyvas & Balcells, 2010; Kuran, 1998; Lake & Rothchild, 1998; Weyland, 2010, 2012, 2014). It is also possible that proto-rebels mobilize when they observe active rebellion occurring in a state with a similar regime type. This was plainly evident during the Arab Spring, when dissidents acted out against personalist dictatorships, during the revolutions of 1848, which saw mobilization against monarchies, and during the events of 1989, in which mobilization occurred against communist-party dictatorships (Kuran, 1991; Saideman, 2012; Weyland, 2010, 2012, 2014). From this discussion, I therefore posit the following hypothesis:

Hypothesis 1: As the number of civil conflicts or revolutionary regimes increases in similar states worldwide, the greater the likelihood that proto-rebels will mobilize.

Although some of the above anecdotes are very dramatic, the nature of international information constrains the learning mechanism. For instance, proto-rebels may be isolated by a remote geographic location and so news of global revolutions may not reach them. Moreover, it is possible that even if learning is occurring, its impact is overshadowed by the direct, material factors that operate in the regional environment. For example, rebel success may very well produce mobilization in a neighboring state, but the mechanism is drowned out by refugee flows. This logic leads many to conclude that diffusion is limited by the geographic distance between the site of a conflict and the potential site of new conflict, and that the potential for learning and demonstration effects is quite limited (Ayres & Saideman, 2000; Buhaug & Gleditsch, 2008; Forsberg, 2013; Saideman, 2012). Thus, I reason that although the emergence of revolutionary regimes may have a global effect, those effects will be felt most notably in geographically proximate states. Proto-rebels near a revolutionary state are more likely to be alerted to the possibilities of revolutionary action if it occurs close at hand. By way of this logic, I hypothesize as follows:

Hypothesis 2: As the distance of a proto-rebel from the source of learned information increases, the less likely a proto-rebel is to rebel.

Revolutionary regimes may further provide a way for proto-rebels to learn from sources beyond the local region. For example, Weyland (2009) argues that stunning rebel success – and the establishment of revolutionary regimes – inspires in even distant proto-rebels an almost euphoric desire to topple their own regime and a willingness to take risks to do so. When such sources of information are present in the international system, the probability increases substantially that their effects will penetrate the noise of international politics, rising above the constraints imposed by geographic proximity.

Although international revolutions and civil wars may or may not have a limited appeal to domestic audiences, the fear of successful revolution is actually a major force in foreign and domestic policy. For example, regimes threatened by conflict abroad may bolster themselves by transforming into a police state, undertaking military intervention, or even engaging in foreign policy balancing against the perceived threat. Long-distance learning by proto-rebels is thus a process with well-known international consequences (Kathman, 2010).

Very few scholars have studied the relationship between non-contiguous conflict diffusion and the existence of revolution regimes using quantitative methods. Forsberg (2013) and Black (2013) each consider some aspect of the question, but they find, respectively, that evidence of such is non-existent, or very rare. To an extent, this is a product of the dependent variable in each study, that of armed civil conflict. By contrast, I am interested in the global diffusion of proto-rebel collective action which, by definition, occurs prior to the onset of conflict.

There is some anecdotal evidence in favor of my theorized relationship. Ted Robert Gurr notes that Ghana’s independence in 1957 raised the expectations of political independence among Africans throughout the continent, indirectly contributing to political violence in places as far away as the Belgian Congo or Angola (Gurr, 1970: 97). And this phenomenon is not limited to the modern era with its instant communications and easy travel. As far back as the 1790s, the Marquis de Lafayette threatened to present Europe with the ‘contagious example of a dethroned king’ (Haas, 2005: 7).

I therefore anticipate that although proto-rebels may extract information from many different sources, including active civil wars, revolutionary regimes are a more significant example from which to learn. In this way, revolutionary regimes may inspire proto-rebel mobilization globally. I therefore posit the following additional hypothesis:

Hypothesis 3: Proto-rebel learning from revolutionary regimes is less likely to be degraded by geographic distance.

Data and research design

In this section, I describe data and a quantitative research design that assesses the above hypotheses. I anchor my analysis to the replication data provided by Buhaug & Gleditsch (2008). The dataset contains a global sample of 6,591 state-years covering the years 1950–2001, although this sample is reduced depending upon the coverage of the variables. Because my theoretical focus is on the timing of proto-rebel mobilization, I use data describing the timing of the emergence of militant organizations in my dependent variable. I use two general classes of independent variables: one that captures similarity of states containing proto-rebels to those undergoing armed conflict, and one that captures similarity of states containing proto-rebels to those hosting a revolutionary regime.

Dependent variable

In order to measure proto-rebel mobilization, I use data on the emergence or formation of militant organizations. Such data capture the moment, as closely as possible given current data, that proto-rebels overcome the collective action problem and assemble an organization. There are two reasons why researchers should consider a dependent variable like militant group emergence, rather than the onset of armed conflict. First, the onset of armed conflict does not capture the theorized mobilization process. Armed conflict onset is an escalatory, action–reaction sequence that is observed in the common datasets only after that conflict yields a certain number of battle-deaths in a given year.² By such a definition, learning and mobilization have already occurred. If proto-rebels are indeed learning from global information, then it is necessary to obtain information from a stage of the conflict process prior to the onset of armed conflict or war.

