In the Eye of the Storm: Rebel Taxation of Artisanal Mines and Strategies of Violence

On Vicious Rebels and Natural Resources

The share of scholarly work on natural resources and conflict that addresses the role of rebel groups is substantial. A centerpiece is Collier and Hoeffler’s Greed and Grievances in Civil War (2004), who argue that the initiation of rebellions does not occur in response to grievances but to greed: rebel groups incite civil wars because natural resources provide favorable economic opportunities for self-enrichment. Employing the share of primary commodity exports of gross domestic product as a proxy for such possibilities, the authors reveal a curvilinear effect on civil war onset. The study has received considerable public as well as academic attention and thus crucially shaped the development of the research agenda.

This subsequent literature can be sorted into four strands. The first addresses Collier and Hoeffler’s (2004) work directly and examines similar data. The second takes a closer look at the types of natural resources and their production method. The third is devoted to disaggregation with geographically fine-grained data. Common to all three groups is either the notion that rebel groups with access to natural resources have a conflict-promoting effect—or the ambition to invalidate this finding. This demarcates the fourth category emphasizing that rebels sourcing revenues from natural resources should profit from stable production conditions and therefore have incentives to decrease violence. In the following, I will briefly discuss a nonexhaustive sample of studies from each strand and end with a key omission in debate.

In line with Collier and Hoeffler’s (2004) rationale, De Soysa (2002) finds that being an oil exporter almost doubles a countries’ risk of experiencing civil war. Using Collier and Hoeffler’s (2004) sample, Fearon (2005) likewise substantiates the adverse effects of oil production but invokes a rival mechanism. As it is typically governments that can collect oil rents through their control of distributive systems, these tend to suffer from weaker institutions. Rebels should thus be unable to finance their struggles with oil revenues but instead have strong incentives to capture governmental control. Humphreys (2005) criticizes that the hitherto interpretation of primary commodity exports may be spurious. In anticipation of civil conflict, industries unrelated to natural resources may abandon a country and leave behind only the extracting sectors inapt to relocation. Moreover, he specifies the theoretic arguments on rebels’ behavior advanced by Collier and Hoeffler (2004) into distinct causal mechanisms: the greed mechanism is in place if rebels gain from natural resources through criminal activities, state control, or secessionism in resource-rich regions. The feasibility mechanism, in turn, enables rebels to finance conflicts that were initially started for other reasons. In his examination, the author nevertheless finds little evidence that natural resources affect civil war via rebel groups. Other researchers, such as Brunnschweiler and Bulte (2009), outright disagree with the view that natural resources promote intrastate conflict based on the finding that oil and mineral reserves have a negative association with civil war onset.

Despite these partly challenging findings, the idea that well-resourced states experience civil war because rebels are enabled to act out on their violent propensity remained influential. Studies grouped in the second strand thus focus on conducive scope conditions—foremost the resource type and method of extraction. For a sample of thirteen civil wars, Ross (2004) finds qualitative evidence that rebels frequently access natural resources after a civil war has commenced, which prolongs hostilities. A crucial feature in this regard is the lootability of a natural resource by rebels, which is pronounced if the mining is conducted by unskilled workers in small groups that are easy to control (but see Ross 2006). Snyder and Bhavnani (2005) similarly argue that natural resources with high economic barriers to entry—which distinguishes industrially produced minerals from labor intensive artisanal extraction—could favor a ruler’s access to revenues and facilitate his or her ability to govern. Owing to the barrier, rebels cannot profit from natural resources and are hence unable to give into their thuggish disposition. According to the authors, this could explain the absence of civil war in resource-rich countries. In line with this logic, Lujala, Gleditsch, and Gilmore (2005) advance quantitative evidence of an association between alluvial diamonds and the start and continuation of intrastate conflicts. Since alluvial deposits occur above ground, are spread over larger areas, and can be mined with simple artisanal tools, rebels in countries endowed with this type of resource can increase the incidence of civil war. Nevertheless, there are also arguments against conflictual effects of resources easy to access for rebels. Dal Bó and Dal Bó (2011), for instance, propose a formal model predicting that positive price shocks to minerals being extracted with capital rather than labor-intensive methods raise the risk of civil conflict. This follows from the returns of appropriation that surmount the opportunity costs of war in case of capital-intensive natural resource production. Positive shocks to labor-intensive natural resource industries, in contrast, decrease this risk.

