Master Thesis:
The dark money of insurgency
Exploring criminal funding strategies of rebel groups
and the role of relative & state power
Submitted by Andrea Fabio Denz s2073366
MSc Political Science & Public Administration (Research) Word Count (excl. appendix): 9,474
Supervisor: Dr. Matthew di Giuseppe Second Reader: Dr. Roos van der Haer
Conflict scholarship has shown that insurgent group funding influences civil war in a variety of ways, but less research has looked systematically at why rebel groups get involved in particular financial strategies. I explore rebel-government dynamics as predictors of participation in a variety of crimes. Accordingly, I hypothesize that stronger rebel groups and lower levels of state power should lead to institutional types of crime being more attractive–i.e. drug crimes, smuggling and extortion. I conduct a logistic regression analysis to explore participation in five different crimes for a total of 112 insurgent groups. While the findings show some support for the importance of environmental factors related to state power such as rugged terrain and distance to conflict, less to no support is found for military capacity itself and infrastructure. In addition, I find evidence that stronger rebel groups are less likely to participate in episodic crimes such as kidnapping for ransom. Overall, the results suggest that contextual variables are an important part of predicting insurgents’ participation in crime, but more detailed, regionally-specific data would improve further analysis.
At some point, every insurgency needs to face the inevitable question of how to finance its fight against the government. The end of the Cold War brought along a sharp decrease of foreign state support for many rebel groups (Cornell & Jonsson 2014: 1, Makarenko 2004: 130). As a result, it became an essential task for insurgents all around the world to search for alternative income streams. Many groups looked for new opportunities in the so-called illicit economy– whether it was kidnapping tourists or trafficking narcotics. Frequently, these new sources of wealth turned out to be extremely valuable. When the Islamic State (IS) rapidly gained a foothold in Syria, former U.S. Secretary of Defense Chuck Hagel stated that the group is “as sophisticated and well-funded as any organization we’ve seen” (NPR 2014). It appeared that the IS was much less reliant on wealthy donors than previous terrorist organizations, because it managed to build up an extensive financial framework of its own–including strategies like extorting the local population or smuggling oil (Washington Post 2015).
Examples like this one make it easy to see how the topic of insurgent funding continues to be relevant, because it is tightly connected to some of the most violent conflicts of our time. The basic notion that the manner by which rebels acquire wealth can influence the course of civil conflict is widespread. This thesis, on the other hand, looks more closely at factors that can explain why non-state actors get involved in criminal funding strategies in the first place. In particular, I propose that the conflict itself and the dynamics surrounding it are likely candidates to feed back into financing decisions of rebels. I expect that stronger rebel groups, respectively, low levels of state power would boost participation in institutional crimes, due to an increase in funding opportunities and higher financial expenditures. Indeed, the findings suggest that the power dynamics between rebels and the government influence involvement in crime. While the results are mixed with regards to theory and mechanism, they do corroborate the notion that the context of the conflict is a crucial part of understanding rebel finances. This, in turn, is important to make sense of existing scholarship that often views insurgent funding as a given.
Existing research: the consequences of rebel finances
A prominent perspective in conflict studies has been that rebel finances have the power to shape civil war in various ways. Famously, Collier & Hoeffler (1998) put forward the contention that, first and foremost, financial feasibility–and not grievances–increases the risk of civil war. In this view, rebels are seen as quasi-criminal actors motivated mainly by the opportunity to enrich themselves in conflict. This greed perspective of how conflicts start has been controversially discussed, and also qualified and refuted by some (e.g. Ross 2004, Brunnschweiler & Bulte 2009). Nevertheless, empirical research has demonstrated the importance of rebel funding, for instance, by showing that conflicts last longer when rebels can profit from certain financial strategies (Ross 2004). In Liberia and also in the Democratic Republic of Congo, peace agreements failed or were not adhered to because economic incentives to exploit the enormous resource wealth prevailed (Ross 2004: 53). In a recent contribution, Conrad et al. (2019) find evidence that insurgent groups that smuggle natural resources are embroiled in significantly longer conflicts than those who extort local populations. They theorize that smuggling in particular gives them the necessary flexibility and mobility to resist government repression.
Figure 1 - Classic perspective of insurgent group funding
Perhaps most prominently, a lot of evidence has accrued that some economic endowments can alter the internal dynamics of rebel groups by creating new incentives for war (Cornell 2007, Weinstein 2006). In his seminal contribution Inside Rebellion, Weinstein (2006) shows that rebel groups that rely on substantial endowments that are independent of the civilian population tend to apply more indiscriminate violence. The underlying logic is that these financial rewards attract more opportunistic and short-term oriented individuals to join a rebel group. For instance, groups like the Resistência Nacional Moçambicana (Renamo) substantially benefitted from international financial support, which, according to Weinstein (2006), can partly explain their responsibility for a large part of civilian killings and destruction of property in the Mozambican Civil War (212-213). Building on Weinstein’s (2006) insights, Fortna et al. (2018) found a similar dynamic regarding the decision to use terrorism. When the source of wealth is less intertwined with the general population–e.g. in the case of gemstone looting or drug trafficking–rebel groups have a lower risk of alienating their civilian base with violent strategies such as terrorism. In Uganda, for example, the Lord’s Resistance Army (LRA) began to engage in deadly, mass terror attacks after it lost much of its civilian support–including financial support–in the beginning of the 1990s (Fortna el al. 2018: 786).
Shifting perspective: determinants of rebel funding & the role of conflict dynamics
This classic perspective in scholarship has shown that how an insurgent group acquires wealth appears to have serious consequences on many levels. There is, however, an equally important question that remains to be answered: Are there factors that drive insurgent groups into particular financial strategies? One could consider natural resource deposits a sort of ‘given’ financial blessing, but many other sources of income for rebels probably cannot. To the contrary, the decision to engage in many funding strategies is likely itself influenced by other factors. This seems to be especially true for illicit funding, where opportunities to become involved in crime are often–in principle–abundant, but many groups do not seize them.