I collect data on militant group emergence from a variety of sources. The primary source of these data is the Jones & Libicki (2008) study on terrorist group life-cycles. Groups in this dataset are identified according to a definition of terrorism in which organizations employ violence of a ‘political nature [involving] the perpetration of acts designed to encourage political change’ (Jones & Libicki, 2008: 3). Under this definition, 648 organizations are identified. The temporal domain of these data is the late 1960s until 2006. I therefore choose 1968 as the starting point for my analysis.

In order to extend the empirical domain of this study, I expand the Jones & Libicki (2008) data by adding the founding date of 248 additional groups. Sources of this expansion include the Terrorist Organization Profiles (TOPs), a source available from the National Consortium for the Study of Terrorism and Responses to Terrorism (START),³ as well as the list from the Federation of American Scientists (FAS) of Liberation Movements, Terrorist Organizations, Substance Cartels, and Other Para-State Entities.⁴ This expansion effort yielded 896 total organizations. I then prune from the data those groups whose status as an actual organization is questionable.⁵ After carrying out this operation, 623 groups remain.

Although the use of terrorism data may seem questionable in a civil war study, research has suggested considerable overlap between the two concepts (Butler & Gates, 2009; Findley & Young, 2012), with recent works arguing that the pursuit of insurgency or terrorism is a tactical choice by rebels that is dependent upon geographical or regime factors (Bueno de Mesquita, 2013), the success of the state’s repressive apparatus (Carter, 2015), or the rebel ability to control territory (Sánchez-Cuenca & De la Calle, 2009). It is appropriate, then, not to limit the empirical scope of a study of political violence based upon the tactical choices made by rebels, particularly at that point in time when they are only just beginning to achieve collective action and, almost by definition, lack territorial control. With this research complete, I obtain a 0/1 variable indicating the emergence of one or more groups in a given state-year. Few states experience more than one group formation in any given year; thus, collapsing the data in this way is justified. I call this variable Militant group emergence. Figure 1 shows the frequency of Militant group emergence over time within this dataset.

There are alternatives to this measure. Data on terrorist groups have been widely used in studies on the life-cycles of such organizations (e.g. Aksoy, Carter & Wright, 2012; Aksoy & Carter, 2014; Blomberg, Engel & Sawyer, 2009; Blomberg, Gaibulloev & Sandler, 2011; Boutton, 2014; Carter, 2012; Cronin, 2009; Gaibulloev & Sandler, 2013, 2014; Young & Dugan, 2011, 2014). Typically, these studies merge a list of militant groups with a list of terrorist attacks, then key the date of group emergence to its first attack. Yet, as Dugan (2012) notes, this approach has limitations in that it conflates

Figure 1.

System-wide militant group emergence, 1946–2006

group emergence with the first attack attributed to that group. My approach, mirroring that of Jones & Libicki (2008: 5), counts a group as having emerged based upon the first indication that the group exists and is dedicated to the use of violence. However, for the sake of robustness, I make use of the Young & Dugan (2011) data on terrorist groups. This dataset combines a list of actors extracted from the Global Terrorism Data (GTD) (e.g. LaFree & Dugan, 2007) with attacks listed in the same source. This approach is also used by Aksoy, Carter & Wright (2012), Aksoy & Carter (2014), Boutton (2014), and Carter (2012). Tests using this alternate dependent variable are included in the Online appendix.

Independent variables

Similarity

In order to operationalize learning, I construct measures of the similarity of proto-rebels to each informational source. I define the sources of information from which proto-rebels learn as follows: (a) information emanating from armed conflicts in other countries; and (b) information emanating from revolutionary regimes.

Armed Conflicts are identified using the Uppsala Conflict Data Program’s (UCDP) Conflict Termination Data (Kreutz, 2010). This dataset contains episodes of ongoing conflicts during the 1946–2010 period. As this theory does not address proto-rebels inside the government conspiring to replace it, as occurs in a coup, such cases are excluded from the data.

States hosting revolutionary regimes are identified for the entire 1946–2011 period, and coded as originating from civil wars in which rebels are victorious, as defined by the Conflict Termination Data (Kreutz, 2010).⁶ The start-year for a revolutionary regime is keyed to the termination year of those conflicts that UCDP identifies as rebel victories. The termination date for each revolutionary state is keyed to changes to the form or type of government brought to power by armed conflict, as determined by regime-data contained in Colgan & Weeks (2015) and Geddes, Wright & Frantz (2014), or a 30-year cutoff that I impose. I follow this protocol because revolutionary regimes will undoubtedly decline over time in their appeal to foreign dissidents.

There are some alternatives to this coding scheme. For example, Colgan (2012), in his study of revolutionary regimes, defines such governments as those that include a revolutionary leader in their governing structures, and which pursue an agenda of radical social and political change. The result, for Colgan, is a set of regimes that includes those that came to power by methods other than military victory. My theory, however, is strictly limited to those that originated in violent regime change. Nevertheless, I execute a series of robustness checks in which I use Colgan’s list of revolutionary regimes in place of my own. These results show no statistically significant linkage between proto-rebel mobilization and this alternative list of regimes. This builds support for the idea that proto-rebels globally learn from sudden and violent regime change.

My coding criteria result in set of regimes in which rebels overthrow the government, as well as several new states formed from secession. These regimes are listed in the Online appendix. There are 61 regimes in the data, with a minimum duration of 0 complete years, a maximum of 30 years, a mean of 10.5 years, and a standard deviation of 8 years. I collapse the data in such a way as to yield a state-year variable, Revolutionary regimes, which is a count of the number of regimes persisting in the international system. Figure 2 reports the trend in the frequency of Revolutionary regimes for the period 1968–2001.

