Natural Resources and the Dynamics of Civil War. Duration and Outcomes

Natural Resources and the Dynamics of Civil War Duration and Outcomes Hirotaka Ohmura Faculty of Economics, Shiga University h-ohmura@biwako.shiga-u.ac.jp March 2012 Working in Progress. Please do not cite without permission. Abstract How do natural resources affect the dynamics of civil war duration and outcomes? While numerous studies have attempted to examine the relationship between the duration of civil war and natural resources, they fails to find a consistent empirical evidence for the alleged links between abundant resources and prolonged civil war. In this paper, to modify the existing explanation for the link between natural resources and civil war duration, I introduce three important factors into the analysis and examine them. The first focus is disaggregating natural resources into lootable and non-lootable resources, and comparing the effect of each natural resource on civil war process. In the second, this paper takes into account the types of conflict termination. Finally, the relative capacity between a government and rebel before starting conflict is considered. In an empirical part, this paper examines the effect of natural resources on the dynamics of civil war duration and outcomes during the period 1946-2003, using a competing risks model approach. Findings from quantitative analysis reveal that the effects of natural resource on duration vary across the types of conflict ends: government victory, rebel victory, and negotiated peace settlement. 1

Introduction How do natural resources affect the dynamics of civil war duration and outcomes? In recent years, increasing attention has been devoted to understanding the relationship between the onset and duration of civil war and natural resources (e.g. Collier and Hoeffler, 1998, 2004; Collier, Hoeffer, and Soderbom 2004). 1 The greed hypothesis suggested by Collier and his colleagues argues that resource wealth increases the motivation of rebels for accumulating private gain and provides rebel leaders enough funding opportunities to continue rebel activity, thereby increasing the likelihood of civil war (see also Ross, 2003, 2004a). 2 While a number of subsequent studies (Buhaug, Gates, and Lujala 2009; Buhaug and Lujala 2005; Fearon 2004; Humphreys 2005; Lujala, 2010; Ross 2006) have attempted to examine that abundant natural resources tends to strengthen the feasibility to continue an insurgency and lengthen the duration of civil war, there is no consensus in terms of empirical results. I believe that the reason no consensus has been reached in the empirical analyses that have examined the effect of natural resources on civil war duration is that they have paid little attention to civil war outcomes. Every civil war eventually ends, but it ends due to different reasons: military victory by a government or rebel group, negotiated peace settlement, or some other type of outcome. 3 Mason and Fett (1996) and Mason, Weingarten, and Fett (1999) indicate that the manner in which a civil war terminates 1 Natural resources can be broadly classified as renewable (e.g. water or agricultural products) and non-renewable resources (e.g. energy or mineral resources). In this paper, natural resources indicate only non-renewable resources. 2 Humphreys (2005) and Ross (2004b, 2006) provides a comprehensive review of causal mechanisms and empirical findings on the link between natural resource wealth and civil war. 3 Since the trend of civil war outcomes after WWII shows that negotiated peace settlements are the dominant type of termination after the Cold War, existing studies on civil war outcomes tends to investigate the determinants of successful peace negotiation (e.g., Walter 1997, 2002). 2

depends on its duration. More importantly, since some factors have different (or oppositional) effects on different outcomes, pooled analysis may lead to biased estimates (Brandt et al. 2008). For example, additional resources on the government side reduce the duration that a government achieves a decisive military victory, but positively related to the duration that a civil war ends in a rebel victory. Since existing research on resource wealth and civil war duration do not consider the aspect that when civil war ends is related to how it ends, and that some covariates have different effects varying across termination types, they cannot claim to have fully explained the actual influence that natural resources have on civil war duration and outcomes. 4 In order to address the relationship between natural resources and the dynamics of civil war duration and outcomes, this paper simultaneously examines the effect of natural resources on civil war duration and outcomes using a competing risk model approach. This article is structured in the following manner. The next section reviews existing empirical literature on natural resources and civil war duration. The third section explains how the lootability of natural resources affects the duration and outcomes of civil war. In this section, I also argue the importance of relative capability of rebels to a government to consider the effect of resource wealth on the consequence of civil war. The fourth section outlines the research design for conducting empirical study, including data, variables, and method. The fifth section presents several quantitative tests and interpretation of the results. The last section concludes the paper by summarizing the main findings of this paper and suggesting improvements for future research. 4 Humphreys (2005) is only one exception that examines the effect of natural resources on civil war duration and outcomes. However, most studies have paid little attention to civil war outcomes. 3