The persisting empirical ambiguity also gave rise to demands for subnational studies. Since natural resources are unevenly distributed within countries, scholars pointed out that their geographic locality should influence rebels’ ability to extract revenues and finance conflict (Buhaug and Gates 2002; Gilmore et al. 2005; Le Billon 2001). Several articles—summarized under the third strand—have taken up this suggestion. Østby, Nordås, and Rød (2009) reveal for sub-Saharan Africa that relative deprivation at the regional level in combination with the presence of natural resources increases the risk of conflict. From this, the authors infer that conflict entrepreneurs have at hand funds necessary to promote rebellions as well as poorer populations that are easier to mobilize. Exploiting an exogenous price shock, Angrist and Kugler (2008) scrutinize the effect of Columbian coca production on violence in regional departments. The results indicate that the shock provided rural paramilitaries with the finances essential to provoke civil conflict. For the same country, Dube and Vargas (2013) likewise present indication of a rapacity effect at the municipality level, driven by a positive price shock in labor-intensive natural resource sectors such as the oil industry. Following the researchers, this supports the idea that armed groups target regions rich in natural resources to extort rents.

Other scholars use rectangular grid data to disaggregate the resource-conflict nexus within countries. For African cells between 1970 and 2001, Buhaug and Rød (2006) find the probability of governmental intrastate conflict to be positively and that of secessionist wars to be negatively affected by proximity to diamond deposits. The authors expound this in favor of warlords being encouraged by natural resources to overthrow regimes. In a resembling analysis, Berman et al. (2017) provide two intriguing pieces of evidence in support of the greedy rebels hypothesis. First, they illustrate that rebel groups are more likely to be involved in conflict if their ethnic homeland contains mines. Second, they demonstrate that rebel groups are more likely to be involved in conflict onset after having previously won battles in mining grids. The authors consequently argue that rebels can capture on mineral rents through exaction in their territories, which increases their fighting capacity and leads to an escalation of violence across time and space. Drawing on gridded data covering Africa, Arezki, Bhattacharyya, and Mamo (2015) refute an effect of natural resources on conflict. Their results show that the discovery of oil and mineral resources is not associated with an increased conflict risk but instead with higher local income as indicated by nightlight measures. Examining conflict zones, Lujala (2009) discovers that in areas with lootable gemstones, conflicts claim larger death tolls and last longer. The author reads this as a validation of Weinstein’s (2007) proposition that resource-rich rebel groups become more violent as they attract increasingly opportunistic members. Moreover, she reasons it to be a rebuttal of Addison, Le Billon, and Murshed’s (2003) claim that resource-looting rebels become less prone to conflict out of a need to substitute violent activities with mining.

Such conclusions seem to represent the dominant view in the conflict literature. Where easily lootable natural resources occur, most scholars assume that profit generation by rebel groups could only be evidenced through a positive effect on conflict, while a negative or insignificant association is seen in favor of rebels that are unable to interfere in oil and mineral production.

Under the fourth strand, I subsume a number of examinations that put the picture of violent rebels into perspective by stressing their dependence on local populations and economic constraints. For a global sample of ethnoregions, Sorens (2011) shows that the presence of natural resources does not lead to looting by rebels in ethnically concentrated areas. In such cases, the groups instead participate in secessionist conflicts, which the author attributes to rebels’ need of support from communities composed of co-ethnics. Scrutinizing the 1991 to 2002 civil war in Sierra Leone, Bellows and Miguel (2009) point out that chiefdoms with diamond mines were only more likely to experience battles that are not targeted against civilians. Maystadt et al. (2014) show for the DRC that a higher number of mining concessions have no significant effect on monthly conflicts at the territory level. Within more aggregated districts, in contrast, it strongly increases the likelihood of violence. The authors report this as a protection effect that follows from mining companies’ efforts to relegate nearby conflicts for uninterrupted production.