Figure 2.

Frequency of revolutionary regimes per year, 1968–2001.

The number of such regimes is relatively low early in the history of this time frame, jumping at two key points – the late 1970s, which saw revolutions in Iran and Afghanistan, and the early 1990s, a period characterized by new state formation at the end of the Cold War.

Having identified the sources of information available to proto-rebels, I now turn to operationalizing similarity of proto-rebels to states hosting these phenomena in order to test Hypothesis 1. Simmons & Elkins (2004) and Danneman & Ritter (2014) operationalize cultural similarity by measuring whether or not two states share a dominant language or religion. I follow this approach by counting the number of states hosting armed conflicts or revolutionary regimes that share a dominant language or religion with another state. Language and religion data are extracted from Ellingsen (2000), in which nine different religions and 132 languages are coded. This yields two variables: Conflict cultural similarity and Revolutionary cultural similarity. To account for the contrary cases, it is also necessary to include variables that account for dissimilarity. I thus split out two additional variables, Conflict cultural dissimilarity and Revolutionary cultural dissimilarity. All similarity variables are paired with their dissimilarity counterparts in all models in which they appear.

It is possible that similarity operates in dimension other than the cultural. To account for that possibility, I create a measure of political similarity by counting the number of states hosting armed conflicts or revolutionary regimes that share regime type with another state. Regime type is determined according to the six-fold classification scheme provided by Cheibub, Gandhi & Vreeland (2010) (hereafter called the CGV data).⁷ This yields two variables: Conflict political similarity and Revolutionary political similarity. As with the cultural measures, I split out dissimilarity variants of these two variables. Each of these measures is lagged by one year.

Proximity

In order to test the expectations pertaining to proximity (Hypotheses 2 and 3), I disaggregate all of my variables into two types. I do so by way of the coding of direct contiguity as defined by the Correlates of War Contiguity dataset (Stinnett et al., 2002). Each variable is split into two variants in each state-year: (1) similar states that are contiguous to the unit of observation at time t;⁸ and (2) similar states that are non-contiguous to the unit of observation at time t.⁹ A complete list of the disaggregated variables is available in Table I.

Disaggregating by proximity is useful for a number of purposes. First, I am interested in showing that learning over great geographic distances is possible. The literature in this area typically uses measures that degrade the impact of civil wars upon proto-rebels over distance. It is therefore not surprising that such literature finds little evidence of learning. Yet, if my theory is correct, and proto-rebels are learning from revolutionary regimes on a global basis, then disaggregation is a useful tool to find evidence of such.

This strategy also serves a purpose beyond hypothesis testing. Because conflict tends to occur in clusters (Salehyan, 2009), it is difficult in the extreme to separate out learning mechanisms from those direct mechanisms, like refugee flows, occurring among directly contiguous states. Moreover, all diffusion mechanisms overlap to some extent (Wood, 2013). Consider that Sudan in the late 1990s was nearly surrounded by states whose governments had been empowered by successful rebellion, each of which could have inspired Sudanese proto-rebels. Yet, these revolutionary states were each consumed by their own civil conflicts. Porous international borders permitted refugee flows into each state’s territories, states sanctioned and supported rebellion against one another’s

Table I.

Logit models of armed conflict onset and militant group emergence, 1968–2001

Variable	Armed conflict onset (1)	Militant group emergence (2)
Neighboring civil war	0.52**	0.58***
	(0.17)	(0.16)
Civil war		0.71***
		(0.18)
Neighborhood democracy	0.19	0.30
	(0.47)	(0.34)
Neighborhood GDP	0.04	0.23
	(0.15)	(0.15)
Democracy	−0.39	0.39*
	(0.24)	(0.16)
GDP(ln)	−0.35**	0.07
	(0.12)	(0.13)
Population(ln)	0.26***	0.41***
	(0.07)	(0.06)
Constant	−2.91*	−9.17***
	(1.22)	(1.14)
χ²	77.17	645.4
Log-likelihood	−649.4	−1028.17
N	4,787	4,787

Coefficients with robust standard errors in parentheses; cubic polynomials and year dummies not reported; significance levels are two-tailed, ***p < 0.001, **p < 0.01, *p < 0.05.

governments, and states attacked and invaded one another in a complex of interlocking wars. Providing evidence of learning over and above the effect of such a conflict zone is the central challenge of this article. In order to provide evidence of learning, it is therefore necessary to show that the non-contiguous versions of my variables are associated with proto-rebel mobilization.

I also execute a series of robustness tests, reported in the Online appendix, in which I replace my independent variables with simple measures of the minimum distance between a given country and the nearest armed conflict or revolutionary regime. These tests add support to my theoretical story, and are discussed in the Online appendix.¹⁰

Control variables

I extract most of my control variables from the Buhaug & Gleditsch (2008) replication dataset, which provides a standard benchmark for studies of civil conflict diffusion. The first is Neighboring civil war, a 0/1 indicator of civil conflict in at least one of a given country’s contiguous neighbors. This variable is used primarily in a set of baseline analyses that establish the fact that international factors are critically important to militant group emergence.

Another variable is Civil War, a 0/1 dummy variable indicating the presence of an organized armed challenge against the government of a state, resulting in at least 25 battle-deaths in a given year (Themńer & Wallensteen, 2015). Because poverty has been shown to be an important correlate of civil conflict, I insert the natural logarithm of GDP per capita at t–1 (GDP per capita (ln)) into the model. This variable is ultimately derived from the Gleditsch (2002) Expanded Trade and GDP dataset. A neighborhood average of this variable is also included, Neighborhood GDP per capita. The natural logarithm of population is also included, derived from the Correlates of War National Material Capability (NMC) data (Singer, 1993).