Figure 2 - Determinants of insurgent funding
The empirical picture (e.g. Asal & Rethemeyer 2015) shows that involvement in criminal funding strategies often fluctuates even throughout the course of conflict. But if many revenue streams cannot appropriately be characterized as exogenous endowments, this changes the way we interpret some of the previous studies. This project seeks to add to the filling of this gap in conflict scholarship. Thanks to newly available data, researchers have already begun to systematically answer the question as to why insurgent groups might exploit illicit markets. In a large-N analysis, Asal et al. (2019) find that a variety of criminal funding strategies are significantly correlated with control of territory. Furthermore, involvement in institutional crime (specifically drug crimes and extortion) appears to become more likely as a group ages, whereas involvement in episodic crime is rarer in cases where groups are involved in social service provision.
While most of the factors analyzed by Asal et al. (2019) concern group characteristics and social factors, it remains less clear if and to what degree factors related to the conflict itself might shape the choice of certain funding strategies. It is here, in particular, where this thesis ties in with existing scholarship. What appears obvious is that insurgent groups cannot make decisions in a political vacuum, because they are usually entangled in a violent clash with the government and–sometimes–other rebel contenders. Consider the example of drug trafficking. Anderson & Worsnop (2019) point out that “gaining control of a domestic drug industry is not easy: it takes significant investment, effort and time.” (100). Nevertheless, drugs are often treated in the literature as a type of natural resource. The authors argue that involvement in the narcotics
business is not merely a question of the a priori existence of drug crops. Rather, it should be considered a conscious funding strategy that rebel groups decide to engage with if they, for instance, are in particular need of resources to fight the government.
Hence, it is plausible that the environment and characteristics of the conflict would influence a group’s decision to invest in particular strategies. Being in a military confrontation with government forces is a very costly undertaking, and Wennmann (2011) notes that “the cost to maintain armed conflict is disproportionately higher – and less predictable – than the cost to start a conflict” (341). Apart from variable military costs, insurgent groups may also have different opportunities to participate in crime depending on their standing in the conflict. For instance, a rebel group’s choices might become limited if counterinsurgency is very effective, such as when the Tamil Tigers’ maritime smuggling network was interdicted successfully during the Sri Lankan civil war (Smith 2011). Thus, particularly the nature of an insurgent group’s relationship with the government should have implications for its methods of financing. This feedback effect of civil war dynamics on rebel funding itself (see Figure 3) could be a missing component that helps to complete the picture presented by existing research.
Figure 3 – Effect of civil war dynamics on funding?
Relative & state power and involvement in criminal funding strategies
The main proposition of this thesis is therefore that an insurgent group’s participation in crime might in part be a consequence of these rebel-government dynamics. Specifically, to what degree is the military power balance skewed towards the government or the rebel group? How capable is a government to project its power? And what does this imply for engagement in different types of crime?
A good starting point is the model of criminologist R. T. Naylor (1993). He outlines how an ideal-typical insurgent group uses the black market to finance its operations, going through three different stages in its development. At first, an insurgent group often has to rely on hit-and-run operations–or what he calls blue-collar crime. As Naylor (1993) argues, when “a guerrilla group exists in a pronounced stage of geographic and political insecurity in the face of determined opposition from the state […], it might finance itself in much the same way as a blue-collar criminal gang” (20). These financial strategies require relatively limited expenditure
and can include theft–e.g. bank robberies–, ransom kidnapping, maritime cargo hijacking, etc. (Naylor 1993: 24-25). If a group reaches a more advanced stage, there is often a shift from predatory means of funding towards parasitical strategies–i.e. what might be called white-collar
crime. Such strategies typically include embezzlement, extortion of the civilian population,
‘revolutionary taxation’ or resource smuggling (Naylor 1993: 28-30). Importantly, these result in more stable streams of income for the group. Finally, in a third stage of more extensive rebel control, illicit funding becomes symbiotic and merges with the parallel economy established by the group.
Depending on its relative strength and control, an insurgent group might thus focus on different types of crime. A case study of the Revolutionary Armed Forces of Colombia (FARC) by Hough (2011) presents further empirical evidence for this notion. According to his analysis, the FARC’s turn towards typical ‘blue-collar’ crime (e.g. kidnapping) at the end of the 1990s was likely a necessary reaction to the increasing militarization of the Colombian government. Especially relevant, according to Hough (2011), was the influx of military aid from the U.S., which substantially ramped up the capacity of the Colombian armed forces and the number of military bases. It was notably during this period, when the FARC began to look for more efficient, short-term strategies such as “the direct pillage of local banks and stores and the kidnapping of selected individuals who could provide lucrative ransom in exchange for their release” (Hough 2011: 400). Abuza (2010) illustrates a similar dynamic in the case of the Abu Sayyaf Group (ASG) in the Philippines. He notes how the ASG “re-degenerated back into a kidnapping for ransom gang” (Abuza 2010: 13) after the Philippine military succeeded in weakening and splintering the group. In contrast, McDougall (2009) shows that areas of weak state presence in Colombia are often the ones where rebel groups are heavily involved in the drug trade, suggesting a favorable environment for institutional types of crime. Similarly, a lack of state presence is associated with the rise of narcotics smuggling in other conflicts such as during the civil war in 1990s Tajikistan (Dadmehr 2003: 245).
In terrorism scholarship, Picarelli & Shelley (2007) likewise note that particular types of crime require higher or lower capabilities and also have different entry costs. Like Naylor (1993), the authors expect terrorist groups to initially gravitate towards strategies that do not require a great deal of expert skills (Picarelli & Shelley 2007: 45), such as kidnapping or armed robberies. Conversely, more ‘market-oriented’ strategies are typically considered harder to implement. Importantly, strategies like drug trafficking or smuggling very often require support from existing criminal organizations (Picarelli & Shelley 2007: 45). These crime-terror connections, in turn, “are more likely to occur in areas of the world where the state has the least presence and means of control – that is, areas with large shadow economies and regional conflicts” (Shelley & Picarelli 2005: 57). In fact, Asal et al. (2015) discover that one contextual
factor in particular is negatively correlated with terrorist groups’ participation in drug crimes, namely the coercive power of the state.
Episodic vs. institutional crime
Based on these theoretical and empirical insights, I hypothesize that the relative strength of conflict actors and, relatedly, the level of state power should influence which criminal strategies rebel groups adapt. To study this proposition, I follow the categorization between episodic and
institutional crime proposed by Asal et al. (2019). Analogous to what Naylor (1993) terms
blue-collar crime, strategies such as robbery or ransom kidnapping can be considered episodic crimes. They are characteristically non-systematic, usually do not require a great deal of expertise and are relatively low-cost. On the other hand, institutional crime–e.g. drug cultivation and trade–is typically costlier and requires more investment on behalf of the insurgent group. Importantly, though, these strategies also provide a more stable revenue stream. While some forms of institutional crime do not require particularly advanced skills (e.g. extortion), they are nonetheless systematic in nature.
