Natural Resources and Civil Conflict

Natural Resources and Civil Conflict: An Analysis of Existing Literature

Lejla Delic

Ottawa, Ontario April 14th, 2015

ABSTRACT There is a currently a growing literature investigating the relationship between natural resources and civil conflict. The general question of this strain of research seeks to answer whether natural resource abundance relates to the propensity for conflict or to economic growth. This paper is a review of existing literature of three empirical works undertaken to provide better-specified characteristics of this relationship. The main specific question that is addressed in this paper is whether there is a relationship between natural resources and the associated severity of civil conflict, while specifically focusing on the mechanisms that link natural resources to conflict, the location of resources relative to rebel groups, and the types of natural resources. The methods used by the authors for their different analyses range from the switching regression model, the Weibull distribution model and a standard logit model. It is argued that literature should move away from focus on the recurring rebel greed hypothesis underlying the relationship between natural resources and conflict. Specifically, other mechanisms that could be explain the effects of natural resources on violence. Another finding in response to this argument holds that the relationship is largely driven by rebel groups’ incentives and opportunities when considering the location of resources and rebel access. Finally, when specifying the type of natural resource, with respect to the ease of its expropriation, a strong link between diamonds, a very lootable resource, and conflict onset is established.

1

1. Introduction There is a currently a growing literature investigating the relationship between natural resources and civil conflict. The general question of this strain of research seeks to answer whether natural resource abundance relates to the propensity for conflict or to economic growth. It is often argued that the wealth arising from natural resources is an explanation for inclination towards civil conflict, and that rebels may gain power over the resources through extortion or appropriation and cause conflict as a result of this. The underlying argument here is that rebel groups are presented with opportunities to fund their activities through the rents natural resources provide, or through the money they extort from firms or institutions in power of those resources. This paper is a review of existing literature of three empirical works undertaken to provide better-specified characteristics of this relationship. The main question addressed in this paper is whether there is a relationship between natural resources and the associated severity of civil conflict. This is done by specifically focusing on all the possible mechanisms that link natural resources to conflict, not just on the motivation of greedy rebels. In addition, the physical location of resources relative to rebel movements is considered when trying to characterize the relationship between natural resources and conflict, as well as the types of natural resources. In this paper’s context, the types of resources are either lootable or non-lootable, The methods used by the authors for their different analyses include the switching regression model when examining the varying effects of rival mechanisms, the Weibull distribution model for considering the impact of location and rebel access to natural resources, and a standard logit model when splitting up the effects lootable and non-lootable resources have on conflict. The main findings of the selected papers, in a way, build upon each other. Humphreys (2005) argues that existing literature should move away from focus on the recurring rebel greed 2

hypothesis underlying the relationship between natural resources and conflict, stating that conflict relating to natural resources does not necessarily imply that it is driven by rebel greed or motivation. Specifically, he provides and discusses other mechanisms that could explain the effects of natural resources on violence, such as the status of state strength and capacity, or grievances caused by natural resources. In addition, he holds that policy implications and reforms would be improved with literature findings that suggest changes in how states manage their revenues and maintain their relations with the public. In her research, Lujala (2010) responds to this argument and holds that the relationship is largely driven by rebel groups’ incentives and opportunities, especially when considering the location of resources and rebel access. She finds that the location of natural resources is critical, and specifically that when resources are located inside a region with conflict, the duration of conflict doubles. Finally, when specifying the type of natural resource, with respect to the ease of its expropriation, Lujala, Gleditsch and Gilmore find strong link between diamonds, a very lootable resource, and conflict onset is established. Noting that secondary diamonds are considered lootable resources, whereas primary diamonds are considered non-lootable, they also find that primary diamonds do not have this same effect on conflict. The paper is divided into three sections. The following section provides a review of the three empirical works chosen for this paper. The first subsection discusses a paper that examines rival mechanisms that underlie the relationship between natural resources and the propensity for conflict. The second subsection considers the role of location of natural resources and rebel access to them when determining the effect of natural resources on conflict. The third subsection considers different types of natural resources, namely those of lootable and non-lootable nature, and their opposing effects on conflict. This third section concludes the paper.