Previous Studies of Natural Resources and Civil War Duration Why are resource conflicts believed to last longer than others? How do natural resources prolong the duration of civil war? 5 After the seminal works of Collier and Hoeffler (1998, 2004), numerous scholars have attempted to investigate the relationship between natural resources and duration as well as onset of civil war. Whereas theoretical explanations regarding the linking mechanism of natural resources and conflict duration are occasionally conflicting, there is a basic consensus that civil war in resource-rich countries, on average, is likely to be longer than one in resource-poor countries. However, the empirical results of quantitative studies are mixed at best. Collier et al. (2004) conduct statistical analysis for 55 civil wars from 1960 to 1999. In the empirical analysis, they use the ratio of primary commodity exports in gross domestic product (GDP) as a proxy for resource abundance or dependence 6 ; contrary to expectation, they find that there is weak or statistically insignificant relationship between the primary commodity exports and civil war duration. They also find that decline in primary export prices are negatively related to civil war duration. Since primary commodity export variable, which Collier and Hoeffler use in empirical analysis, lumps several categories of natural resources together such as mineral, energy, and agricultural resources; drugs, diamonds, and other gems that are believed to be the most influential resources for greedy rebels are not included. Other empirical studies believe that treating all natural resources as the same is not a better way; thus, they 5 Humphreys (2005) and Ross (2004a, 2006) offer several possible causal mechanisms for explaining the reason why natural resources prolong the duration of civil war: the feasibility mechanism, the conflict premium mechanism, the balance of power mechanism, the fragmented organizational structures mechanism, the possibility of pork mechanism, the international conflict premium mechanism, and the sparse networks mechanism. 6 Sachs and Warner (1995) originally use the rate of primary commodity export in GDP to examine the effect of natural resources on economic growth. 4

examine the effect of specific resources or specific types of resources on civil war (Ross 2004b). Most research tends to focus on three relevant categories of natural resources, namely, oil, gemstones (including diamond), and drugs, instead of the primary commodity export variable. Fearon (2004) examines 128 civil wars that occurred in the period 1945 1999 and shows that rebel forces mainly obtain financial resources from contraband goods such as opium, coca, and gems, thereby prolonging the duration of war. Buhaug and Lujala (2005) and Regan and Aydin (2006) also find that gemstones, opiates, and coca are related to longer war. Fearon and other scholars conclude that the easily accessibility of resource wealth leads to a longer civil war as it increases the feasibility of rebel activities. Since it is difficult for rebels to organize their own groups and continue an insurgency without adequate financing to procure arms and belligerents, wealth from abundant natural resources increases the feasibility of rebellion as it provides rebel leaders with sufficient funding. Some studies (Le Billon 2001; Lujala, Gleditsch, and Gilmore 2005) indicate the importance of geographical characteristics of natural resources and resource s lootability to determine who can obtain certain resources. Since natural resources are natural, factors such as where they are mined, how difficult the production is, and how easily the transportation is cannot be artificially determined. In order to provide rebels with the feasibility of armed conflict, natural resources must be something that can be looted by rebel groups. Therefore, even if resources are abundant, they do not influence the duration of civil war if they cannot be or are difficult to be looted by rebel groups. On the other hand, because non-lootable resources are likely to be controlled by a government, such resources strengthen the power basement of government. 5

In addition, natural resources in the conflict zone provide rebel groups with greater opportunities to access resources during a conflict. Sufficient funding opportunities from abundant resources enable rebel leaders to procure arms and recruit belligerents with the aim of disempowering a government. The empirical results of Lujala (2010) and Buhaug et al. (2009) support the argument that the lootability (or accessibility) of natural resources increases the feasibility of insurgency and lengthens civil war duration. According to Lujala (2010), lootable resources such as gemstones and secondary diamonds and non-lootable resources such as oil and natural gas in conflict zones are strongly related to longer duration of conflict. Further, Buhaug et al. (2009) also show that gemstones and petroleum production in conflict areas prolong conflict duration, but that drugs are not related to conflict duration. Keen (2000) argues that lootable resources can stimulate rebels incentive or motivation for continuing civil war as well as increase the feasibility of fighting. In wartime, individual rebels can more easily continue to commit criminal activities (looting or illegal commerce of spoils) in order to accumulate personal gains than they can do in peacetime. As mentioned earlier, rebels initiate an insurgency against the government in order to obtain private wealth; therefore, favorable benefits in wartime provide a rebel group with incentive to prolong the duration of conflict (Addison, Le Billon, and Murshed, 2002). Lujala (2010) interprets the result that lootable resources are positively and significantly related to duration as evidence that rebels prefer the continuation of wartime to a decisive military victory or negotiated settlement (i.e., civil war termination). However, some studies show the results that natural resources are not related to civil war duration. Ross (2006) disaggregated oil and diamonds into offshore oil and 6