Three recent studies apply an economic framework to explain the behavior of rebel groups directly. Parker and Vadheim (2017) exemplify the unintended adverse effects of the US 2010 Dodd–Frank Act that followed from the de facto ban of Congolese minerals. As the prices of these products dwindled, rebels were deprived of important revenues. This, in turn, decreased their appeals to protect miners in return for taxation and made violent looting of civilians more profitable. In line with this hypothesis, the authors provide evidence that territories in the DRC with more mines affected by the policy had a higher probability of experiencing violence against civilians. This finding is supported by the study of Stoop, Verpoorten, and van der Windt (2018) who explore updated and more comprehensive data. Another examination by De la Sierra (2020) advances a related logic but emphasizes the occurrence of taxation—a vital factor that does not feature empirically in Parker and Vadheim’s (2017) study. The author shows that if minerals are more challenging to tax, rebel groups establish intrusive administrative systems. These permit taxing the consumption of goods in villages of artisanal miners and in exchange provide protection against violent attacks.

Since the initial cognition of the link between valuable minerals or oil and intrastate conflict, the bulk of the literature thus appears absorbed with why rebels could account for it and whether it should result in a positive or negative association. So far, this discussion about eventuality and directionality neglects that the necessary answers first require a plausible rationale of how a rebel group would connect violence and profits from natural resources. In part, this could stem from a lacking sound rationale of how rebels exactly behave in resource-rich contexts (Korf 2011). Buhaug, Gates, and Lujala (2009), for instance, are one of the few scholars pointing out that rebels might employ small-scale violence systematically to preserve their role in the mining of lootable gemstones. The illegal nature of this activity, the authors contend, would make its pursuance difficult under peaceful times, which decreases rebels’ inclination to fully withhold violence. The next chapter develops this rationale further, by combining Olson’s (1993) framework of roving and stationary bandits with Kalyvas’ (2006) logic of violence in civil war. Its validation naturally requires details on rebel access to revenues from natural resources. Owing to a new data set—to be introduced below—about armed groups’ interference at artisanal mines in the eastern DRC, it becomes possible for the first time to extensively scrutinize the effects of resource-profiting rebels on violence.

The Calculus of Rebel Groups

Policy makers and activists seeking to diminish the link between natural resources and civil war often advance the DRC as an exemplary case of a country plagued by the resource curse. The minerals gold and the 3Ts—tantalum, tin, and tungsten—according to this view are a factor that not only significantly affected the Second Congo War but continues to spur human rights violations presently occurring in the country’s eastern region (Lalji 2007).

Pointing toward the complexity of the resource-conflict nexus, many researchers suggest more nuanced accounts in three regards. First, ethnic tensions and clashes about land property rights rather than natural resources were the strongest triggers for the civil war. It was only when the initial financial means of the different parties involved dwindled that revenues from valuable minerals became influential for the conflict’s dynamic. Once in force, this turn was nevertheless substantial—not a single armed group refrained from exploiting the DRC’s mineral wealth at some point during the war (Le Billon 2012, 109).

This association is still prevalent today. In the Virunga National Park, for instance, the rebel group Forces Démocratiques de Libération du Rwanda (FDLR) systematically uses violence to sustain businesses that produce and trade natural resources (Dranginis 2016). With the same means, the Islamist Alliance of Democratic Forces pursues both criminal and political goals (Irin 2014). Even groups as heterogeneous as the Mai-Mai militias are reported to have gone rogue to enrich themselves with natural resources (Verweijen and Marijnen 2018).