I expect that participation in institutional crime in particular should become more likely the better an insurgent group is able to confront the power of the state. From the literature discussed, two plausible, causal links can be derived. First, comparatively stronger rebel groups can be expected to need more systematic and institutional funding arrangements to finance their fight, because higher capacities usually imply higher costs. Secondly, they should have greater
opportunities to participate in these types of crime, because they are under less pressure from
the state. Conversely, if a government is capable of maintaining a high level of state power, one might expect that a rebel group has to resort to other forms of more episodic crime. The causal mechanism I propose is depicted in Figure 4 below.
Figure 4 – Proposed causal mechanism
While stronger insurgent groups should have incentives to pursue more institutional crime, the question remains as to what the implications for episodic crime are. On the one hand, one could expect a kind of substitution effect, i.e. groups would abandon episodic crime such as kidnapping or robbery if they successfully established more institutionalized funding schemes– which seems to be what Naylor (1993) implies in his model. On the other hand, it also seems plausible that stronger rebel groups simply maintain all types of revenue-creating crime as long
as they are a) able to do so and b) as long as these strategies do not interfere with the provision of social services (see Asal et al. 2019). I argue that both scenarios seem likely, which is why I focus my hypotheses on what factors make insurgents choose institutional crimes instead.
State power
Intrastate conflict is almost by definition an asymmetric clash between a strong government and a weaker insurgent group. To assess what role power plays in a conflict, it is essential to analyze the power of the adversarial state. In order to do so, I follow Mann’s (1984) conceptualization of state infrastructural power. In distinction to despotic power–i.e. the range of actions state elites are empowered to make (Mann 1984: 189)–, Mann (1984) describes infrastructural power as “the capacity of the state to actually penetrate civil society, and to implement logistically political decisions throughout the realm“ (189). In other words, I am interested in what is often termed state capacity, respectively, state reach (Soifer & vom Hau 2008: 230). In a conflict situation, the most relevant indicator of state power is likely going to be its military capacity, as well as existing transport and communication infrastructure as means of transmitting and projecting this power (Mann 1984: 196-197, Mann 2008: 358). In the following section I will outline what I consider to be appropriate indicators of the power balance in intrastate conflicts, as well as geographical and infrastructural factors relevant for the projection of power.
Theoretical expectations
Military capacity & relative rebel strength
A central pillar of (state) power is the military capacity of a government. One would expect that the level of military capacity influences the ability of insurgent groups to maintain their operations–which include any such related to the funding of the organization. In particular,
institutionalized structures of funding should be more prone to disruption if groups face a
government that is highly capable. The smuggling of goods, for instance, often requires a substantial amount of expertise as well as complex support structures that existing crime groups have to provide (Picarelly & Shelley 2007: 45). Similarly, drug cultivation is relatively labor-intensive and usually depends upon a fairly stable environment (Lujala 2009: 54-55). Criminal strategies that involve a number of supply chains–e.g. drug trafficking (see e.g. Kenney 2007)– or are more physically exposed–e.g. extortion of local agriculture–should therefore represent a more attractive target for a more powerful government than typical ‘blue-collar’ crime. All else equal, a rebel actor facing a stronger military threat by the government should therefore be less capable of sustaining institutional revenue streams.
Obviously, the military capacity of a government only tells one part of the story. The rebel group’s own ability to keep government forces at bay is a mitigating factor of state power. In other words, a strong government might have to face exceptionally capable rebel actors, which would decrease its capacity. To account for this fact, I analyse the military power balance with the concept from conflict scholarship called relative rebel strength, i.e. an insurgent group’s military capacity in relationship to the one of the government. Applying this concept, Holtermann (2016) shows that Maoist insurgency groups in the Nepalese civil war would rely on strategies such as raiding police stations while their relative capacity was still limited and only later–with increased military power–would, for instance, systematically extort the local population. In the same manner, my expectation is that relatively stronger rebel groups should be more capable of institutional crime.
Hypothesis 1: The higher an insurgent group’s relative military power, the higher the chance it relies on institutional crime as a funding strategy.
Geographic factors & infrastructure
Aside from raw military power, certain geographic factors can make it more difficult for governments to implement state power in a conflict area (Soifer 2008: 238). Such characteristics can be considered additional, environmental factors that increase or decrease state power–or the relative capacity of rebel groups–but are not merely captured by accounting for regular military strength. Buhaug et al. (2009) show that such geographic factors seem to interact with the capacity of rebel groups to influence, in that case, the length of conflict.
One such factor is the topography of the areas in which rebel groups operate in. Rough, mountainous terrain is “ideal for guerilla warfare and difficult for a government army to control” (Buhaug & Gates 2002: 422). It is also thought to give insurgent groups a defensive advantage in case the power imbalance is already severe (Buhaug et al. 2009: 547, Holtermann 2016). Fearon & Laitin (2003) find that such terrain is associated with significantly higher rates of civil war, arguing that this is likely the consequence of being a favorable environment for rebellion. When comparing different conflicts all around the world, groups that can rely on a significant topographic advantage should be able to implement institutional funding strategies more easily, mainly because of lower state capacity in this area. For some forms of institutional crime such as drug cultivation, remote mountainous terrain moreover presents ideal conditions, as illustrated by the role of opium poppy in conflict-ridden areas such as Myanmar (Tian et al. 2011: 279). All else equal, my theoretical expectation is therefore that rugged terrain increases the likelihood that a rebel group decides to get involved in institutional crime. Thus, I propose:
Hypothesis 2a: Rugged terrain in a conflict area increases the chance that a rebel group relies on institutional crime as a funding strategy.
Another factor that might be relevant is the distance from the location of central state power to the conflict area. Similar to rough terrain, it makes sense that rebel groups that are located further away from the locus of state control would have a comparative advantage over other groups. For a government, operating in remote areas is expensive, because it needs to face considerable logistical obstacles (Buhaug et al. 2009: 550). Holding raw military strength and terrain equal, armed forces closer to a conflict area should be able to better project state power than armed forces having to deal with insurgencies in remote areas. For instance, Schutte (2015) argues that counterinsurgency becomes less accurate over larger distances, because the applied violence cannot be targeted as well as over a short distance. Indeed, Schutte (2015) finds some evidence for his claims in the form of higher rates of indiscriminate violence and rebel victories in countries where much of the population is concentrated far away from the capital city. I argue that for insurgents, in turn, the difficulty in projecting state power would imply that they are able to invest in a wider variety of funding strategies, because the government is comparatively less capable of interfering with these funding efforts.