onshore oil and primary diamonds and secondary diamonds on the basis of the resource lootability and finds that there is no statistically significant relationship between lootable or non-lootable resources and civil war duration. Montalvo and Reynal-Querol (2010) also show that oil resources have a weak or no significant effect on the incentive of continuing an insurgency. More importantly, Humphreys (2005) examines the effect of oil production and reserves and diamond production on the duration of civil war; he argues that natural resources production is related to shorter civil war. He interprets the results as implying that when the stronger side obtains natural resources, the duration of conflict becomes shorter than before. 7 Humphrey s argument regarding the balance of power between rebels and the government indicates that when rebels or governments obtain resource wealth, they can strengthen their relative power (e.g., purchasing arms and recruiting soldiers) by obtaining resources, thereby changing the power equation in favor of the side that has greater access and control over natural resources. The empirical evidences for a causal connection from resource abundance to conflict duration is less consistent than the theoretical argument. As mentioned above, existing studies have attempted to solve the resource-conflict duration by examining how different resources (oil or diamonds) and different types of resources (lootable or non-lootable resources) are linked to the duration of civil armed conflict. However, these empirical studies tend to focus solely on the resource-war duration link and do not pay enough attention to termination types. While lootable resources increase the feasibility of rebels, non-lootable resources can reinforce the power of government. Lootable (non-lootable) resources may different effects on the duration that a civil war ends in rebel victory or government victory. Additionally, since resource wealth 7 Of course, this mechanism also implies that natural resources can lengthen civil war duration when the weaker side gains resource wealth. 7

strengthens the side that control and exploit natural resources, the relative capability between a government and rebel before conflict start also play a significant role in duration and outcomes. In order to reveal the relationship between natural resources and civil war duration, we need to take into account the termination types of conflict and the relative capability of rebels to government. Natural Resources and the Dynamics of Civil War Duration and Outcomes Previous studies indicate that abundant natural resources enable the sustenance of insurgency and facilitate rebel activities. Furthermore, rebels find it easy to conduct illegal acts during a civil war; acts such as seizing control of resources and selling them illicitly in peacetime would be criminal. Therefore, the availability of natural resources strengthens the capability of rebels and prolongs a civil war when the resources are easy for rebels to loot and can be used to fund their military campaign. 8 However, these arguments may assume that weak rebels increase the relative capability of the government to win a civil war by looting resource wealth. 9 This is 8 Lootable resources are precisely defined as those that (1) are scattered over a large area (Le Billon, 2001); (2) can be easily explored, mined, and produced by individuals or small groups of unskilled workers (Ross, 2003); and (3) are not controlled under strong regulation for export and traffic of resources (Ross, 2003). Drugs, alluvial gemstones, agricultural products, and timber can be identified as examples of lootable resources. In contrast, oil and natural gas or deep-shaft minerals are clustered in a narrow area and require an enormous amount of investment in resource exploration and mining (e.g., sophisticated exploration technology, considerable mining equipment, and numerous skilled workers). All other things being equal, these non-lootable resources may be more difficult for an insurgent group to pillage. 9 Of course, several scholars argue that the government tends to be stronger than the rebels when a civil war starts (Mason and Fett 1996; Mason et al. 1999). It would seem to be appropriate that considerable research tends to assume the initial condition. Nevertheless, there are some cases that the power balance between the government and the opposition is roughly equal or the relative power of rebels to government is high (Cunningham et al. 2009; Gent 2008). These observations suggest that considering the war conditions other than the case of weak rebel and strong government gives us the 8

because when strong rebels obtain sufficient funding opportunities to enhance their military capability, they can achieve military victory in a shorter duration than before. The effect of lootable resources on duration can vary across the capability of rebels relative to the government. Contrary to lootable resources, non-lootable resources, such as offshore oil or primary diamonds, provide a government with sufficient financial resources to reinforce its power. If natural resources change the balance of power in favor of the government, the duration of civil war is shortened. However, in the case of a weak government, non-lootable resources make the balance of power equal and this lengthens civil war duration. This may indicate that the effect of natural resources on duration may depend on the relative capability of rebels and the government prior to conflict. It is thought that a change in the relative power of rebels (government) as compared to the government (rebel) considered as one of the most important factors determining civil war duration and outcomes (Mason and Fett 1996; Mason et al. 1999; Cunningham et al. 2009). Greater rebel strength is significantly related to longer duration of conflict and is more likely lead to a rebel victory (Cunningham 2009). Some previous studies on third-party intervention examine the relationship between biased intervetion and civil war duration and outcomes (Balch-Lindsay et al. 2008; Gent 2008). They explain that a third-party intervention on the side of the government or rebels can increase the likelihood of military victory for supported side because supports from external actors increase the availability of political, military, and economic capability and change a power balance in favor of the supported side. The effects of biased intervention on conflict duration vary across the power balance between government and rebel and how crucial implication for the relationship between natural resources and the civil war duration and outcomes. 9