Several additional variables control for those institutional features associated with militant mobilization. These are Democracy and Neighborhood democracy. Democracy is a 0/1 indicator, coded 1 if a state is of a democratic type. Neighborhood democracy, per Maves & Braithwaite (2013), is that proportion of states within 3,000 km of any given state that are democratic. These variables are constructed using the CGV data. This is appropriate because the traditional indicator of democracy, the Polity scale (Marshall, Jaggers & Gurr, 2004), includes features of political violence within its measure, thus increasing the danger of endogeneity in the model (Vreeland, 2008).

Figure 1 demonstrates an increasing number of militant groups emerging over time. This may be an actual empirical phenomenon or, more likely, it is due to bias in the underlying data sources (Drakos & Gofas, 2006). To control for this bias, I add year dummies into the model. Finally, I control for temporal dependence by inserting a count of the number of years between incidences of Militant group emergence, as well as the squared and cubed variants of this term (Carter & Signorino, 2010). Descriptive statistics for each of these variables is included in the Online appendix.

Analysis

In order to analyze the data, I execute logit regressions with robust standard errors clustered by country. This is an appropriate model given the binary nature of the dependent variable and the time-series–cross-sectional structure of the data. In order to find support for my theory, I undertake a multistep strategy. First, I demonstrate that neighboring civil wars are significantly related to both Armed conflict onset and Militant group emergence (Table I). The primary independent variable of interest is Neighborhood civil war. Model 1 is a replication from the Buhaug & Gleditsch (2008) study, with years prior to 1968 removed in order to match those samples in the remainder of the article. As Model 1 demonstrates, this variable is significantly related to Armed conflict onset, as we would expect from the literature. Model 2 replaces the dependent variable with Militant group emergence and uses time dummies and cubic polynomials. The model shows that Militant group formation is also significantly related to Neighborhood civil war.

This finding demonstrates that empirical leverage is to be gained by changing the analytic focus of civil war onset to one that is more theoretically connected to collective action. Moreover, this finding is an innovative and significant contribution to the literature. Existing studies of civil conflict utilize an indicator of armed conflict onset in the dependent variable, thereby aggregating the various prewar action–reaction processes that occur between proto-rebels and governments.

Having established that mobilization and the onset of war are both significantly related to neighboring conflict, and that analytic utility is to be gained, I undertake an analysis of militant group emergence as a function of learning from armed conflicts, with learning defined according to cultural and political similarity (Table II, Models 3 and 4). I then show in Table II, Models 5 and 6, the impact of revolutionary regimes upon militant group emergence. Because the variants of the conflict and revolutionary learning variables are significantly related, I cannot insert them into the same model. This makes sense, as a revolutionary government, by definition, emerges in a country that has experienced civil conflict.

Before turning to my analysis, a note on presentation is required. Owing to the fact that many different learning variables have been constructed, I do not report Armed conflict similarity and Revolutionary similarity, or their dissimilarity-based variants, in separate rows of Table II. Rather, I label these variables according to that contextual feature whose similarity is being measured, that is, Cultural similarity or Regime similarity. It must be understood, however, that these variables are measuring separate concepts in Models 3–4 and Models 5–6.

In Model 3, the impact of Cultural conflict similarity (contiguity) is statistically distinguishable from zero. This should come as no great surprise – armed conflict within a state is highly likely to increase the odds of mobilization

Table II.

Logit models of learning and military group emergence, 1968–2001

Variable	Armed conflict		Revolutionary regimes
Variable	(3)	(4)	(5)	(6)
Cultural similarity (contiguous)	0.50*		0.28*
	(0.24)		(0.13)
Cultural similarity (non-contiguous)	0.26		0.15†
	(0.23)		(0.08)
Cultural dissimilarity (contiguous)	0.28		0.11
	(0.27)		(0.12)
Cultural dissimilarity (non-contiguous)	0.24		0.14†
	(0.23)		(0.08)
Political similarity (contiguous)		0.34		−0.01
		(0.24)		(0.16)
Political similarity (non-contiguous)		0.23		0.27**
		(0.22)		(0.10)
Political dissimilarity (contiguous)		0.40^†		0.13*
		(0.24)		(0.07)
Political dissimilarity (non-contiguous)		0.24		0.16*
		(0.21)		(0.06)
Civil war	0.92***	0.89***	0.72***	0.73***
	(0.27)	(0.26)	(0.18)	(0.17)
Neighboring civil war			0.56***	0.57***
			(0.16)	(0.17)
Neighborhood democracy	0.24	0.27	0.27	0.33
	(0.34)	(0.36)	(0.34)	(0.34)
Neighborhood GDP	0.17	0.22	0.24†	0.24
	(0.14)	(0.15)	(0.14)	(0.15)
Democracy	0.41**	0.32	0.40**	0.15
	(0.16)	(0.20)	(0.15)	(0.23)
GDP(ln)	0.05	0.02	0.07	0.07
	(0.14)	(0.14)	(0.13)	(0.13)
Population(ln)	0.43***	0.37***	0.41***	0.41***
	(0.09)	(0.07)	(0.06)	(0.06)
Constant	−12.86**	−12.30**	−10.8***	-10.78
	(4.06)	(3.78)	(1.30)	(1.24)
χ2	634.5	647.7	730.2	741.7
Log-likelihood	−1,024	−1,029	−1,024	−1,021
N	4,787	4,787	4,787	4,787