Second, the role that mineral extraction plays in financing armed groups is less straightforward than often implied in popular portrayals. In most cases, rebel groups extract revenues from the artisanal mining sector. Artisanal denotes a mode of production involving simple tools such as shovels and sieves instead of the mechanized equipment that is part of industrial mining. The manual method requires a considerable labor force that is estimated to range between 200,000 and 350,000 workers in the DRC. The World Bank estimates that each miner supplies work for an additional five people within the sector, whereby the occupational group would affect between 9 and 17 percent of the population in the eastern region. As this assessment does not regard other economic branches, such as the service industry that is substantially connected to artisanal mining, it may still underrate the true importance of the sector. Another crucial characteristic of artisanal mining is informality. Minerals extracted in artisanal mines commonly do not feature in official records even though in the case of gold the produced share lies at approximately 98 percent of the DRC’s exports (Geenen 2015, 5). Still, absent formalization in the artisanal sector does not imply that the extraction is unsystematic. Much to the contrary, artisanal miners need to comply with a system of rules that regulate their access to sites (Geenen 2015, 73).

Due to its unregulated and dynamic nature, the state faces severe obstacles in establishing a monopoly of control over artisanal mining (Snyder and Bhavani 2005). This feature, in combination with a low capital intensity and the high-value-to-weight ratio of minerals, in contrast, makes the method attractive to rebel groups (Van Bockstael and Vlassenroot 2009). In particular where rebel groups establish a lasting presence, the sites see the emergence of tax systems that involve collecting fees from workers. Miners are usually required to pay lump sum taxes for working permissions ranging from 1,000 to 2,500 Congolese francs per week (Weyns, Hoex, and Matthysen 2016).

Multiple groups are known for such illegal taxation at extraction sites. In the Mambasa territory, Mai-Mai Simba rebels control about thirty artisanal gold mines on which they levy taxes. The Nduma Defense of Congo (NDC) extracts revenues in the same fashion from over eighty gold mines in the Walikale territory. In the southeast of Walikale, the Mai-Mai Kifuafa also taxes several cassiterite mines that deliver over US$2,000 per month to the group. Still within the same territory but toward its western end, the Raïa Mutomboki is similarly involved in illegal taxation. The FDLR has developed the most extensive system of illegal taxation by a rebel group in the Lubero territory. Here, it established parallel administrative systems to maintain a firm grasp over gold mines (Weyns, Hoex, and Matthysen 2016).

Apart from minerals, rebel groups also profit from a diverse set of other natural resources such as charcoal, drugs, and even livestock (Laudati 2013). Still, this is does not imply that the role of minerals is negligible—the United Nations (UN) estimates that illegal natural resource exploitation values up to US$1 billion each year. Roughly 98 percent of its profit benefit transnational criminal networks and provide financing to as many as twenty-five armed groups that are active in the region (Zorba, Bambara, and Toure 2017). This significance is also reflected in the finding that more than half of the mines experience the presence of an armed group (Weyns, Hoex, and Matthysen 2016). It does, however, indicate that violence is unlikely to end if rebel groups loose access to mineral resource rents only (Stoop, Verpoorten, and van der Windt 2018). Past policies such as the Dodd–Frank Act, aimed at achieving precisely this, have unintendedly deteriorated local living conditions by raising rebels’ incentives to loot civilians (Parker and Vadheim 2017). Curtailing the effects resulting from rebel access to artisanal minerals and comparing their magnitude vis-à-vis alternative resources of financing therefore necessitate an understanding of how revenues from the mines are connected to violence in the first place.

Third, experts disagree with the common perception that artisanal miners are forced to work at gunpoint. Instead, the coercive nature of illegal taxation stems from the protection that armed groups offer in exchange for the rents. This more subtle system of coercive governance characterizes the wider Congolese economy as a remnant of the civil wars (Geenen 2015, 6). A substantial share among the population also sustains its livelihood from artisanal mining, and artisanal miners identify professionally with their work—sometimes to an extent that they would turn down alternative job opportunities (Geenen 2015, 206). Thus, it seems paramount to also recognize the opportunities that many people attain through the sector and to derive policy solutions that seek to retain the advantages of artisanal mining while diminishing the violent effects of rebels’ profit extraction.