civil war ends. The idea of above research is closely related to the argument of this paper. The power relationship between rebel forces and the government can be simply classified into three patterns: (1) weak rebel forces and a strong government, (2) power evenly balanced between rebel forces and the government, (3) strong rebel forces and a weak government. If the balance of power does not change during the course of the civil war, it is very likely that the stronger side will win and end the war. On the other hand, if power is evenly balanced, then hostilities are likely to be prolonged. Therefore, the relationship between civil war duration and natural resources is not solely determined by the balance of power before conflict, but also fluctuates depending on termination types. Depending on both the power balance between rebels and the government and civil war outcomes, one of the following situations could exist at any given time. Case 1: Weak rebels and strong government: If a country with abundant natural resources can be easily controlled by rebel forces, these resources will constitute a source of funds for the rebels and will sustain their activities and organization, thereby augmenting their power. Therefore, power balance between the rebel forces and the government will become more evenly balanced. In this case, the availability of lootable resources, which are easy for rebel forces to control, increases the probability of victory for rebel forces and decreases the probability of government victory. Under equal balance of power, government and rebel have different incentives. Rebels continue insurgent activities and avoid military defeat in order to accumulate private gains in wartime. However, because power is now more evenly balanced than before the acquisition of resources, the hostilities last longer and the government s accumulated 10

costs of fighting increase; thus, the government has an incentive to conclude a negotiated peace agreement. On the other hand, the advantages afforded by abundance of natural resources that are difficult for rebel forces to control (and easy for the government to control) further bolster the government s power base and tip the power balance further away from the rebel side. In this case, the existence of non-lootable resources that are difficult for the rebel forces to control reduce the probability of the rebel forces victory and thus increase the probability of the government s military victory. Fearon (2004) explains that the longer civil war endures, the greater the likelihood that it ends with a negotiated civil war settlement. According to Fearon s argument, when the government is much stronger than rebels, the government cannot offer a rebel group any credible commitments to implement peace agreements even though the government would prefer a negotiated peaceful settlement to the costs of war. However, resource wealth provides the government with greater leverage over the terms of negotiation and more incentive to reach a negotiated agreement quickly (Balch-Lindsay et al. 2008). In sum, when the relative capability of rebels to government is low, lootable resources prolong the duration of civil war and reduce the duration that a conflict ends in a negotiated settlement. Under such situation, non-lootable resources increase the likelihood of government victory or peace agreement. Case 2: Evenly balanced power: If power is evenly balanced between rebel forces and the government, the outcome of civil war depends on whether the natural resources abundant in the country can be easily appropriated by rebels. If the rebel forces can easily control the resources, they are able to use the advantages afforded by the resources to change the course of the war in their favor. On the other hand, if the 11

resources are easily controlled by the government, the probability of the government achieving a military victory increases. When the government s military ability is strong in the battlefield, rebels simply go into hiding and continue guerrilla activity. Case 3: Strong rebels and weak government: In this scenario, the availability of natural resources has an impact that is opposite to that in the previous case. If abundant natural resources that can be easily controlled by the rebel forces exist, the already strong rebel forces become stronger and it s even easier for them to defeat the government forces. Strong rebels can use sufficient funding to defeat a government quickly because the winning side can control the government, wealth from natural resources, and resource areas following a military victory. However, since rebel groups have to first overthrow former government regime and then establish a new regime to run the country, rebels are reluctant to do costly works even though they can obtain greater wealth than that in wartime. Collier (2000) and other scholars argue that rebels prefer a wartime situation to outright military victory because they can easily accumulate private gains. Therefore, it is ambiguous that lootable resources lead to shorter duration of conflict. When there are abundant non-lootable resources in civil war countries, resource wealth can generate an almost equal relationship between the government s military capability and that of the rebels. As is the case 1, the equal power balance lengthens civil war duration and expected war costs; therefore, the government is more likely to seek to settle a conflict through peaceful means. As can be seen, the consideration of not only the duration of a civil war but also the manner in which it ends and the power balance between government and rebels clarifies that the availability of natural resources does not influence the duration of all civil wars 12