3

2. Literature Review This section will provide a review of the three papers chosen for this analysis. The authors of all three works seek to better identify the characteristics of the link between natural resources and civil conflict. The first subsection will discuss a paper by Macartan Humphreys, who holds that the mainstream rebel greed hypothesis is not sufficient in describing the relationship, and that other mechanisms and factors need to be considered. The second subsection discusses Païvi Lujala’s rebuttal to this argument, holding that greed and motivation by rebels is indeed crucial to the relationship between natural resources and conflict, especially when considering the location of natural resources relative to rebel groups. The final subsection will discuss research conducted by Lujala, Gleditsch and Gilmore that analyzes the link as well by disaggregating diamonds into their lootable and non-lootable forms, specifically, secondary and primary diamonds. 2.1. “Natural Resources, Conflict and Conflict Resolution: Uncovering the Mechanisms” In his research, Humphreys identifies and discusses other mechanisms that underlie the link between natural resources and violence, arguing that the recurring rebel greed mechanism is but one of many that could explain the link. In addition to identifying the mechanisms, Humphreys implements different methodologies for identifying which of the rival mechanisms would apply to which situation. He underlines the importance of considering other mechanisms besides rebel greed, noting that “until the different mechanisms are understood, advice of conflict scholars will be of limited use” (Humphreys, 2005, page 509). Humphreys considers an African sample for one of his tests and uses a global sample for the remainder. These samples include countries that have relatively little natural resource abundance, but have experienced longer duration of conflicts, such as in Afghanistan, Ethiopia and Somalia; 4

the samples also include countries that are abundant in natural resources but have experienced relatively shorter conflict duration, such as in Nigeria, Yemen and Croatia (Humphreys, 2005, page 531). Contrary to previous studies examining the link between natural resources and conflict, Humphreys (2005) argues that there are at least six other mechanisms that could explain the link (page 510). He defines them as follows. The first is the greedy rebels mechanism, whereby rebel movements can take place to either “benefit from resources independent of the state,” or because “natural resources increase the prize of capturing the state” (page 511). The greedy outsiders mechanism may occur when neighboring states are motivated to “engage in or foster civil conflicts” (page 511). The grievance mechanism may be as a result of unequal distribution of wealth and gains, or whereby those that are forced to migrate due to resource extraction experience grievance because of their loss of land rights (page 511). The feasibility mechanism is defined as the occurrence when rebel movements that have begun for reasons other than natural resources are financed by natural resource appropriation. The weak states mechanism argues that state structures may be weaker in countries with natural resource dependence (Humphreys, 2005, page 512). Specifically, Humphreys he describes the effect as coming from both the state and society sides of the link; citizens who have less information or power over their government have less incentive to monitor its activity, whereas governments “that rely on natural resources rather than taxation have weakened incentives to create strong bureaucratic institutions,” leading to states such as oil dependent ones being “more likely to have weak structures because they have less need for intrusive bureaucracies to raise revenues” (page 513). Finally, the sparse networks mechanism signifies that countries with economic fragmentation, whereby there are difference enclaves of production, have a higher risk of conflict (page 513).