Research Design

The central conjecture of this article is that rebel groups extracting tax revenues from an artisanal mine will thwart violence in its nearby environment but further it in areas that lie beyond a distance coverable on flight. Thereby, they can uphold the politically fragile situation that enables their profit-making—collecting taxes in return for protection—and encourage miners to work at their sites. Testing the hypotheses naturally requires details on where rebels collect taxes and engage in conflict. The following section describes a new data source which, for the first time, makes his information accessible and delineates the operationalization of indicators for violence in nearby and distant surroundings.

Data Sources

The unit of observation in this analysis is the individual artisanal mine. Data on the location of the mines as well as the presence and taxation of various armed groups on-site are drawn from the Belgian International Peace Information Service (IPIS). In making this data accessible, the IPIS provides information to international stakeholders who seek to address the links between armed groups’ financing modalities and mineral exploitation. The goal of the research institute is to provide a structural solution to the need for accurate and up-to-date information through a system that permanently monitors artisanal mining activities and the involvement of armed groups. The project was funded by the World Bank through its PROMINES program and the Belgian Ministry of Foreign Affairs (Weyns, Hoex, and Matthysen 2016, 6).

Between 2009 and 2016, IPIS surveyed 3,035 artisanal mines in the eastern DRC in collaboration with the Congolese Ministry of Mines. It first collected specifics on the location of the mines and subsequently sent out two-person field teams to visit the sites. The teams recorded the exact geo-coordinates of the mines, gathered information on the production and destination of the minerals, and interviewed several respondents per mine on the presence and interference of armed groups. An advantage of this data set in comparison to others such as the DIADATA (Gilmore et al. 2005) is therefore that it reports where and how armed groups are de facto accessing natural resources rents rather than where they could be. At a mine where armed groups taxing miners are observed, it is coded with the dichotomous variable taxation. This indicator forms the independent variable.

Information on violent conflict events around mines is sourced from the Uppsala Conflict Data Program Georeferenced Event Dataset (Sundberg and Melander 2013). ² It records individual conflict events with at least one casualty and informs about their location, the participating sides, and estimated casualties. This analysis uses this particularly low threshold of violence since civilian miners—being unable to defend themselves against atrocities—should react sensitively as soon as violence turns deadly. I consider all 1,418 conflict events that occurred in the DRC between 2009 and 2016. Figure 1 depicts the locations of artisanal mines and violent conflicts and evidences a clustering of both phenomena in the eastern half of the country.

OPEN IN VIEWER

Figure 1.

Artisanal mines and conflict. The crosses show the location of Uppsala Conflict Data Program conflict events in the Democratic Republic of the Congo (DRC) between 2009 and 2016 with one casualty or more. The points depict the position of all artisanal mining sites in the eastern DRC. These have been surveyed by the International Peace Information Service between 2009 and 2016.

Violence around Mines

The geographic areas labeled immediate and distant surrounding in the hypothesis are operationalized in four steps. To examine the spatial association between artisanal mining and the incidence of violence, I first measure the distance between each mine and every conflict using a geographic information system program. This yields a matrix with 4,303,630 distances that connect each site to every conflict event. In the second step, I use the temporal information on the start of the conflicts and the field teams’ mine visits to drop the connections where the violence occurred before a mine had been surveyed. This is done to avoid reverse causality in case that rebels tend to establish taxation at mines that have been peaceful in the past. ³

Another concern is that field teams could be more likely to visit less violent mines to avoid danger. Yet, there is little reason to suspect that this creates a selection bias specifically in favor of rebel-taxed mines. The precise interference by rebels should be unknown in advance—otherwise, actual visits would be futile.