in the same manner. Previous research may be considered insufficient since power relationship and termination types have not been considered in the analyses of these studies. The availability of lootable or non-lootable resources is most likely to be inversely proportional to the war s duration depending on which party wins. If an analysis does not take into account this assumption or the outcome of the civil war, then the resource effect is not likely to be included in the results of the analysis. Data, Variables, and Method Data In order to test the relationship between natural resources and civil war duration and outcomes, I use the Nonstate Actor Conflict (NSA) data (Cunningham et al. 2009) to identify a sample of 288 civil war episodes for the period 1946 2003. Cunningham et al. (2009) incorporate the information on nonstate actors involved in civil war into the Uppsala/PRIO Armed Conflict Data (ACD) (Gleditsch et al. 2002). Similar to ACD, NSA data employs the 25 battle-death threshold rule to identify the samples. 10 In order to investigate the effect of natural resources on duration and outcomes, this dataset is preferable to others for two main reasons. First, the most important features and usefulness of NSA data is the information on the duration and outcomes of each government-rebel dyad. While there are multiple government-rebel dyads in the same civil war, most previous studies have not disaggregated rebels into each rebel and not collected the data pertaining to each government-rebel dyad. Second, relating to the above, NSA collect the data on the relative strength of each rebel to the government. 10 The UCDP/PRIO dataset defines intrastate war as a contested incompatibility which concerns government and/or territory where the use of armed force between two parties, of which at least one is the government of a state, results in at least 25 battle-related deaths. 13

The NSA data measures rebel strength against that of the government in terms of four different criteria: clear central command, mobilization capacity, ability of rebels to procure arms, and fighting capacity, and rebel strength relative to that of the government is measured on a 5-point scale: much stronger, stronger, equal, weaker, and much weaker. Ideally, to examine the theoretical argument in this paper, I need to conduct the statistical analysis for three separate cases: (1) strong rebels, (2) equaled power, and (3) weak rebels. However it is difficult to have sufficient observations to conduct statistical analysis for the former two cases. This paper, therefore, focuses on only the case of weak rebel and strong government. Dependent Variable In this research, the dependent variable is duration and civil war outcome that are based on the NSA dataset. NSA defines civil war outcomes following the definition of Kreuz (2010). NSA data distinguish civil war outcomes into four categories: government victory, rebel victory, negotiated settlement, and low activity/other. Here, government victory is defined as a decisive defeat by the government of a rebel group in the battlefield. Rebel victory is defined as the defeat of the government s military forces in the battlefield by rebel groups, followed by the subsequent overthrow of the incumbent regime and the establishment of a new regime. Moreover, rebel victory includes secessionist civil wars in which a rebel group secedes from the government and builds a new state or autonomous region in the territory that was controlled by the defeated government. Further, a negotiated settlement is a formal treaty in which a government and rebel group agree to stop fighting or promise to cooperate with each 14

other in a new political order with a formal peace agreement or a ceasefire agreement. Low activity/other is when conflict activity continues but does not reach the UCDP threshold with regards to fatalities or other conflict activity that does not fulfill the criteria of UCDP. The analysis employ the former three types of civil war termination; negotiated peace agreement, government victory, and rebel victory because the outcome low activity and other may be difficult to be theoretically distinguished from ongoing war. Therefore, this paper examines the relationship between natural resources and three civil war outcomes. There were 69 government victories, 62 rebel victories, and 46 negotiated settlements between 1946 and 2003. Therefore, conflict duration is the time between the beginning of fighting and civil war termination by a negotiated peace settlement or decisive military victory by a government or rebel group. When a civil war has not ended, the duration of civil war is right-censored as of December 31, 2003. Independent Variables This article analyzes how abundant natural resources affect civil war duration and outcomes. I use the three categories of natural resources, oil, diamonds, and drugs. Data on natural resources are from PETRODAT (Lujala, Rød, and Thieme 2007), DIADATA (Gilmore et al. 2005), and Lujala (2010). These datasets cover the period 1946-2003. PETRODATA and DIADATA include the geographical information on resource production and reserves, and oil and diamonds can disaggregated into sub-categories (onshore or offshore oil; secondary or primary diamonds) based on their lootability. Oil and gas are located either in narrow onshore or offshore areas. Offshore exploitation requires a large outlay for exploration, production equipment, and a number of skilled 15

workers before any extraction can begin. Therefore, offshore oil and gas are far more difficult for rebels to exploit and pillage. Secondary (alluvial) diamonds are easily accessed and exploited by unskilled workers and insurgent groups and they are easily smuggled to other areas due to porous regulation regimes. On the other hand, primary (kimberlitic) diamonds require expensive extraction equipment and sophisticated skilled labor to be extracted, and most of these diamonds are under the Kimberly Process. Since disaggregating diamond resources into primary (kimberlitic) and secondary (alluvial) diamonds enables us to better understand the impact of lootability on the duration and outcomes of armed conflict, I separately examine the effect of primary and secondary diamonds production. PETRODATA also provides us the geographical information on oil production and reserves in onshore and offshore areas. However, oil production is strongly related to conflict duration when the oil is in conflict zone or rebels fight over oil areas (i.e. secessionist conflict) regardless of onshore or offshore areas (Buhaug et al. 2009; Lujala 2010). Therefore I examine the relationship between oil production and the dynamics of civil war in the analysis. The resource production variables are coded 1 for the first production year and for each year thereafter. Control Variables Once civil war begins, several factors may influence whether the war will last a long time or how it will end. First, I take two measurements related to the relative capability of rebels and the government. Next, I take the characteristics of the civil war country into account. There are four characteristics economic condition, population size, level of democracy, and ethnic heterogeneity. Further, I also include two 16