5

Humphreys further argues that “internal trade is associated with greater levels of social cohesion,” and that “dense trade networks reduce conflict risks” (page 513). This main data used in this research was collected on the lootable resource in question, diamonds, as well as oil, which is a less less lootable, but still appropriable, resource. His new diamond data is more fine-grain than that used in previous literature, and although it did not report whether the diamond was primary or secondary, the dataset provides specific quantities for Humphreys to use in his research. It also does not solely rely on export data, as it includes “information gathered from actors in the industry and mining corporation” (page 523). Secondary diamonds are considered more lootable than primary diamonds, as secondary diamonds are generally found above ground, or very close to the surface, whereas primary diamonds are less easy to find and exploit. The data was taken from three different datasets: Mining Annual Reviews, Metals and Minerals Annual Review, and the Diamond Registry. The data used for oil production and reserves, taken from the BP Statistical Review of World Energy and BP Statistical Review of World Oil Industry, allows for Humphreys to distinguish between oil that has not yet been extracted and oil that has been produced in the past. In addition, the dataset excludes oil re-exports as to allow for differentiation between oil extraction, which provides rents, and “the more oil processing sector” (page 523). Finally, Humphreys (2005) uses the recorded share of agricultural value in national income as a rough measure of economic structures (page 523). The methodology used by Humphreys (2005) differs among his tests. He differentiated between “Type B” and “Type A” mechanisms “in which rival mechanisms relate to each other” (page 518). He explains that “Type B” situations arise when many mechanisms may be applied but that have opposing effects on the outcome of interest, disaggregating the dependent and

6

independent variables allows for more precise distinguishing between initially unobservable effects. An example of this is breaking a conflict variable into three variants, whereby the conflict persists, ends with a military victory for either side, or ends through negotiation (page 520). “Type A” situations occur when there are two possible ways in which the independent variable can be linked to the dependent variable, but that both cannot simultaneously occur. There are two methods that follows: either constructing an interactive term with a third variable if it is known that the independent variable has an impact on the dependent variable through the third variable; the second method is proceeding with the switching regressions method “to determine the individual characteristics of each rival mechanism” (page 521). This subsection will now present Humphrey’s findings. When testing for whether state strength plays a role in the link between lootable resources and conflict, Humphreys constructs three proxies as measures for state strength and weakness. These can be seen, separated into three columns, for each of the three natural resource measures in Table I. Instability is a measure of political instability, measured by whether a state has experience a large change in political institution over the past three years. Strong is a measure of instability and anocracy, taking a value of 1 of the state is a democracy or dictatorship, and 0 if otherwise (Humphreys, 2005, page 527). The Weberian variable measures whether the state has a monopoly over the legal use of force within a region. As seen in Table I, both Instability and Strong generally enter significantly into the models. The interaction term, however, is insignificant across the table with the exception of oil production. Focusing on Models I, II and III, it can be seen that there is weak evidence that oil production has a harmful impact across all three measures of state strength, but specifically for weak states. This effect is not observed for the other two resources measures.

7

Humphreys concludes that the sparse networks mechanism is supported in place of the rebel greed mechanism, as production enters significantly into the tests. This signifies that natural resource production in the past has an impact on civil conflict, whereas the mere existence of resource reserves, which are potential for future production and gains, do not have that same impact. In performing another test, he finds evidence of the sparse netoworks, specifying a relationship between primary commodities that is motivated in part by agricultural resource dependence. In addition, when testing for the link between natural resources and the length and duration of wars, he finds that natural resource wars tend to be shorter than other wars, and that they tend to end with a military victory for either side than by negotiation. He adds that there is no evidence that natural resources obstruct or aid negotiation means in terms of conflict, and that “external actors have no incentives to work to bring wars to a close when natural resource supplies are threatened” (Humphreys, 2005, page 508). This finding combined with the previous finding comes into direct opposition with the argument that rebels will prolong conflict if they have sufficient motivation and opportunity to do so. Humphreys argues that existing and leading research has conducted tests on the relationship between natural resources and conflict without accounting for mechanisms besides rebel greed that underlie the correlations. His paper provides the rival mechanisms as well as the tests and methods required to differentiate between them when testing for the outcome. He notes that although his findings provide preferred explanations than literature that focuses on the rebel greed mechanism, “the tests still suffer from severe data and specification issues” (Humphreys, 2005, page 534). In order to obtain even more precise results, as well as to be able to test all the mechanisms, more fine-grain data and better measures for the indicators are required.