To account for distance, existing studies have used the distance of conflict areas from the capital city of a country as an indicator (Buhaug et al. 2009, Raleigh 2010, Clayton 2013). While it is not a perfect proxy measure, it should nevertheless–in combination with rugged terrain–be appropriate to account for the remoteness of a conflict area, because capital cities are often the main government stronghold in terms of military power. Accordingly, I expect:
Hypothesis 2b: The higher the distance between a state’s capital and the conflict area, the higher the chance that a rebel group relies on institutional crime as a funding strategy.
Herbst (2014) points out that, given the geographical reality a country has to handle, the question arises “how leaders confront their geographic and demographic endowments by constructing the infrastructure of power” (111). An indicator of how a state manages its geography is the density of the transport infrastructure. Implementing control–especially military control–requires the state to physically relocate manpower and equipment. Herbst (2014) identifies the low road density in many African countries–which often only marginally increased since the wave of decolonization in the 20th century–as a key reason why African governments have often lacked effective state power. On the other hand, tragic incidents of state control such as the extraordinarily rapid genocide in 1994 Rwanda or the high levels of repression in certain areas of Zimbabwe during the 1980s are considered to be linked to the
comparatively good (inherited) infrastructure in these countries (Herbst 2014: 115-116). Similarly, McDougall (2009) views road density as an appropriate proxy of state power in Colombia. He finds a negative relationship between road density as a measure of state presence in Colombia and the location of rebel groups and drug cultivation. If high road density increases the reach of the state, one could thus expect:
Hypothesis 2c: The lower the road density in a conflict area, the higher the chance that a rebel group relies on institutional crime as a funding strategy.
Finally, what might mitigate the reach of the state is the location of a conflict in a border region. While conflicts in border areas are shown to last longer (Buhaug et al. 2009), it is not obvious why a conflict location close to an international border per se would increase a rebel group’s ability to invest in more institutional criminal strategies, because a government might have a strong presence in such an area–for instance, if the capital city is located in its vicinity. However, border regions should provide insurgent groups with a natural advantage when it comes to one institutional crime in particular, namely smuggling. Asal et al. (2019) argue that especially in border regions “the ability to transport illicit goods across borders may require less infrastructure, experience, or skill than drug crimes or extortion” (10). Thus, while conflicts in border regions might not necessarily deteriorate a government’s power per se, one could expect that it should lower its ability to impede groups from smuggling goods and individuals over international borders. Hence, I expect:
Hypothesis 2d: Conflicts located in a border region increase the chance that a rebel group relies on smuggling as a funding strategy.
Other factors
Evidently, when it comes to participation in crime, relative & state power are not the only explanatory variables that come into question. In my analysis, I control for the factors that have so far been theorized and tested in the analysis of Asal et al. (2019).
Control of territory has been hypothesized to drive participation in crime, because of higher costs and better opportunities to establish criminal funding strategies (e.g. Asal et al. 2015). Indeed, territorial control predicts involvement in all of the criminal strategies analysed by Asal et al. (2019). A notable contextual factor is conflict intensity, which has been hypothesized to increase participation in episodic crime, due to reliance on more easily implemented criminal strategies (Hough 2011). Partially confirming this theory, the incidence of robbery seems to
increase with higher conflict intensity (Asal et al. 2019). Furthermore, alternative funding such as state support might decrease reliance on illicit strategies in general.
In terms of social & organizational features, a previously analyzed factor is the ideology of insurgent groups. It has been noted that certain ideologies such as ethnic or religious ones are strongly associated with crime and certain types of crime (e.g. Paoli & Reuter 2008). Moreover, network connections to other groups were hypothesized to increase the propensity to become involved in criminal alliances (e.g. Shelley & Picarelli 2005) and have been shown to be associated with extortion and kidnapping (Asal et al. 2019). In terms of organizational structure, non-hierarchical leadership could lead to greater collaboration with crime groups and thus to increased participation in crime (e.g. Dishman 2005). Asal et al. (2019) also find evidence that the provision of social services seems to inhibit participation in episodic crimes. This is tied to an assumed negative effect of crimes like robbery and kidnapping on the legitimacy of insurgent groups in the population, as social services have shown to be an important instrument to increase legitimacy (e.g. Flanigan 2008). In addition, some institutional crimes (drug crimes and extortion) are found to be more prevalent among older groups, which is theorized to be a consequence of higher skills necessary to enter into these strategies. Finally, increased size of an organization could necessitate involvement in high-revenue crime such as drug trafficking (Asal et al. 2015).
Research design
In order to analyze participation in illicit funding strategies I make use of the Big Allied and Dangerous (BAAD) 2.0 dataset (Asal & Rethemeyer 2015). The BAAD provides information on 140 violent non-state organizations for the period of 1998-2012. Data on independent variables was available for most, but not all observations. In total, I am able to analyze 112 organizations, resulting in 1’079 single observations (dyad-years). Based on the homebase of insurgent groups, the analysis covers 44 different countries–20 in Africa, 12 in Asia, five in the Middle East, four in Europe, two in South America (Colombia & Peru), as well as Haiti in the Caribbean.
Dependent variables
The BAAD enables to account for participation in five different types of crime: drug trafficking, extortion, robbery, smuggling and kidnapping for ransom. For all insurgent groups in the dataset, dummy variables indicate if a group was involved in a particular strategy in a given year. Involvement in drug trafficking implies that an insurgent group was trafficking narcotics and/or was an original grower and supplier of such products. Extortion denotes the collecting of resources by threating local businesses, communities or individuals with violence. Robbery
was coded if the group was involved in the theft of money, including bank robberies and looting. Smuggling accounts for the smuggling of goods but also human trafficking. Narcotics smuggling, however, is coded separately, as noted above. Finally, the data captures if insurgent groups were involved in kidnapping for ransom. Any other type of kidnapping–e.g. for political goals–is excluded in this measurement.