characteristics of conflict: stake of conflict and the number of parties participating in a civil war. All control variables have been taken from Cunningham et al. (2009). First, I incorporate two variables related to rebel strength that are not directly affected by resource wealth territorial control and legal political wing. When rebels have a territorial base, they can easily organize their group and efficiently counter the military threat posed by the government. If the government legally guarantees political activity, rebels are more likely to achieve their demands without resorting to violent means. Next, in order to control for economic conditions, GDP per capita income, which is based on Gleditsch (2002), is included in the empirical analysis. It is argued that countries with a higher level of economic development indicate higher opportunity costs of insurgency. GDP per capita income is expected to have a negative impact on civil war duration. Next, population size is included in the analysis and may be positively related to duration. The level of democracy in a civil war country may have significant influence on the dynamics of civil war duration and outcomes. I use the Polity democracy-autocracy scale to measure the level of democracy during civil war. Finally, in order to control for ethnic diversity, I use an Ethnic Fractionalization (EFL) index, which is measured by the probability that two randomly chosen persons in a country belong to different ethnic groups, and ranges between 0 and 1. The types of civil war affect its duration and outcomes. The stakes in ethnic conflicts involving ethnic, religious, racial, and linguistic issues are less divisible than politico-economic conflicts regarding ideological and socioeconomic interests. Ethnic conflicts are assumed to involve higher stakes; thus, compromise and accommodation are much more difficult to achieve in such conflicts than they are in 17

politico-economic conflicts. Moreover, a greater number of actors participating in a conflict make it more difficult to reach a compromise in the peace process and lengthen the duration of civil war (Cunningham 2006). Method In this analysis, since the dependent variable is the time to multiple events of different types, I employ a competing risk model, which is one method of duration analysis, to estimate the effect of natural resources on civil war duration and outcomes. 11 As mentioned earlier, natural resources can potentially have a different effect on the duration and outcome of civil war. Treating all the outcomes leads to inadequate empirical results when a certain covariate has a positive effect on one outcome and a negative effect on another outcome. For example, this paper assumes that lootable resources shorten the duration in which rebels achieve military victory, but lengthen the duration in which the government gains victory over a rebel group. A competing risks model is appropriate when an event can end in different ways and there are different factors that influence the hazard of the risks (Box-Steffensmeier and Jones 2004; 166). In the analysis, I separately estimate three duration models and treat events other than the one being examined as randomly right-censored in each model. 12 The competing risk model approach assumes that the risks (i.e. outcomes) are conditionally 11 Box-Steffensmeier and Jones (2004) is a good introduction to competing risk model. 12 One of the most important advantages of duration analysis is that it enables an appropriate estimation for censored data. Since the observation of spells is censored on December 31, 2003, the data is right-censored. In other words, at the end of observation, the relevant event (the end of civil war) has not yet occurred (the spell end date is unknown); thus, the total duration of civil war from entry to and exit from the war spells is unknown. Therefore, duration analysis permits scholars to employ a more precise degree of variation in the dependent variable. When we fail to account for right-censored data, the empirical analysis can lead to biased results (Box-Steffensmeier and Jones 2004). 18

independent. There are a number of duration models that differ depending on the assumption of functional forms for the hazard rate. The hazard function is the distribution of the probability of failure across time. If the functional form of the hazard rate can be assumed with reasonable confidence at the outset of the analysis, parametric models should be employed. If there is no theoretical argument regarding the shape of the hazard functions and hazard rates (parameter effects) need to be determined, semi-parametric analysis should be used (Clevens et al., 2004). In this paper, I do not have a good prior reason for using functional forms. Therefore, the Cox model is employed. Results This section discusses the findings from the data analysis. 13 Table 1 presents the results from the four competing risk models for all civil war cases. In Table 2, I focus on only weak rebels cases to examine the effect of relative capability on the relationship between natural resources and civil war process. Table 1 and 2 show the results from the various Cox models for the entire post-wwii period (1946-2003). 14 Table 1 around here 13 All estimations are conducted using STATA 12. 14 Since the Cox model assumes that the hazard rate is proportionate over time, we should test for the proportional hazards (PH) assumption before conducting empirical analysis. The violation of PH assumption leads to substantial and methodological problems (Box-Steffensmeier et al. 2003). The results from the Schoenfeld test for the PH assumption indicate that several variables violate the PH assumption, and thus I adjust the violation using interaction term between the covariates and the log of time. 19