For each mine, I isolate the conflict event with the smallest distance in the third step. The idea behind this focus on the nearest violence is that events after the closest event should be less relevant to miners as they primarily respond to the area that is entirely exempt from violence. ⁴

In the fourth step, I compute three areas around mines in which the incidence of the closest conflict is examined. The first immediate area is a circular zone with a 40 km radius around each mine. This initial value is based on the earlier note that rebels’ mines are predominantly accessible by foot only, which relates the security zone to the maximum distance walkable within a day. The second distant area covers the occurrence of the nearest violence within a ring-shaped zone stretching from an inner radius of 40 km to an outer radius of 56 km. The upper limit of this area is chosen such that both areas are of the same size of roughly 5,000 km²—as otherwise conflict is more likely in the larger area. ⁵ This requires that without its center void, the second area is twice the size of the first, whereby the subtraction of the first leads to an equally sized geographic area. ⁶ The third area is calculated in the same fashion. ⁷ Figure 2 depicts the three areas schematically, together with the surrounding conflict events and marks the advances of the nearest violence—most relevant to this analysis—in red. Each area is represented by a separate dependent variable which takes the value 1 if the closest conflict event takes place in it and 0 otherwise.

OPEN IN VIEWER

Figure 2.

Areas around mines.

Estimation Strategy

Due to the binary nature of the dependent variable, I estimate a logistic regression based on the following equation:

Y i j = β 0 + β 1 × r e b e l t a x a t i o n j + β 2 × X j + ∊ j .

Y reports the incidence of the closest conflict in area i around mine j. The coefficient of interest is β₁ that estimates the effect of rebel taxation at mine j on the likelihood that the nearest conflict occurs within area i. X denotes the vector of control variables referring to area i around mine j. ε_ij is the error term. ¹⁰

Findings and Discussion

Table 1 depicts the regression results that show the effect of rebel taxation on the likelihood that the nearest violence occurs in each of the three areas surrounding a mine. Transforming the log odds of the first model shows that rebel taxation significantly decreases the probability of the closest violence taking place within a radius of 40 km by 35 percent. Moreover, within the ring-shaped area of 40 to 56 km—examined in the second model—it significantly raises this probability by 76 percent. In the last model, which refers to the ring-shaped area of 56 to 69 km, the coefficient loses significance. The results thus support the hypothesis: taxing rebels shape the patterns of violence in areas surrounding artisanal mines by exempting its immediate vicinity from it but not its more distant environment.

OPEN IN VIEWER

Table 1.

Closest Conflict Event around Mines and Rebel Taxation.

Variables	(1)	(2)	(3)
	Area 1: 0–40 km	Area 2: 40–56 km	Area 3: 56–69 km
Rebel taxation	−0.596*** (.188)	1.173*** (.169)	−0.524* (.278)
Rebel presence	0.562*** (.158)	0.0935 (.216)	−0.742** (.315)
State service	−1.446*** (.0975)	−0.622*** (.141)	0.0187 (.165)
Forced labor	−1.411*** (.332)	0.271 (.380)	0.985*** (.351)
Digging	0.769** (.301)	−0.594 (.448)	0.757** (.362)
Pillaging	−1.034*** (.257)	−0.663** (.309)	1.170*** (.302)
3T	0.699*** (.168)	0.315 (.236)	−0.339 (.302)
Gold	0.263 (.168)	−0.119 (.240)	0.232 (.297)
Diamond	0.252 (.280)	−0.251 (.452)	0.958** (.388)
Conflict before survey	1.234*** (.0883)	1.180*** (.139)	1.799*** (.208)
Rebel-taxed mines	0.00288 (.00373)	0.0318*** (.00461)	0.0284*** (.00585)
Constant	−1.069*** (.177)	−2.331*** (.243)	−3.032*** (.305)
Observations	3,035	3,035	3,035

Note: Logistic regressions with occurrence of nearest conflict event are given as dependent variable. Rebel taxation at mining site is given as independent variable. Standard errors are given in parentheses.

*p < .1.