Coefficients with robust standard errors in parentheses; year dummies not reported; significance levels are two-tailed, ***p < 0.001, **p < 0.01, *p < 0.05, ^† p < 0.1.

in a culturally similar state that is within close proximity. The remaining similarity and dissimilarity variables in Model 3 are not significant. Although I have shown that neighboring conflicts are associated with both conflict onset and militant mobilization (Table I), it can be seen that disaggregating neighboring conflict by Cultural similarity does not add much to the model in terms of explanatory power. This is consistent with the literature. Conflict tends to cluster as a result of its underlying determinants, while diffusing as a result of migration-based mechanisms, like refugee flows, irrespective of the languages spoken among neighbors. The same result holds in Model 4, which analyzes the impact of Political similarity. This supports Hypothesis 2.

It is in Models 5 and 6, however, that evidence of learning is apparent. Model 5 assesses the impact of a state’s cultural similarity to the international system’s revolutionary regimes. Most of the learning-based independent variables are at least marginally significant. This lends mixed support to many of my hypotheses. The significance of Cultural similarity (non-contiguity) supports the contention that proto-rebels learn from international events, even those in non-neighboring states. Interestingly, Cultural dissimilarity (non-contiguity) is

Figure 3.

Militant group emergence as a function of cultural revolutionary similarity (contiguous), 1968–2001

also significant. It seems, then, that proto-rebels may learn from dissimilar cases as well. Recall that I theorized that proto-rebels are more likely to learn from a source that is highly visible, per Hypothesis 3. In other words, cultural similarity does not restrict those sources from which proto-rebels learn, provided that said source is a revolutionary regime. Similar findings are evident in Model 6, concerning learning based on political similarity.

Together, these findings paint an interesting, if nuanced, picture. Empirical leverage can be gained by directly measuring proto-rebel mobilization in the dependent variable, rather than the onset of conflict, and by using a measure of revolutionary success as the independent variable. Moreover, the emergence and survival of a revolutionary state is a radical shock to the international system that can inspire proto-rebels globally, irrespective of cultural or regime similarity. Thus, Hypotheses 1 and 2 receive marginal support, while Hypothesis 3 is well supported.

In order to discern the actual impact of revolutionary regimes upon Militant group emergence, I extract predicted probabilities for the statistically significant learning variables from Models 5 and 6. I set all continuous variables to their means and all dummy variables at their modes and then generate a number of scenarios, which are plotted in Figures 3, 4 and 5. Figure 3 plots a 166% change in the probability of Militant group emergence as Revolutionary cultural similarity (contiguous) varies from 1 to 5. Figure 4 presents the effect of Revolutionary cultural similarity (non-contiguous) as it varies over the same range, and shows a concomitant 100% increase in the probability of Militant group emergence. There is thus a reduction in the absolute probability of Militant group emergence,

Figure 4.

Militant group emergence as a function of cultural revolutionary similarity (non-contiguous), 1968–2001

Figure 5.

Militant group emergence as a function of political revolutionary similarity (non-contiguous), 1968–2001

as well as a reduction in the relative change in probability as distance from the site of a revolutionary regime increases. This supports Hypothesis 2, concerning the impact of geographic proximity. A similar finding is seen in Figure 5, concerning Revolutionary political similarity (non-contiguous) – the variable shows a 75% change over the displayed range. This result solves the puzzle that opened this article – proto-rebels react to events in similar regimes across the world. This was the pattern seen in the European revolutions of 1848 and 1989, the Color Revolutions of the mid-2000s, and the Arab Spring of 2011 (Beissinger, 2007; Kuran, 1991; Weyland, 2010, 2012, 2014).

Finally, the significance of non-contiguous variables, as well as many of the dissimilarity variables, suggests a possible role for ideology in overpowering the attenuating impact of proximity and dissimilarity. In my opening anecdote, Peru and Nepal were connected by proto-rebels adhering to a common ideological frame – Maoism in this case. Certainly the contemporary actions of ISIS and like-minded rebels seem to fit within this framework. I do not address ideological diffusion in this article, but it is an important area for future research.

Conclusion

During the Arab Spring in 2011, mass uprisings in Tunisia were credited with inspiring rebellion in neighboring states, such as Libya and Egypt, and influencing levels of dissent in distant states, such as Syria and Mali. Yet, these early rebellions did not diffuse to a number of proximate and culturally similar states. These patterns of conflict diffusion in the Middle East suggest that rebellion is not solely a function of proximity to civil war, or of the spillover of civil war conditions. Rather, I contend that rebellion is, in part, a function of observation, interpretation, and learning by individuals considering whether to engage in rebellion, individuals that I refer to as proto-rebels. At the same time, I seek to move the literature’s focus away from a sole preoccupation with direct, proximate and material causes of diffusion, and toward indirect forms.

This study has several implications for the broader analysis of conflict. First, victory by rebels and the governments that they subsequently lead have a powerful impact on the decision to rebel. Second, the diffusion of civil conflict is more than simply a matter of purely local determinants or neighborhood effects; rather, it is a phenomenon with global dimensions and can be transmitted via information flows that are not grounded solely in direct, material experience.