I begin with the result of natural resources and civil war duration without considering the types of civil war termination (pooled analysis). In Table 1, the positive and statistically significant coefficient for primary diamonds in the pooled model indicates that primary diamonds increase the likelihood that a civil war ends more rapidly than in the absence of primary diamonds. In contrast, the negative and statistically significant coefficient for secondary diamonds means that secondary diamonds increase the likelihood of prolonged conflict. The results of drugs and oil variables are not statistically significant related to civil war duration, but the signs are negative and positive. In line with prior findings, the estimated coefficients of resource variables show that lootable resources lengthen the duration of civil war, but non-lootable resources reduce the duration of civil war. Previous research interprets this results that more accessible natural resources, that is, lootable resources, give rebels more financial resources to maintain an armed organization and much incentive to prolong the duration of civil war and accumulate the resource wealth. However, the following empirical analysis which discriminate civil war termination into three separate outcomes shows the effect of natural resources on duration vary across the types of termination. Turning now to the results corresponding to three separate outcomes model, the estimated coefficient of secondary diamonds variable is not statistically significant in the formal agreement model and the government victory model, but surprisingly the variable is negatively and statistically significant in the rebel victory model. Contrary to the expected result, this result suggests that the abundant secondary diamonds, namely lootable resources, lengthen the duration that rebels will archives a military victory. Rebels tend to choose the continuation of civil war because they are more likely to 20

obtain resource wealth than they are in peacetime, even though the probability of rebel victory is higher than before the onset of civil war. Drugs have no significant effect on all civil war outcomes. The primary diamond production variable is negatively and statistically significantly related to the duration that a civil war ends in negotiated peace agreement. The positive and statistically significant coefficient for oil production in the government victory model indicates that oil wealth reduces the duration that the government will achieve a military victory. The interaction between this variable and the natural log of time is negatively signed and statistically significant indicating that the effect of oil production on government victory decreases over time. As mentioned before, non-lootable resources reinforce the government s capability, and thereby increase the likelihood of decisive military victory by government and provide the government with greater leverage in a peace process. The results are supportive evidences for the argument. Table 2 reports the results of resource wealth and civil war duration and outcomes when the relative capability of rebels to government is low. The result of the pooled model is similar to the analysis in all civil war cases. Lootable resources (drugs) positively affect the length of civil war. The primary diamonds tend to shorten the conflict duration. Table 2 around here The results of competing risk model for weak rebel cases yield similar results to Table 1. The primary diamond production does reduce the duration of civil wars that 21

end in peace agreement or government victory. In addition, when rebel s capability is lower than government s one, all civil wars with abundant primary diamonds have not experienced the rebel victory. Therefore, to avoid the perfect prediction, the variable for primary diamonds is excluded from the Rebel Victory Model. Secondary diamonds and drugs have no effect on any of the possible civil war outcomes. While oil production has no significant effect on the duration of either negotiated peace agreement or rebel victory, resource wealth from oil increases the likelihood that a civil war will end more rapidly in a military victory by government. The interaction between oil variable and the natural log of time shows that the positive effect on government victory decreases as a civil war endures. The accessibility of lootable resources enhances the rebel s motivation to amass private wealth by looting resources and increases the feasibility of insurgency. When the military capability of rebels is weaker than the capability of government, rebels can use resource wealth from lootable resources to strengthen the organization's activities for government and continue insurgency. However the empirical results of Table 2 may show weakly evidence that drugs have positive effect on civil war duration, it cannot be said with confidence that lootable resource production has a substantive and significant effect on the dynamics of civil war duration and outcomes. On the other hand, abundant non-lootable resources shorten the duration that a civil war ends in government victory or formal agreement. The strengthened capability enables government to suppress the rebels more rabidly, and the greater difference between government s and rebel s power gives them an incentive for concluding negotiated peace agreements. Table 1and 2 also tells us something about the civil war duration and outcomes. While strong rebels that control certain areas can fight against government longer than 22

rebels without territorial control, territorial control reduce the duration of armed conflict when the relative capability of rebels to government is low. Ethnic fractionalization lengthens the duration that a civil war ends in formal agreement or rebel victory. The results of war issue and economic condition indicate that less divisible issues lead to longer duration that a civil war ends in rebel victory and high opportunity costs of insurgency decrease the probability of rebel victory. Democracy has a positive effect on civil war duration, especially the time until government victory. If civil war countries have more than one rebel, they experience longer conflict than other countries. The coefficient of population is positively related to the duration that a civil war ends in formal agreement or rebel victory. Conclusion This article aimed to examine the effect of natural resources on the dynamics of civil war duration and outcomes. Even though previous research has intensely debated the subject of the resource-conflict duration link, it does not fully account for the possibility that natural resources have different influence on different civil war outcomes. This paper investigated the relationship among lootable or non-lootable resources, the types of terminations, and the power balance between rebels and government. The empirical findings presented in this article suggest mainly two important findings which indicate that natural resources have different effect on each civil war outcome. First, while resource wealth enables rebels to strengthen the capability which may increase the likelihood of rebel victory, the results show that lootable resources tend to generate longer conflict. It can be said that a civil war may provide rebel groups 23