**p < .05.

***p < .01.

The control variable which most consistently behaves as anticipated is state service. Where it is present, it decreases the likelihood of violence in the first two areas. The negative effect of forced labor in the most proximate area to a mine is surprising, as the expectation is that it should limit miners’ ability to escape violence. It could indicate that groups strong enough to enforce labor no longer depend on violence to uphold this system. This strength would also explain why mines with forced labor are more likely to experience the nearest violence only in the most distant area.

Digging by soldiers appears to make armed groups less dependent on miners since violence nearby occurs more often if this activity is observed. Pillaging, as exemplified by its negative effect on violence in the first two areas, seems to be conducted in a manner which does not discourage artisanal production. Again, this seemingly requires groups capable of defending their territory against other rebels since violence becomes more likely only in the most distant third area. The nearby effect of minerals is only pronounced for 3T that is associated with more violence. A similar effect pertains to diamond in the third area. Conflict before the survey and rebel-taxed mines predominantly increase the likelihood of close-by conflict.

Two questions naturally arise from these results: Is it the taxing rebels that incite the conflict, and does this in fact reflect strategic behavior? I examine the first in Table 2 by narrowing the sample in two regards. First, I focus on rebel mines, which reduces the sample to the 525 mines that either experience rebel taxation (281) or mere rebel presence (244). Second, I examine only those nearest conflict events in which the group controlling the mine participates. This allows me to estimate the effect of taxation by a rebel group on the likelihood that the nearest violence involving this group occurs in the surrounding. ¹¹

OPEN IN VIEWER

Table 2.

Closest Own-group Conflict Event around Rebel Mines and Taxation.

Variables	(1)	(2)	(3)
	Area 1: 0–40 km	Area 2: 40–56 km	Area 3: 56–69 km
Rebel taxation	−0.893*** (.247)	2.265*** (.428)	1.734*** (.476)
Pillaging	−0.948*** (.322)	−1.641*** (.465)	0.548 (.446)
3T	0.185 (.441)	0.875 (.746)	0.396 (.840)
Gold	0.566 (.466)	0.382 (.775)	−1.797** (.864)
Conflict before survey	1.106*** (.229)	2.221*** (.347)	−1.112 (.783)
Rebel-taxed mines	0.0157*** (.00499)	−0.0401* (.0222)	−0.0604** (.0268)
Constant	−1.679*** (.486)	−3.863*** (.902)	−2.301*** (.856)
Observations	525	525	525

Note: Logistic regressions with occurrence of nearest conflict event involving rebel group present at the mine are given as dependent variable. Rebel taxation at mining site is given as independent variable. Standard errors are given in parentheses.

*p < .1.

**p < .05.

***p < .01.

The results show that taxation by a rebel group significantly decreases the probability that the immediate surrounding of a mine will see the closest violence involving it by 30 percent. In addition, it increases precisely this probability in the second area by 90 percent. The findings thus corroborate the theory that rebels generating taxes from artisanal mining will both decrease and increase violence at different locations in relation to a mine.

Figure 3 addresses the second—namely if rebels’ violence originates from ordered or undisciplined behavior. It depicts the logarithmic cumulative distance distribution between the 525 rebel mines and the first, fifth, and tenth closest civilian casualty of conflicts in which the rebels participated. I obtained the values from the matrix spatially connecting all conflicts and mines by multiplying each event with its number of civilian casualties and sorting the data by distance. This allowed me to select the margin from each mine at which a given number of civilians had been killed by violence involving the group controlling the mine. The plot shows the values for the 281 rebel-taxed mines in red and the 244 untaxed rebel mines in blue. More generally, it corroborates the “eye of the storm” theory: smaller distances between rebel mines and casualties are more common for untaxed mines, and distances above 40 km are more common for taxed mines. This yields the blue shaded area protected by taxing rebels and the red shaded area which is imperiled.

OPEN IN VIEWER

Figure 3.