The findings in this article offer a wealth of options for future research. Although I find that proto-rebels learn on a global basis from revolutionary regimes, irrespective of culture or regime-type, scholars should explore other informational pathways. Cultural similarity and political similarity are commonly used in the literature, yet these are very blunt measures. One obvious starting point for future research is ideology. Indeed, some ideologies make universalist claims that transcend the commonalities of religion or language. It should be noted, however, that ideology is as blunt as any other measure. One way around this problem might be to study the origins of ideological diffusion. For example, the Latin American urban conflicts of the 1970s are known to have inspired urban militants in Europe, especially the Red Army Faction (Midlarsky, Crenshaw & Yoshida, 1980). Although appeals to ideology might explain this linkage, another more theoretically grounded approach would examine the relative mix of urbanized and agrarian economies within linked states.

Finally, although this article focuses on the origins of the conflict process, it has implications for later stages. Existing literature on the international origins of domestic political conflict studies the onset and duration of civil war. Yet, onset and duration are the result of an escalating action–reaction spiral among rebels and governments. In this article, learned information impacts collective action even before there are organized rebel actors that can enter into such a spiral. Additional research should therefore accomplish two additional goals. First, it should examine the state’s response to proto-rebel learning. Second, it should model those militant groups that actually survive state repression and enter into an escalating spiral, particularly as a function of learning from international sources. Such studies would substantially advance the literature.

Footnotes

Replication data

The dataset, codebook, and do-files for the empirical analysis in this article, along with the Online appendix, can be found at . All analyses were conducted using Stata 12.

Acknowledgements

The author thanks Andrew Enterline, Idean Salehyan, T David Mason, J Michael Greig, and Marijke Breuning for their advice, the University of North Texas for research support, and the editors and reviewers at the Journal of Peace Research. All errors are solely attributable to the author.

Notes

References

Aksoy

Deniz

Carter

David B

(2014) Electoral institutions and the emergence of terrorist groups. British Journal of Political Science 44(1): 181–204.

Aksoy

Deniz

Carter

David B

Wright

Joseph

(2012) Terrorism in dictatorships. Journal of Politics 74(3): 810–826.

Ayres

William R

Saideman

Stephen

(2000) Is separatism as contagious as the common cold or as cancer? Testing international and domestic explanations. Nationalism and Ethnic Politics 6(3): 91–113.

Bakke

Kristin

(2013) Copying and learning from outsiders? Assessing diffusion from transnational insurgents in the Chechen wars. In: Checkel

Jeffrey

(ed.) Transnational Dynamics of Civil War. Cambridge: Cambridge University Press, 31–62.

Bas

Muhammet A

(2012) Democratic inefficiency? Regime type and suboptimal choices in international politics. Journal of Conflict Resolution 56(5): 799–824.

Beissinger

Mark R

(2002) Nationalist Mobilization and the Collapse of the Soviet State. Cambridge: Cambridge University Press.

Beissinger

Mark R

(2007) Structure and example in modular political phenomena: The diffusion of Bulldozer/Rose/Orange/Tulip revolutions. Perspectives on Politics 5(2): 259–276.

Black

Nathan

(2013) When have violent civil conflicts spread? Introducing a dataset of substate conflict contagion. Journal of Peace Research 50(6): 751–759.

Blomberg

S Brock

Engel

Rozlyn C

Sawyer

Reid

(2009) On the duration and sustainability of transnational terrorist organizations. Journal of Conflict Resolution 54(2): 303–330.

10.

Blomberg

S Brock

Gaibulloev

Khusrav

Sandler

Todd

(2011) Terrorist group survival: Ideology, tactics, and base of operations. Public Choice 149(3): 441–463.

11.

Boutton

Andrew

(2014) US foreign aid, interstate rivalry, and incentives for counter-terrorism cooperation. Journal of Peace Research 51(6): 741–754.

12.

Braithwaite

Alex

(2010) Resisting infection: How state capacity conditions conflict contagion. Journal of Peace Research 47(3): 311–319.

13.

Braithwaite

Alex

Quan

(2007) Transnational terrorism hot spots: Identification and impact evaluation. Conflict Management and Peace Science 24(4): 281–296.

14.

Brandt

Patrick T

Sandler

Todd

(2010) What do transnational terrorists target? Has it changed? Are we safer? Journal of Conflict Resolution 54(2): 214–236.

15.

Bueno de Mesquita

Ethan

(2013) Rebel tactics. Journal of Political Economy 121(2): 323–357.

16.

Buhaug

Halvard

Gleditsch

Kristian Skrede

(2008) Contagion or confusion? Why conflicts cluster in space. International Studies Quarterly 52(2): 215–233.

17.

Butler

Christopher

Gates

Scott

(2009) Asymmetry, parity, and (civil) war: Can international theories of power help us understand civil war? International Interactions 35(3): 330–340.

18.

Carter

David B

(2012) A blessing or a curse? State support for terrorist groups. International Organization 66(1): 129–151.

19.

Carter

David B

(2015) When terrorism is evidence of state success: Securing the state against territorial groups. Oxford Economic Papers 67(1): 116–132.

20.

Carter

David B

Signorino

Curtis S

(2010) Back to the future: Modeling time dependence in binary data. Political Analysis 18(3): 271–292.

21.

Cheibub

José Antonio

Gandhi

Jennifer

Vreeland

James Raymond

(2010) Democracy and dictatorship revisited. Public Choice 143(1–2): 67–101.

22.

Colgan

Jeff D

(2012) Measuring revolution. Conflict Management and Peace Science 29(4): 444–467.

23.

Colgan

Jeff D

Weeks

Jessica

(2015) Revolution, personalist dictatorships, and international conflict. International Organization 69(1): 163–194.