Independent variables
A variety of independent variables are included by merging the BAAD with other datasets. As a level of observation, dyad-years were constructed based on the conflict dyads in the Uppsala Conflict Data Program (Harbom et al. 2008, Allanson et al. 2017). For every insurgent group, the main government (‘Side A’) from the UCDP dyadic dataset was added. In order to account for the specific regions in which conflict took place, I added up to five sub-national regions per dyad-year. Conflict regions were constructed based on the UCDP Georeferenced Event Dataset (Sundberg & Melander 2013, Croicu & Sundberg 2017). First, I aggregated the number of conflict events per region and dyad in the period of 1998-2012. If there was a clear pattern of conflict activity in specific regions–which was the case for most dyads–up to five conflict regions were added. In a few cases where no data was available or no clear pattern was found– e.g. for the Communist Party of the Philippines (CPP-ML)–no regions were added.
With regards to relative rebel strength, there are existing quantitative and qualitative measures in the literature that came into question. A quantitative measure is the ratio of rebel and government troop strength, with estimates taken from UCDP annual reports (Gent 2011, Clayton 2013). While such a ratio measure is rather straightforward and clearly defined, it has severe limitations. For instance, as to the size of government forces, in most cases only the total national strength of troops is available. Thus, the ratio variable does not give an indication as to how large the force deployed against a specific insurgent group is.
For this reason, I decided to opt for an ordinal indicator of relative rebel strength, as provided in the non-state actors dataset (Cunningham et al. 2013). The non-state actor dataset gives the option of a composite measure of relative rebel strength and three component measures: fighting capacity, mobilization capacity and arms procurement. In the composite measure the vast majority of insurgent groups fall into the two lowest categories (‘much weaker’ and ‘weaker’) of the five-point-scale, but the decision for coding a group as either ‘weak’ or ‘much weaker’ seemed rather unclear. In order to better capture the military power balance in a conflict, I rely on the component variable ‘fighting capacity’ as an indicator of relative rebel strength, (1= low, 2 = intermediate, 3= high), denoting a group’s military capacity relative to the government. Because the number of groups with high relative fighting capacity is very low, these particular observations could have distorted the analysis. For this reason, I constructed a dichotomous measure of relative strength (0=low, 1=intermediate or high) for the final variable. The variable rugged terrain is based on newly available data from Shaver et al. (2019). Their data on terrain ruggedness have two main advantages over existing measurements (e.g.
percentage of mountainous terrain in Fearon & Laitin 2003). First, the data include
observations not only at the national level but also at the primary sub-national level (e.g. state, province, etc.). Secondly, their data are based on precise measures of terrain ruggedness per km2, as opposed to dichotomous measures that code terrain as either mountainous or not. Specifically, “the ruggedness of any given 1 km2 is determined by measuring how the average elevation of that area differs from all those of neighbouring 1 km2 areas. Each such difference is squared so that positive and negative elevation changes contribute equally to the overall ruggedness measure.” (Shaver et al. 2019: 199). By taking the sum of these squared differences and, subsequently, the square root of this sum, a normalized ruggedness measure for every square kilometer is constructed.
Based on this data, I constructed variables for every dyad. If specific regions were identified, the mean of the respective average sub-national measures was calculated. For instance, in the case of the Kachin Independence Army (KIA)-Government of Myanmar-dyad the measures for the two regions Kachin State (337m) and Shan State (289m) resulted in a mean ruggedness measure of 313m. For dyads with no specific conflict regions, the mean national ruggedness measure was added. To construct the final variable, I applied the natural logarithm to reduce the right-skewedness of the data.
In order to construct the variable distance to capital city, I first added the coordinates for the capital cities of all main governments (‘Side A’) to the UCDP Georeferenced Event Dataset1. Capital city coordinates were taken from the United Nations report World Urbanization
Prospects: The 2014 Revision (UN 2015). For every conflict event, the geographical distance
between the event and the respective capital city of the government side was calculated2 by means of the Haversine formula:
where f1, l1 and f2, l2 denote the geographical latitude and longitude in radians of the respective conflict event and the respective capital city. The actual distance in kilometers per conflict event was computed by multiplying the result of the formula with the radius of Earth (6371 km). Consequently, the distances were aggregated at the dyad level in order to compute the average distance between conflict events and the capital city of the government in kilometers in the period of 1998-2012. The final variable is represented by the natural logarithm of this average distance.
The variable road density is based on national-level measures, as unfortunately no sub-national data for enough of the cases was found. Measures were included for the country in which the homebase of the insurgent group is located. The data were recorded in the period 1990-2011 and stem from the Food and Agricultural Organization (FAO) of the United Nations (FAO 2011). The indicator accounts for motorways, highways, as well as main, secondary and regional roads. It measures road length (km) per 100 km2. In the case of the 2012 dyad-years– and other dyad-years where data was missing–, last recorded observations were added instead. For early dyad-years with missing data, first recorded observations were added instead. Ideally, projecting data into the future and past would be avoided, but in the case of road density, the data barely varies over time, which limits the negative consequences. Again, I applied the natural logarithm for the final variable to reduce right-skewedness.
Finally, because existing data that codes border regions was incomplete for the cases analyzed, I account for conflicts in border regions with a dummy variable cross-border activity. The constructed variable takes the value 1 if either the main government, one of the conflict regions or the homebase of the group did not match with one of the other two indicators. As an
1 In the al-Qaida dyad, the main government location was changed from Washington (USA) to Kabul
(Afghanistan)
illustration, the insurgent group All Tripura Tiger Force (ATTF) was coded as 1, because it was recorded to be active in the Indian state Tripura but is itself based in Bangladesh, according to the BAAD.
Control Variables
As outlined earlier, I control for previously explored predictors by including variables from Asal et al.’s (2019) analysis of the BAAD. These are territorial control, size, state funding, battle deaths, network connectivity, ethnic ideology, religious ideology, leftist ideology, hierarchical leadership, social service provision and age. Descriptive statistics that include all of the control variables were added in the appendix.
Method
As the independent variables are binary–participation or no participation–I apply a logistic regression model to test the hypotheses. The panel data requires that time dependence needs to be accounted for. I follow Carter & Signorino (2010) who suggest that, instead of modelling time dependence through cubic splines, it is possible to use the simpler alternative of cubic polynomial approximation, which means including the variables t, t2 and t3 in the regression.