8

In terms of policy implications, Humphreys’ finding of strong support for the weak states mechanism, as well as that natural resources have an especially harmful effect on already weak states, underlines the important of policy priority redirection; he suggests that initiatives such as better focus on proper management of resource extraction processes, as well as on “better usage of resource revenues controlled by states” (page 534). In terms of policies that could lessen the grievance mechanism, Humphreys (2005) suggests that governments better inform the public about the revenue expenditures and allow for public oversight of these expenditures (page 534). Finally, after finding support that resource wars generaly end with a victory for one military side rather than negotiation, Humphreys suggests that “policy responses should focus on establishing criteria for determining what regimes should be supported”, as the results “supported one-sided military interventions” in resource wars (page 535). 2.2. “The Spoils of Nature: Armed Civil Conflict and Rebel Access to Natural Resources” Païvi Lujala seeks to further investigate the link between natural resources and conflict. Given the results in Humphrey (2005), Lujala considers the location of natural resources, specifically hydrocarbon and gemstone production and reserves, as well as the role of rebel access to these resources as a new method not found in existing literature. She argues that if the rebel greed mechanism did not play an important role in the relationship between natural resources and conflict, and that other mechanisms were the only valid explanations, then accounting for rebel access to natural resources would not make a difference. The author conducts her study in the time period 1946 to 2003, looking at a minimum of 252 conflicts. Her data considers 885 onshore and 379 offshore regions of hydrocarbon (crude oil and natural gas) production and reserves found in 111 countries. 98 of these countries had oil, gas or

9

both present. Her diamond dataset allows her to use more than 1,000 diamond deposits worldwide. The theory behind Lujala’s research is that natural resources and their link to conflict is general explained by two different methods. The first argues that relatively lootable natural resources, such as diamonds, may provide rebel movements with the motivation and opportunity to finance their activities and raise the odds of their success. The other stream argues that the link between natural resource abundance and conflict is attributable to weak states and poor governing choices. Lujala (2010) notes that existing literature is based on these two hypotheses, and that advocates of both sides generally use the same indicators and measures in estimating their models, which she considers a weakness of existing literature (page 15). Referring back to the theory found in Humphreys (2005), this paper presents the same alternative mechanisms that could possibly explain the link between natural resources and conflict. However, Lujala (2010) holds that, for example, if the weak states mechanism was truly an explanation for the harmful effects of natural resources on peace, then the “relative location of resources should not matter” (page 16). She hypothesizes that the only occasion in which there will be an effect of natural resources on conflict is if those resources are located in a conflict zone; “those located outside the conflict zone should have a different or no effect on conflict” (page 17). Lujala’s research uses three main types and sources of data. She uses the PETRODATA dataset for her crude oil and natural gas reserves data, which collects from regions worldwide. This dataset allows the author to account and control for the spatial and temporal overlap of resources and conflict (page 18). For the duration analysis, she uses the dataset to construct sec conflict-specific dummies, and for onset analysis, she constructs dummies that coded at the

10

country level. When testing the more lootable resource of the two, Lujala uses the DIADATA dataset to construct dummy variables for secondary diamond production, primary diamond production, as well as a gemstone production dummy. These are all coded at the conflict and country level (page 19). In terms of Lujala’s dependent variables, she constructs two, both using the UCDP/PRIO Armed Civil Conflict dataset. The benefit of this dataset is that is “has a relatively low inclusion rate, implying that low-intensity conflicts are included as well” (Lujala, 2010, page 19). For her duration analysis, she collects data on 252 conflicts. For onset analysis, she is able to collect 238 conflicts; several conflicts are excluded from the second analysis, as there were cases where a conflict would begin in the same year as another conflict in a given country (page 19). Her control variables ranged from income level, logged population, social fractionalization, level of democratization and regions with rough terrain or forest cover. The methodology used in this paper followed a continuous probability distribution, namely the Weibull distribution. This method is preferred to its alternative, the Cox model, as it “reports the most conservative coefficients” (Lujala, 2010, page 21). Table II in the Appendix presents Lujala’s results for the bivariate duration analysis of armed civil conflict. From the first row, it is clear that conflict duration is more than doubled in conflict areas consisting of oil reserves relative to conflict areas without oil reserves. Considering oil production, conflict is again increased by a factor of about 1.8 in regions that have conflict relative to regions with no conflict. Gas reserves also have an adverse effect on conflict in conflict areas, but the same effect is not found for oil production. This may be explained by the fact that rebels may count this as future value that they work towards expropriating. Another finding from Table II is that both secondary diamond and gemstone production increase conflict