24.

Cronin

Audrey Kurth

(2009) How Terrorism Ends: Understanding the Decline and Demise of Terrorist Campaigns. Princeton, NJ: Princeton University Press.

25.

Danneman

Nathan

Ritter

Emily Hencken

(2014) Contagious rebellion and preemptive repression. Journal of Conflict Resolution 58(2): 254–279.

26.

Drakos

Konstantinos

Gofas

Andreas

(2006) The devil you know but are afraid to face: Underreporting bias and its distorting effects on the study of terrorism. Journal of Conflict Resolution 50(5): 714–735.

27.

Dugan

Laura

(2012) The making of the global terrorism database and its applicability to studying the life cycles of terrorist organizations. In: Gadd

David

Karstedt

Susanne

Messner

Steven F

(eds) The Sage Handbook of Criminological Research. Los Angeles, CA: Sage, 175–198.

28.

Dunn

Susan

(2000) Sister Revolutions: French Lightning, American Light. New York: Faber & Faber.

29.

Elkins

Zachary

Simmons

Beth

(2005) On waves, clusters, and diffusion: A conceptual framework. Annals of the American Academy of Political and Social Science 598(1): 33–51.

30.

Ellingsen

Tanja

(2000) Colorful community or ethnic witches’ brew? Multiethnicity and domestic conflict during and after the Cold War. Journal of Conflict Resolution 44(2): 228–249.

31.

Enders

Walter

Sandler

Todd

(2006) Distribution of transnational terrorism among countries by income class and geography after 9/11. International Studies Quarterly 50(2): 367–393.

32.

Fearon

James D

(1995) Rationalist explanations for war. International Organization 49(3): 379–414.

33.

Findley

Michael

Young

Joseph K

(2012) Terrorism and civil war: A spatial and temporal approach to a conceptual problem. Perspectives on Politics 10(2): 285–305.

34.

Findley

Michael G

Piazza

James A

Young

Joseph K

(2012) Games rivals play: Terrorism in international rivalries. Journal of Politics 74(1): 235–248.

35.

Forsberg

Erika

(2013) Do ethnic dominoes fall? Evaluating domino effects of granting territorial concessions to separatist groups. International Studies Quarterly 57(2): 329–340.

36.

Forsberg

Erika

(2014) Diffusion in the study of civil wars: A cautionary tale. International Studies Review 16(2): 188–198.

37.

Gaibulloev

Khusrav

Sandler

Todd

(2013) Determinants of the demise of terrorist organizations. Southern Economic Journal 79(4): 774–792.

38.

Gaibulloev

Khusrav

Sandler

Todd

(2014) An empirical analysis of alternative ways that terrorist groups end. Public Choice 160(1–2): 25–44.

39.

Geddes

Barbara

Wright

Joseph

Frantz

Erica

(2014) Autocratic Breakdown and Regime Transitions: A new data set. Perspectives on Politics 12(2): 313–331.

40.

Gilardi

Fabrizio

(2012) Transnational diffusion: Norms, ideas, and policies. In: Carlsnaes

Walter

Risse

Thomas

Simmons

Beth A

(eds) Handbook of International Relations, 2nd edition. Los Angeles, CA: Sage, 453–477.

41.

Gleditsch

Kristian Skrede

(2002) Expanded trade and GDP Data. Journal of Conflict Resolution 46(5): 712–724.

42.

Gleditsch

Kristian Skrede

(2007) Transnational dimensions of civil war. Journal of Peace Research 44(3): 293–309.

43.

Gurr

Ted Robert

(1970) Why Men Rebel. Princeton, NJ: Princeton University Press.

44.

Haas

Mark L

(2005) The Ideological Origins of Great Power Politics, 1789–1989. Ithaca, NY: Cornell University Press.

45.

Hegghammer

Thomas

(2013) Should I stay or should I go? Explaining variation in Western jihadists’ choice between domestic and foreign fighting. American Political Science Review 107(1): 1–15.

46.

Horowitz

Michael C

(2010) Nonstate actors and the diffusion of innovations: The case of suicide terrorism. International Organization 64(1): 33–64.

47.

Jones

Seth G

Libicki

Martin C

(2008) How Terrorist Groups End: Lessons for Countering al Qa’ida. Santa Monica, CA: Rand.

48.

Kalyvas

Stathis N

Balcells

Laia

(2010) International system and technologies of rebellion: How the end of the Cold War shaped internal conflict. American Political Science Review 104(3): 415–429.

49.

Kathman

Jacob D

(2010) Civil war contagion and neighboring interventions. International Studies Quarterly 54(4): 989–1012.

50.

Kreutz

Joakim

(2010) How and when armed conflicts end: Introducing the UCDP conflict termination dataset. Journal of Peace Research 47(2): 243–250.

51.

Krieger

Tim

Meierrieks

Daniel

(2011) What causes terrorism? Public Choice 147(1–2): 3–27.

52.

Kuran

Timur

(1991) Now out of never: The element of surprise in the East European revolution of 1989. World Politics 44(1): 7–48.

53.

Kuran

Timur

(1998) Ethnic dissimilation and its international diffusion. In: Lake

David A

Rothchild

Donald

(eds) The International Spread of Ethnic Conflict: Fear, Diffusion, and Escalation. Princeton, NJ: Princeton University Press, 35–60.

54.

LaFree

Gary

Dugan

Laura

(2007) Introducing the Global Terrorism Database. Terrorism and Political Violence 19(2): 181–204.