Mine-casualty distance for taxed and untaxed rebel mines. The figure shows percentiles of the distance between taxed and untaxed rebel mines and the first, fifth, and tenth closest civilian casualties. Classification of taxed and untaxed rebel mines is based on International Peace Information Service data collected 2009 to 2016. Untaxed rebel mines see regular rebel group presence but no interference in mining activities. Taxation denotes rebels’ extraction of fees from workers. Civilian casualties refers to the variable deaths_civilians in the Uppsala Conflict Data Program Georeferenced Event Dataset for events in the DRC 2009 to 2016.

To examine whether this results from strategic behavior, I derive two observable implications of undisciplined violence: first, violence should not be a steady function of distance since it does not depend on the costs of projecting authority across space. ¹² Instead, the imperiled area should emerge because taxing groups are stronger and thus able to secure the larger areas needed to incentivize miners. From the point on where security is no longer relevant, the red line should have a meandering slope resembling the blue line—illustrating that spatially determined costs of orchestrating violence are not systematically related to its occurrence. Second, undisciplined violence should predominantly lead to smaller numbers of casualties, as these require less organization and can be inflicted spontaneously and independently of spatial constraints. In this case, we would expect distances to the first civilian death to increase less steadily than distances to larger casualties. All three panels give evidence contradicting the view that violence occurs through lacking discipline. The distances between taxed mines and civilian deaths accumulate steeply at a steady rate and are indistinguishable across casualties. This supports the reading that the pattern of violence surrounding rebel-taxed mines stems from strategic calculus.

Conclusion

The largest share of the resource curse literature on conflict portrays rebels extracting profits from natural resources as being prone to inflict violence, while a smaller body advances incentives to withhold it. This article argues that explanations for the resource-conflict nexus remain incomplete as they do not recognize the dual and spatial nature of the relationship. I theorize that decreasing the likelihood of violence near mining sites enables rebel groups to extract taxes since it creates economic environments suitable for continuous production. Simultaneously, upholding demand for this service requires rebels to manifest the threat they offer protection against. The spatially limited extent to which they need to safeguard mines allows the groups to actively inflict violence at locations that lie beyond their full control. This reminds workers of the security aspect without discouraging them to mine at the taxed site.

I examine my theory with data on the behavior of armed groups in an area notorious for the conflictual effect of natural resources—around over 3,000 artisanal mines in the eastern DRC. The results support the hypotheses, evidencing that taxing rebels create an environment that resembles the eye of a storm: they both increase the relative safety of taxed artisanal mines and decrease that of their wider surroundings. The analysis moreover reveals that this pattern is attributable to strategic behavior by the very groups that levy the taxes, as violent events involving specifically these groups spread across space in the hypothesized fashion.

The findings open up fruitful avenues for future research. First, it is crucial to examine whether the twofold effects of profit-making on violence also hold under different settings, as taxation of minerals is but one of multiple ways for armed groups to attain financial means. At times, rebels also generate revenues from other resources such as charcoal, livestock, and cannabis (Laudati 2013). Second, it would be interesting to learn more about the behavior of the state army around artisanal mines since mines taxed by this actor surmount those taxed by rebels. Third, the effects armed groups have on artisanal mining also depend on how miners react to their interference. It could therefore be sensible to collect information of how sensitive they are to nearby violence, whether they value rebels’ protection, and how easily they can migrate to other mines to avoid it.

Legislators may likewise find the results revealing since they speak to policies such as the Dodd–Frank Act that seeks to curtail rebels’ access to minerals by obstructing their export. Recent contributions stress that they may have unintended detrimental effects at the territory level, as they decrease the groups’ incentives to protect miners (Parker and Vadheim 2017; Stoop, Verpoorten, and van der Windt 2018). This article adds the disaggregated insight that taxing rebels not only safeguard mines but also increase violence in their smaller adjacent mining areas, precisely to maintain a demand for security. Restraining from policy initiatives may thus still leave populations living in affected areas worse off. A more pertinent policy approach may instead be to increase the protection of populations that live in the areas around mines that are more likely to see violence. Thereby, rebels would lose the basis for demanding taxes, as they can no longer implement the violence they offer protection against.

Supplemental Material

Supplemental Material

Supplemental Material