11

duration, whereas primary diamonds do not enter significantly. This may be explained by the fact that primary diamonds are not considered lootable given the difficulty of their expropriation. When secondary diamonds and gemstones are combined into one variable, their effect is even larger and more statistically significant. Table III presents Lujala’s findings for the duration of armed conflict. When including dummies for hydrocarbon reserves, as well as secondary diamonds and gemstones, the results follow closely those of Table II. The effects of the resources is highly significant and more than doubles conflict duration. Model 1 also includes certain control variables, such as rough terrain regions, forest cover, and rainy season, all which enter significantly into the model. The argument here is that these measures “benefit rebels by providing them with hiding places and causing natural breaks in fighting” (Lujala, 2010, 23). Model 2 includes a measure for intensity, which accounts for conflicts that cause relatively high causality rates, which entered significantly into the model as well. Model 3 includes the level of democracy, which entered significantly and suggests that “democracies tend to fight longer wars,” and that a possible explanation for this is that they are “less likely to use overly brutal methods to bring a rebellion to an end” (page 23). Model 4 seeks to determine the effect on secondary diamonds separately, and finds that is it statistically significant. This makes sense, as secondary diamonds are relatively more lootable than other resources. In model 5, hydrocarbon production loses its significance, whereas the effect of reserves on conflict duration still holds. Model 6 tests the effect of oil reserves specifically, and finds a weaker, but still present effect on conflict duration. Lujala (2010) holds that the mere presence of gas and oil may lengthen conflict duration and that production is clearly not necessary for the effect. Model 7 shows that the effect of oil production is still present, but very weak. Model 8 accounts for the fact that there may be an effect being picked up

12

by resource rich countries that generally tend to have longer conflict duration (page 23). By testing all the dummies at the country level, it is shown that the effect of the resources on conflict duration is only present when the resources are in conflict regions, which supports the argument that “rebels prolong conflict duration through their movements” (page 23). Finally, Table III presents Lujala’s findings on onset analysis, where she distinguishes between onshore and offshore regions when considering the difference in risk of conflict onset. Model 9 shows that the democracy enters significantly into the analysis, and specifically that “the most democratic and autocratic countries are likely to experience conflict” (page 24). Other control variables that enter significantly are linguistic fractionalization, mountainous region, and secondary diamond production. Specifically, countries with secondary diamond production increase the risk of conflict onset by a factor of almost 1.5. Model 10 introduces the oil production dummy, and finds that the effect is substantial and statistically significant. This translates to oil production also increasing the risk of conflict onset by a factor of almost 1.5. Model 11 seeks to differentiate between onshore and offshore production regions, and yields results that suggest that the only effect is observed in onshore production; offshore production suggests no effect on conflict onset. Lujala notes then that rebels rarely have access to offshore production regions, and that “the only way in which offshore production can influence conflict onset” is by means of its effects on state institutions, or by the weak state mechanism (page 24). Models 12 and 13 account for former British and French colonies dummies, which then cause the oil variables to lose their significance (page 24). In conclusion, Lujala uses a new method to better characterize the relationship between natural resources and conflict. She does this by considering how the location of natural resources and rebel access affect both conflict onset and duration. Her results suggest that resources that