The variable t hereby refers to the time since the last event–in this case the time period since a funding strategy was implemented last. According to Carter & Signorino (2010) cubic polynomial approximation can model a variety of hazard shapes and performs well compared to cubic splines and time dummies (291). In addition, I also account for possible non-independent observations in the data by including robust standard errors clustered by insurgent groups. All models have furthermore been checked for multicollinearity of predictors. Due to limited space, control variables are not displayed in the regression tables, but complete tables are included in the appendix.
Results
Smuggling
Looking first at participation in smuggling, rebel groups with higher relative military power are no more likely to participate in the crime than weaker ones. In line with expectations, however, I find that conflicts in border regions increase the incidence of smuggling. Specifically, cross-border activity increases the predicted probability of participation in smuggling from 18.7% to 34.6%. Higher road density, on the other hand, is associated with a higher and not lower incidence of smuggling, which is counter to what I expected in Hypothesis 2c. At the lower bound of road density (value 0) the probability of smuggling is predicted to be 5%, while at the upper bound (value 5) the predicted value is 36.6% (see Figure 5). Thus, the counter-theoretical
effect is quite strong. Further indicators related to the projection of state power such as distance to capital and ruggedness of terrain show no effect on smuggling.
Extortion
Turning to participation in extortion, the relative power of insurgent groups, again, does not seem to influence involvement in this type of crime. On the other hand, insurgent groups located further away from the state capital are more likely to invest in extortion as a means of funding, which is evidence for Hypothesis 2b–i.e. that weakened state power over long geographical distances increases participation in institutional crime. The effect is fairly strong: An increase by one unit (natural logarithm) in distance to the capital city, raises the odds of extortion by more than 35%. Figure 6 illustrates that the predicted probability of extortion goes up from 33.3% to 68% between the lower and upper bound values for the distance to the capital city (range of values: 3-8).
In turn, the results for road density–as in the case of smuggling–are in the opposite direction compared to expectations. Higher road density is associated with a higher incidence of extortion, contradicting Hypothesis 2c. The effect size is also among the largest in the analysis: over the range of values for road density (0-5) the predicted probability increases from 26.8% to 74.6%. In addition, ruggedness of terrain and cross-border activity are uncorrelated with extortion.
Drug Crimes
Concerning participation in drug crimes, a surprising result is that relatively weaker rebel groups seem to be more likely to participate in drug crimes. The predicted probability of participation in drug crimes for weaker insurgent groups is 22.6%, while it stands at 6% for stronger groups. This finding contradicts Hypothesis 1, where I expected institutional crime to be lower if rebel groups have smaller military capacities. Conversely, though, rugged terrain increases the incidence of drug crimes, as I expected. The effect size is quite large, as Figure 7 shows. The predicted probability of drug crimes increases from 4.1% (value: -3) to 53.6% (value: 2). This result supports Hypothesis 2a, i.e. a higher precedence of institutional crime when the terrain is more mountainous. I discuss further below how the two above-mentioned findings might relate to each other. Moreover, no effect is found for distance to capital, road density and cross-border activity.
Robbery
In terms of theory, I hypothesized that the independent variables would mainly influence involvement in institutional crime. Turning to episodic crimes, relative power indeed does not seem to drive involvement in robberies. Nonetheless, lower road density is correlated with a lower incidence of robbery. As discussed earlier in the theory section, this could speak for a
substitution effect for episodic crime, where lower state power does not only increase
institutional crime but also decreases episodic crime. However, given the previous counter-theoretical results for road density, this interpretation should be treated cautiously. An unexpected finding is that participation in robbery is predicted by higher cross-border activity as well. Analogous to smuggling, cross-border activity leads to odds of being involved in robbery that are roughly three times as big as those of other groups.
Kidnapping
Finally, turning to kidnapping for ransom, the key finding is that relatively more powerful groups are less likely to use kidnapping as a funding strategy. Again, I mainly expected that increased power would lead groups to ‘diversify’ their operations through higher investment in institutional crime. Analogous to the results for robbery, this result would, however, suggest that increased relative power could lead groups to substitute kidnapping with other financial strategies. The associated decrease in odds of kidnapping is about 77% and thus similar to the one for drug crimes. Weaker insurgent groups are predicted to participate in kidnapping with a probability of 11.5%, while stronger groups are predicted to do so with a probability of 3.8% (Figure 8).
Figure 7 – Predicted Probability of Drug Crimes Figure 8 – Predicted Probability of Kidnapping
Discussion & Conclusion
Overall, the results are mixed with regards to my theoretical expectations. The geographic indicators provide some of the strongest evidence for the theory that low state power makes institutional crime more likely. Rugged terrain in the conflict area is associated with a significantly higher incidence of drug crimes, in line with Hypothesis 2a. It thus supports the notion that institutional crime tends to be attractive in areas that are harder to reach for the government. Furthermore, insurgent groups operating at greater distance from the state capital use systematic extortion more often to fund their operations. This finding provides some evidence for Hypothesis 2b, which predicted participation in institutional crime to be more likely if the government has to project its state power over larger distances. Somewhat reassuringly, these are the two independent variables where I could rely on the most precise data– i.e. the UCDP Georeferenced dataset (Sundberg & Melander 2013, Croicu & Sundberg 2017) and regionally specific terrain data (Shaver et al. 2019). Hypothesis 2d likewise finds support, as groups are indeed more likely to partake in smuggling when they are active across international borders, suggesting that border regions give insurgents a particular advantage for this crime. An unanticipated finding is that robbery is also predicted by cross-border activity. A possible explanation might be that the difficulty of law enforcement across international borders leads to a particular advantage for robberies, because it provides them with a convenient refuge.
The proposed theory lacks evidence when it comes to military capacity itself. I expected the relative power of rebels to be positively correlated with institutional crime, which is not the case. Participation in drug crimes is even associated with lower military capacity of rebel groups. What might explain this latter finding? One way to make sense of this result is that the
effect of relative power could be interrelated with terrain (Holtermann 2016). In other words, the advantage of rugged terrain would explain why insurgent groups do in fact need less raw military power to partake in drug-related crimes. These two cross-cutting effects therefore do not necessarily contradict the general theory. In other words, the relative advantage of rough terrain in terms of providing shelter from state power might still outweigh the relative
disadvantage of having lower military power. However, I did not anticipate sufficiently that
such a cross-cutting effect makes a conclusive interpretation quite difficult.