55.

Lake

David A

Rothchild

Donald S

(1998) The International Spread of Ethnic Conflict: Fear, Diffusion, and Escalation. Princeton, NJ: Princeton University Press.

56.

Levy

Jack S

(1994) Learning and foreign policy: Sweeping a conceptual minefield. International Organization 48(2): 279–312.

57.

Lichbach

Mark Irving

(1995) The Rebel’s Dilemma. Ann Arbor, MI: University of Michigan Press.

58.

Marks

Thomas A

Palmer

David Scott

(2005) Radical Maoist insurgents and terrorist tactics: Comparing Peru and Nepal. Low Intensity Conflict & Law Enforcement 13(2): 91–116.

59.

Marshall

Monty G

Jaggers

Keith

Gurr

Ted Robert

(2004) Polity IV project. Center for International Development and Conflict Management, University of Maryland (http://www.systemicpeace.org/polity/polity4.htm).

60.

Mason

T David

Fett

Patrick J

(1996) How civil wars end: A rational choice approach. Journal of Conflict Resolution 40(4): 546–568.

61.

Mason

T David

Krane

Dale A

(1989) The political economy of death squads: Toward a theory of the impact of state-sanctioned terror. International Studies Quarterly 33(2): 175–198.

62.

Maves

Jessica

Braithwaite

Alex

(2013) Autocratic institutions and civil conflict contagion. Journal of Politics 75(2): 478–490.

63.

McSherry

J Patrice

(2005) Predatory States: Operation Condor and Covert War in Latin America. New York: Rowman & Littlefield.

64.

Midlarsky

Manus I

Crenshaw

Martha

Yoshida

Fumihiko

(1980) Why violence spreads: The contagion of international terrorism. International Studies Quarterly 24(2): 262–298.

65.

Neumayer

Eric

Plümper

Thomas

(2009) International terrorism and the clash of civilizations. British Journal of Political Science 39(4): 711–734.

66.

Neumayer

Eric

Plümper

Thompas

(2010) Galton’s problem and contagion in international terrorism along civilizational lines. Conflict Management and Peace Science 27(4): 308–325.

67.

Piazza

James A

(2008) Incubators of terror: Do failed and failing states promote transnational terrorism? International Studies Quarterly 52(3): 469–488.

68.

Rogers

Everett M

(2003) Diffusion of Innovations, 5th edition. New York: Free.

69.

Saideman

Stephen M

(2012) When conflict spreads: Arab spring and the limits of diffusion. International Interactions 38(5): 713–722.

70.

Salehyan

Idean

(2009) Rebels Without Borders: Transnational Insurgencies in World Politics. Ithaca, NY: Cornell University Press.

71.

Salehyan

Idean

Gleditsch

Kristian Skrede

(2006) Refugees and the spread of civil war. International Organization 60(2): 335–366.

72.

Sánchez-Cuenca

Ignacio

Calle

Luis De la

(2009) Domestic terrorism: The hidden side of political violence. Annual Review of Political Science 12: 31–49.

73.

Signorino

Curtis S

(2003) Structure and uncertainty in discrete choice models. Political Analysis 11(4): 316–344.

74.

Simmons

Beth A

Elkins

Zachary

(2004) The globalization of liberalization: Policy diffusion in the international political economy. American Political Science Review 98(1): 171–189.

75.

Singer, J David (1993) Reconstructing the Correlates of War dataset on Material Capabilities of States, 1816–1985. International Interactions 14(2): 115–132.

76.

Stinnett

Douglas M

Tir

Jaroslav

Diehl

Paul F

Schafer

Philip

Gochman

Charles

(2002) The Correlates of War (COW) project direct contiguity data, version 3.0. Conflict Management and Peace Science 19(2): 59–67.

77.

Strang

David

(1991) Adding social structure to diffusion models: An event history framework. Sociological Methods & Research 19(3): 324–353.

78.

Themnér

Lotta

Wallensteen

Peter

(2015) Armed conflict, 1946–2014. Journal of Peace Research 52(4): 536–550.

79.

Vreeland

James Raymond

(2008) The effect of political regime on civil war: Unpacking anocracy. Journal of Conflict Resolution 52(3): 401–425.

80.

Weyland

Kurt

(2009) The diffusion of revolution: ‘1848’ in Europe and Latin America. International Organization 63(3): 391–423.

81.

Weyland

Kurt

(2010) The diffusion of regime contention in European democratization, 1830–1940. Comparative Political Studies 43(8–9): 1148–1176.

82.

Weyland

Kurt

(2012) The Arab Spring: Why the surprising similarities with the revolutionary wave of 1848? Perspectives on Politics 10(4): 917–934.

83.

Weyland

Kurt

(2014) Making Waves: Democratic Contention in Europe and Latin America Since the Revolutions of 1848. Cambridge: Cambridge University Press.

84.

Wood

Elisabeth Jean

(2013) Transnational dynamics of civil war: Where do we go from here? In: Checkel

Jeffrey

(ed.) Transnational Dynamics of Civil War. Cambridge: Cambridge University Press, 231–258.

85.

Young

Joseph K

Dugan

Laura

(2011) Veto players and terror. Journal of Peace Research 48(1): 19–33.

86.

Young

Joseph K

Dugan

Laura

(2014) Survival of the fittest: Why terrorist groups endure. Perspectives on Terrorism 8(2) (http://terrorismanalysts.com/pt/index.php/pot/article/view/334/html).