13

are located within actual conflict zones cause conflict duration to be doubled. When testing the variables at the country level, the effects disappeared, strengthening her hypothesis that the effect is only present in conflict regions. She also finds that production is not necessary for an effect of resources on duration to be present, and that the mere presence of reserves is enough to make an impact. For Lujala’s onset analysis, she found that only onshore production had a harmful effect on conflict onset, also through its rebel movements. Countries with secondary diamond production are also at a higher risk of conflict onset. These results suggest that rebel motivations and greed are more prominent as mechanisms underlying the link between natural resources and conflict than are other mechanisms that work through state or political institutions. 2.3. “A Diamond Curse? Civil War and a Lootable Resource” This final study, conducted by Lujala, Gleditsch and Gilmore (2005), also examine the relationship between natural resources and civil war, but do so by looking at the specific type of natural resource in question. The authors consider one key resource, diamonds, as they have “emerged as a prominent factor in explanations of civil war” (page 538). The authors apply a new method not seen before in previous literature, and disaggregate the resource into its lootable and non-lootable form, secondary diamonds and primary diamonds, respectively. As mentioned earlier, the argument that secondary diamonds are lootable than are primary diamonds is due to their relative location to the surface, and thus the ease at which they can be extracted. For their research, the authors considered the time period of 1945-1999. The considered regions with diamond discoveries and production as provided to them by the dataset they selected. This narrows their research down to 25 countries with primary diamond discoveries, 17 of which are producers. In addition, the authors use another 32 countries that have reported secondary diamond discoveries, 26 of which are producers.

14

As with Lujala (2010), the theory behind this paper lies in the rebel greed hypothesis. The authors note, “the general argument has been that abundant natural resources provide a pool in which rebels can acquire a stake to finance their warfare” (page 539). Referring to the three factor model of rebellion, which was largely inspired by Gurr in 1970, Lujala, Gleditsch and Gilmore (2005) note that rebels movements are decided based on motivation, opportunity and identity; specifically, the rebels need to feel either grievance or greed, need to be free of barriers to achieve their goal, and they need a common identity for group formation (page 539). In addition, previous scholars have suggested that “natural resource abundance may increase the risk of conflict onset,” mainly because “rebels can loot primary product commodities to finance their fighting” (page 540). Lujala, Gleditsch and Gilmore (2005) also note that in 2002, Addison, Le Billon and Murshed found that the relationship between natural resources to conflict is dependent on the lootability of the resource (page 541). The authors hypothesize that a country producing diamonds is more prone to civil war outbreak. This hypothesis has two variants: primary diamonds do not affect the risk of civil war onset, whereas secondary diamonds cause the risk to be higher. Another hypothesis the authors make that countries with secondary diamonds are associated with a higher incidence of conflict, whereas countries with primary diamonds are associated with a lower incidence of conflict. A third hypothesis is that “the presence of secondary diamonds is positively associated with the onset and incidence of civil war in countries with high ethnic fractionalization” (Lujala, Gleditsch & Gilmore, 2005, page 545). The fourth is that poorer countries experience a stronger effect of secondary diamond mining on their risk of civil war than do relatively richer countries. Lujala, Gleditsch and Gilmore (2005) used the DIADATA dataset to collect their information on diamonds. The dataset includes 23 countries across the world, and as mentioned

15

earlier, the data is divided into primary or secondary diamonds, and the data also includes geographic coordinates (page 546). The authors also note that they exclude countries that have sporadic diamond occurrences, such as in Nigeria. The methodology used by Lujala, Gleditsch and Gilmore is taken from a previous study by Fearon and Laitin in 2003, which studied the onset of civil war in relation to natural resources. The authors use a logit model, which is modeled as the following: y* = β0+β1x1+β2x2+β3x3+ε Whereby the dependent variable, y*, indicates whether or not there was a war. It takes value 1 if there was, 0 if not. β1 includes the diamond dummies; β2 includes various control variables; β3 includes a set of variables to control for time dependence; ε is the associated error term. The control variables in this paper are also taken from Fearon and Laitin (2003): income per capita, logged population, rough terrain, petroleum abundance, recent independence, instability, ethnicity, control for time dependence, and the number of years of peace before onset of conflict. When conducting a bivariate analysis between civil war onset and diamond presence and production, the authors constructed six dummy variables. This can be seen in Table 5 of the appendix. One was for aggregated diamond presence, the second for aggregated diamond production, the third for primary diamond presence, the fourth for primary diamond production, the fifth for secondary diamond presence and the sixth for secondary diamond production. They found that the aggregated dummy variable entered significantly, translating to all diamond production and presence contributes to civil war onset. However, primary diamonds are no longer significant after disaggregation, and only secondary diamonds remain highly significant.