More in accordance with expectations is the finding that weaker groups are also less likely to participate in kidnapping for ransom. This effect appears to be independent from social service provision, which has been shown to reduce participation in episodic crime (Asal et al. 2019). The finding would suggest that there might exist a substitution effect for kidnapping, i.e. comparatively stronger groups have to rely less on episodic crimes financially and therefore drop it, respectively, do not consider it (Naylor 1993). However, there is no corresponding evidence that, in turn, participation in institutional crime increases for stronger groups. What remains possible is that such an effect could not be appropriately captured with the current data. For instance, if stronger rebel groups drop kidnapping as a strategy because already available financing strategies become more profitable, this could not be accounted for with binary data about participation or non-participation.
Finally, the findings for road density most clearly do not align, as the effects run counter to the theory for both extortion and smuggling. Accordingly, the results of the analysis contradict Hypothesis 2c–i.e. that increased state capacity through better infrastructure makes institutional crime less attractive. In hindsight, the relationship between smuggling and road density in particular might have been influenced by another factor I did not take into account: While higher road density might bolster state capacity and, therefore, limit insurgents’ opportunities to smuggle, it could–at the same time–increase opportunities by facilitating the transportation of contraband. In the case of extortion–which is a more localized crime–a similar explanation is more difficult. What I cannot exclude, however, is that the operationalization of road density at the national level was possibly insufficient and could have influenced these findings. Indeed, while I was able to find serviceable solutions to operationalize terrain ruggedness and conflict distance, better and more detailed geographical data would be a manner by which the current analysis could be improved. With more time and resources, raw data for road networks like the Global Road Inventory Dataset (Meijer et al. 2018) or border length data could be used to construct variables that account for the sub-regions in which insurgent groups operate. Although difficult to implement on a global scale, assessing the distance between conflict areas
and major military outposts–instead of capital cities–could better account for state power projection. With regards to military capacity, it might be a fruitful approach to also consider government counterinsurgency (COIN) strategies. Apart from military power per se, the choice of light- or heavy-force counterinsurgency could be another factor that influences a rebel group’s participation in crime. Unfortunately, at the time of this research project, a forthcoming dataset on government COIN strategies by Sullivan & Kareth (2019) was not yet available.
Of course, any data on the precise degree to which rebel groups are financially dependent on specific types of crime would be enormously valuable, because it could give a clearer idea about how certain strategies relate to each other. As mentioned already, this could, for instance, corroborate or disprove the idea that some strategies are substituted with others. However, I am well aware that these types of data are extremely difficult to gather systematically for large-N analyses.
In conclusion, the analysis shows that, in particular, environmental & infrastructural variables typically associated with state power and capacity are important to predict which types of criminal funding strategies insurgent groups choose. Partially backing the theory, there is evidence that participation in institutional crimes–i.e. extortion, smuggling and drug crimes–is increased, in each case, by mountainous terrain, the remoteness of the conflict zone and cross-border activity. But overall the theory is only somewhat supported, because military power itself is uncorrelated with institutional crime and better transport infrastructure increases the rate of participation for extortion and smuggling. With regards to episodic crime, there is some evidence that kidnapping for ransom, in particular, becomes less attractive when rebel groups are stronger, which could suggest a substitution effect towards higher-revenue income streams. Aside from testing the proposed theory with more sophisticated data, further research could also look beyond the differentiation between institutional and episodic crime. While the results for the theory proposed are inconclusive, there are other interesting patterns that might warrant explanations. For instance, it is notable that smuggling and robbery are both predicted by the same variables–high levels of road density and cross-border activity. It could be that these crimes tend to cluster in specific environments regardless of, for instance, the different costs and capabilities they require. In any case, the current analysis has shown that the context of the conflict should be taken seriously, if research wants to accurately predict why insurgent groups enter into the world of crime.
References
Abuza, Z. (2010). The Philippines chips away at the Abu Sayyaf Group’s strength. CTC
Sentinel, 3(4), 11-13.
Allansson, M., Melander, E., & Themnér, L. (2017). Organized violence, 1989–2016. Journal
of Peace Research, 54(4), 574-587.
Anderson, N., & Worsnop, A. (2019). Fatality Thresholds, Causal Heterogeneity, and Civil War Research: Reconsidering the Link Between Narcotics and Conflict. Political Science
Research and Methods, 7(1), 85-105.
Asal, V., Milward, H. B., & Schoon, E. W. (2015). When terrorists go bad: Analyzing terrorist organizations’ involvement in drug smuggling. International Studies Quarterly, 59(1), 112-123.
Asal, V. H., & Rethemeyer, R. K. (2015). Big Allied and Dangerous Dataset Version 2. Asal, V. H., Rethemeyer, R. K., & Schoon, E. W. (2019). Crime, Conflict, and the Legitimacy Trade-Off: Explaining Variation in Insurgents’ Participation in Crime. The Journal of
Politics, 81(2), 000-000.
Brunnschweiler, C. N., & Bulte, E. H. (2009). Natural resources and violent conflict: resource abundance, dependence, and the onset of civil wars. Oxford economic papers, 61(4), 651-674. Buhaug, H., & Gates, S. (2002). The geography of civil war. Journal of Peace Research, 39(4), 417-433.
Buhaug, H., Gates, S., & Lujala, P. (2009). Geography, rebel capability, and the duration of civil conflict. Journal of Conflict Resolution, 53(4), 544-569.
Carter, D. B., & Signorino, C. S. (2010). Back to the future: Modeling time dependence in binary data. Political Analysis, 18(3), 271-292.
Clayton, G. (2013). Relative rebel strength and the onset and outcome of civil war mediation. Journal of Peace Research, 50(5), 609-622.
Collier, P., & Hoeffler, A. (1998). On economic causes of civil war. Oxford economic
papers, 50(4), 563-573.
Conrad, J. M., Greene, K. T., Walsh, J. I., & Whitaker, B. E. (2019). Rebel Natural Resource Exploitation and Conflict Duration. Journal of Conflict Resolution, 63(3), 591-616.
Cornell, S. E. (2007). Narcotics and armed conflict: Interaction and implications. Studies in
Conflict & Terrorism, 30(3), 207-227.
Cornell, S., & Jonsson, M. (Eds.). (2014). Conflict, Crime, and the State in Postcommunist
Eurasia. University of Pennsylvania Press.
Croicu, M., & Sundberg, R. (2017). UCDP GED codebook version 17.1. Department of Peace
and Conflict Research, Uppsala University, 1-38.