16

This confirms the authors’ hypothesis that only secondary diamonds have an impact on the risk of civil war, and that primary diamonds do not. When testing for the onset and incidence of all civil war, the authors initially find that higher income levels decrease the likelihood of conflict onset, whereas a larger population size, mountainous terrain, dependence on oil exports and political instability all have an adverse effect on conflict. Ethnic and religious fractionalization does not enter significantly into the model. The authors report that there is no relationship found between aggregated or disaggregated diamond production and the onset overall, they argue that this is an issue related to the selection of the dependent variable. When testing for conflict persistence, the authors now find that the dummy for ethnic fractionalization now becomes significant. When adding an interaction term that combines secondary diamond production and ethnic fractionalization, the term yields a significant result. This result suggests that “secondary diamond production in ethnically heterogenous countries tends to lead to more persistent conflicts while in more homogenous countries it does not (page 552). This confirms the hypothesis linking secondary diamonds to conflict in countries with ethnic fractionalization. In their third regression, when testing for the onset and incidence of ethnic civil war, as noted above, it is suggested that secondary diamonds are associated with conflict in ethnically heterogenous countries. In this analysis, religious fractionalization is no longer significant, but the rough terrain and instability measures enter significantly. When testing the aggregated diamond variable, it resulted insignificant; when disaggregating, primary diamonds remained insignificant, while secondary diamonds entered significantly. The authors also tested the effect of diamond production on poor countries relative to rich countries. They found that poor countries are at risk of conflict if they have secondary diamond production, but that the “effect

17

evaporates when an interaction term for secondary diamond production and ethnic fractionalization is added” (page 553). In poor countries, there is no impact made by the production of primary diamonds on the incidence of conflict. In addition, it is noted by the authors that primary diamond production decreases the risk of ethnic conflict onset. In general though, the authors find that the production of secondary diamonds increases the probability of the incidence of ethnic war by more than 200%, while the production of primary diamonds decreases the probability by 80% (page 556). Finally, the authors note that they cannot “neglect the possibility that the diamond dummies are picking up an Africa effect,” whereby more than half the countries with diamond deposits or production are located in Africa, which is overrepresented when it comes to conflict (page 558). When running their analysis with a dummy for sub Saharan Africa, the results remain robust. Another sensitivity analysis conducted is to control for the possible effects of colonialism; when the authors include dummies for British and French colonies, the results remain robust as well. In conclusion, Lujala, Gleditsch and Gilmore (2005) find that the type of resource matters when conducting analysis on the relationship between natural resources and civil conflict. They found substantial support for a strong link between secondary diamonds to civil conflict, but no support linking primary diamonds to conflict. This is in accordance with the idea that secondary diamonds are more lootable as compared to primary diamonds. In addition, the authors found that the presence of diamond deposits does not have nearly as big of an effect as the actual production of the resource—again, this effect differs between primary and secondary diamonds. This paper highlights the importance of disaggregating natural resources with respect to their lootability, as to avoid obtaining biased results (page 559).

18

3. Discussion and Conclusions The three empirical works reviewed in this paper all aimed to better specify the relationship between natural resources and the propensity for conflict onset and duration. The study by Humphreys (2005) directed focus away from the common rebel greed hypothesis, by aiming to shed light on several other possible mechanisms that underlie the link between natural resources and conflict. Although he was unable to test all the mechanisms and although there were data and methodology issues, he provided sufficient evidence that the consideration of other mechanisms could prove to benefit conflict scholars in their search for answers regarding relationships such as this one. Further, Humphreys notes that research undertaken with several mechanisms in mind could lead to more promising and effect policy reforms and other changes in government that reduce grievances and conflict in general. The second paper was in a way a rebuttal to Humphrey’s analysis. Lujala (2010) accepted that there may be other mechanisms at work, but specified that if certain factors were controlled for, such as the location of lootable resources and rebel access to them, then the true effect could be seen. If other rival mechanisms were truly the appropriate explanations for why conflict occurs in resource abundant countries, and not the rebel greed hypothesis, then there would be no observable effect in tests considering location of resources and rebel access. Her research, however, provides results supporting that location and rebel access are crucial in identifying a link between natural resources and conflict. Furthermore, she finds that production of natural resources is not necessary for conflict to arise; the mere presence of oil reserves or gemstones underground is enough to prolong conflict. Finally, Lujala (2005) looks at the relationship once again, this time focusing on the type of resource with respect to its lootability degree. Lujala, Gleditsch and Gilmore (2005)

19

disaggregate the diamond resource into its lootable and non-lootable form, namely, secondary diamonds and primary diamonds. They find that, in accordance with their hypotheses, that secondary diamonds are associated with conflict, whether they are produced or simply discovered underground. In contrast, primary diamonds do not have that effect. In fact, primary diamonds have a softening effect on conflict in some cases, whereby the results suggest that secondary diamonds almost always have an adverse effect on conflict. What is interesting is the finding in Lujala, Gleditsch and Gilmore (2005) that the presence of diamond deposits does not have nearly as big of an effect as the actual production of the resource. This follows what is found in Humphreys, who argues that reserves of natural resources do not provoke conflict by means of rebel greed. However, a few years later in her 2010 study, Lujala disproved this by showing that the mere presence of oil reserves and gemstones was enough to lengthen conflict duration and that production was not necessary. As noted by the authors themselves, conducting further research while using more finegrained data, better measures for variables and indicators, and disaggregating variables into logical forms would help provide more significant, thorough and reliable results. This would help research by providing clear results, as opposed to the ambiguous findings currently found in existing literature.

20

4. Appendix

Table I: The Political Economy of Extraction (Global Sample) I Measure of natural resources Measure of Strengths/Weaknesses Natural Resources

Strengths/Weaknesses

Interaction Term

Observations

II

III

Oil Production (Per Capita)

IV

V

VI

Oil Reserves (Per Capita)

Diam

b

Weberian

Instability

Strong

2.415

8.537

-89.911

0.336

0.956

1.603

1.11

(3.94)***

(3.36)***

(1.70)*

(5.67)***

(1.25)

(0.16)

(4.27)**

0.595

-0.466

-0.056

0.646

-0.51

-0.051

0.618

(2.39)**

(1.67)*

(0.72)

(2.57)**

(1.80)*

(0.66)

(2.49)*

10.978

-5.313

18.572

0.974

-0.588

1.368

-0.004

(1.82)*

(1.93)*

(1.93)*

(0.43)

(0.76)

(1.24)

(0.01)

5,170

5,167

1,339

5,170

5,167

1,339

5,170

Instability

Strong

b

VII

Weberian

a. Computational problems encountered in estimating equation IX. b. “Strong” is given by (1-Instability)x(1-Anocracy) *Significant at 10%. **Significant at 5%.***Significant at 1%. Source: Humphreys, Macartan (2005) ‘Natural Resources, Conflict and Conflict Resolution: Uncovering the Mechanisms,’ The Journal of Conflict Resolution 49, 508-537

Table II: Bivariate duration of analysis of armed civil conflict, 1946-2001 Independent Variable Time ratio Oil reserves, conflict zone 2.171 Oil production, conflict zone 1.790 Gas reserves, conflict zone 1.687 Hydrocarbon reserves, conflict zone 2.556 Hydrocarbon production, conflict zone 1.752 Secondary diamond production, conflict zone 1.939 Gemstone production, conflict zone 4.667 All gemstones, conflict zone2 3.164

p-value 0.009* 0.082* 0.044* 0.001* 0.088* 0.017* 0.000* 0.000*

*p