Cunningham, D. E., Gleditsch, K. S., & Salehyan, I. (2013). Non-state actors in civil wars: A new dataset. Conflict Management and Peace Science, 30(5), 516-531.
Dadmehr, N. (2003). Tajikistan: regionalism and weakness. In: Rotberg, R.I., eds.: State failure
and state weakness in a time of Terror. Washington DC: Brookings Institution Press, 245-260
Dishman, C. (2005). The leaderless nexus: When crime and terror converge. Studies in Conflict
& Terrorism, 28(3), 237-252.
FAO (2011). Road density (per 100 square km of land area). Retrieved from https://landportal.org/book/indicator/fao-21017-6124
Fearon, J. D., & Laitin, D. D. (2003). Ethnicity, insurgency, and civil war. American political
science review, 97(1), 75-90.
Flanigan, S. T. (2008). Nonprofit service provision by insurgent organizations: the cases of Hizballah and the Tamil Tigers. Studies in Conflict & Terrorism, 31(6), 499-519.
Fortna, V. P., Lotito, N. J., & Rubin, M. A. (2018). Don't Bite the Hand that Feeds: Rebel Funding Sources and the Use of Terrorism in Civil Wars. International Studies
Quarterly, 62(4), 782-794.
Gent, S. E. (2011). Relative rebel strength and power sharing in intrastate conflicts. International Interactions, 37(2), 215-228.
Harbom, L., Melander, E., & Wallensteen, P. (2008). Dyadic dimensions of armed conflict, 1946—2007. Journal of peace research, 45(5), 697-710.
Herbst, J. (2014). States and power in Africa: Comparative lessons in authority and
control (Vol. 149). Princeton University Press.
Holtermann, H. (2016). Relative Capacity and the Spread of Rebellion: Insights from Nepal. Journal of Conflict Resolution, 60(3), 501-529.
Hough, P. A. (2011). Guerrilla insurgency as organized crime: Explaining the so-called “political involution” of the revolutionary armed forces of Colombia. Politics & Society, 39(3), 379-414.
Kenney, M. (2007). The architecture of drug trafficking: network forms of organisation in the Colombian cocaine trade. Global crime, 8(3), 233-259.
Lujala, P. (2009). Deadly combat over natural resources: Gems, petroleum, drugs, and the severity of armed civil conflict. Journal of Conflict Resolution, 53(1), 50-71.
Makarenko, T. (2004). The crime-terror continuum: tracing the interplay between transnational organised crime and terrorism. Global crime, 6(1), 129-145.
Mann, M. (1984). The autonomous power of the state: its origins, mechanisms and results. European Journal of Sociology/Archives européennes de sociologie, 25(2), 185-213. Mann, M. (2008). Infrastructural power revisited. Studies in Comparative International
Development, 43(3-4), 355.
McDougall, A. (2009). State power and its implications for civil war in Colombia. Studies in
Meijer, J. R., Huijbregts, M. A., Schotten, K. C., & Schipper, A. M. (2018). Global patterns of current and future road infrastructure. Environmental Research Letters, 13(6), 064006.
Naylor, R. T. (1993). The insurgent economy: Black market operations of guerrilla organizations. Crime, Law and Social Change, 20(1), 13-51.
NPR (2014, August). Islamic State ‘Beyond Anything We’ve Seen’ Hagel Says. Retrieved from https://www.npr.org/sections/thetwo-way/2014/08/21/342232267/isis-beyond-anything-we-ve-seen-hagel-says?t=1563360213836
Paoli, L., & Reuter, P. (2008). Drug trafficking and ethnic minorities in Western Europe. European Journal of Criminology, 5(1), 13-37.
Picarelli, J., & Shelley, L. I. (2007). Organized crime and terrorism. Terrorism Financing and
State Responses, 40.
Raleigh, C. (2010). Seeing the Forest for the Trees: Does Physical Geography Affect a State's Conflict Risk?. International interactions, 36(4), 384-410.
Ross, M. L. (2004). How do natural resources influence civil war? Evidence from thirteen cases. International organization, 58(1), 35-67.
Schutte, S. (2015). Geography, outcome, and casualties: A unified model of insurgency. Journal of Conflict Resolution, 59(6), 1101-1128.
Shaver, A., Carter, D. B., & Shawa, T. W. (2019). Terrain ruggedness and land cover: Improved data for most research designs. Conflict Management and Peace Science, 36(2), 191-218. Shelley, L. I., & Picarelli, J. T. (2005). Methods and motives: Exploring links between transnational organized crime and international terrorism. Trends in Organized Crime, 9(2), 52-67.
Smith, J. O. (2011). Maritime interdiction in Sri Lanka's counterinsurgency. Small Wars &
Insurgencies, 22(3), 448-466.
Soifer, H. (2008). State infrastructural power: Approaches to conceptualization and measurement. Studies in Comparative International Development, 43(3-4), 231.
Soifer, H., & Vom Hau, M. (2008). Unpacking the strength of the state: The utility of state infrastructural power. Studies in Comparative International Development, 43(3-4), 219. Sullivan, P. L., & Karreth, J. (2019). Strategies and Tactics in Armed Conflict: How Governments and Foreign Interveners Respond to Insurgent Threats. Journal of Conflict
Resolution, 0022002719828103.
Sundberg, R., & Melander, E. (2013). Introducing the UCDP georeferenced event dataset. Journal of Peace Research, 50(4), 523-532.
Tian, Y., Wu, B., Zhang, L., Li, Q., Jia, K., & Wen, M. (2011). Opium poppy monitoring with remote sensing in North Myanmar. International Journal of Drug Policy, 22(4), 278-284.
UN, D. (2015). World urbanization prospects: The 2014 revision. United Nations Department
of Economics and Social Affairs, Population Division: New York, NY, USA.
Washington Post (2015). How the Islamic State makes its money. Retreived from
https://www.washingtonpost.com/news/wonk/wp/2015/11/18/how-isis-makes-its-money/?utm_term=.ccd19cdc9b95
Weinstein, J. M. (2006). Inside rebellion: The politics of insurgent violence. Cambridge University Press.
Wennmann, A. (2011). Economic dimensions of armed groups: profiling the financing, costs, and agendas and their implications for mediated engagements. International Review of the Red
Cross, 93(882), 333-352.
Appendix
Complete logistic regression tables (including control variables)