economic opportunities for accumulating private wealth that are not present during peacetime, rebels may prefer continued fighting to the end of civil war. Next, contrary to lootable resources, non-lootable resources generally reduce the conflict duration. Especially, resource wealth from non-lootable resources shortens the duration that a civil war ends in formal agreement or government victory. Non-lootable resources are more likely to be controlled by a government. Natural resources change the balance of power in favor of the government by reinforcing power basement of government and therefore resources are significantly, and positively related to government victory and peace settlement. Non-lootable resources may bring about the desired results to a government. 24

Reference Addison, Tony, Philippe Le Billon, and S. Mansoob Murshed (2002) Conflict in Africa: The Cost of Peaceful Behaviour. Journal of African Economics 11(3): 365-386. Aydin, Aysegul (2010) Where Do States Gone? Strategy in Civil War Intervention. Conflict Management and Peace Science 27(1):47-66. Balch-Lindsay, Dylan, and Andrew J. Enterline (2000) Killing Time: The World Politics of Civil War Duration, 1820-1992. International Studies Quarterly 44 (4): 615-642. Balch-Lindsay, Dylan, Andrew J. Enterline, and Kyle A. Joyce (2008) Third-Party Intervention and the Civil War Process. Journal of Peace Research 45(3): 345-363. Box-Steffensmeier, Janet M. and Bradford S. Jones (2004) Event History Modeling: A Guide for Social Scientists. New York: Cambridge University Press. Box-Steffensmeier, Janet M., Dan Reiter, and Christopher Zorn (2003) Nonproportional Hazards and Event History Analysis in International Relations. Journal of Conflict Resolution 47(1): 33-53. Brandt, Patrick, David Mason, Mehmet Gursesc, Nicolai Petrovsky, and Dagmar Radin (2009) When and How the Fighting Stops: Explaining the Duration and Outcome of Civil Wars. Defense and Peace Economics 19(6): 415-434. Buhaug, Halvard and Scott Gates (2002) The Geography of Civil War. Journal of Peace Research 39 (4): 417-433. Buhaug, Halvard, Scott Gates, and Päivi Lujala (2009) Geography, Rebel Capability, and the Duration of Civil Conflict. Journal of Conflict Resolution 53(4): 544-569. Buhaug, Halvard and Päivi Lujala (2005) Accounting for Scale: Measuring Geography in Quantitative Studies of Civil War. Political Geography 24(4): 399-418. Cleves, Mario A., William W. Gould, and Roberto G. Gutierrez (2004) An Introduction to Survival Analysis Using Stata. College Station: Stata Press. Collier, Paul (2000) Doing Well out of War. In Mats R. Berdal and David M. Malone, eds. Greed and Grievance: Economic Agendas in Civil Wars. Boulder, CO: Lynne Rienner. Collier, Paul and Anke Hoeffler (1998) On Economic Causes of Civil War. Oxford Economic Paper 50(4): 563-573. Collier, Paul and Anke Hoeffler (2004) Greed and Grievance in Civil War. Oxford Economic Paper 56(4): 563-96. Collier, Paul, Anke Hoeffer, and Mans Soderbom (2004) On the Duration of Civil War. Journal of Peace Research 41 (3): 253-273. DeRouen, Karl R. Jr. and David Sobek (2004) The Dynamics of Civil War Duration and 25

Outcome. Journal of Peace Research 41 (3): 303-320. Fearon, James D. (2004) Why Do Some Civil Wars Last So Much Longer than Others? Journal of Peace Research 41(3): 275-301. Gent, Stephen E. (2008) Going in When it Counts: Military Intervention and the Outcome of Civil Conflicts. International Studies Quarterly 52(4): 713 735. Gilmore Elisabeth, Gleditsch Nils Petter, Lujala Paivi, and Rød Jan Ketil (2005) Conflict Diamonds: A New Dataset. Conflict Management and Peace Science. 22(3): 252 72 Gleditsch, Nils Petter, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg, and Havard Strand (2002) Armed Conflict 1946-2001: A New Dataset. Journal of Peace Research 39(5): 615-637. Humphreys, Macartan (2005) Natural Resources, Conflict, and Conflict Resolution. Journal of Conflict Resolution 49(4): 508-537. Keen, David (2000) Incentives and Disincentives for Violence. In Mats R. Berdal and David M. Malone, eds. Greed and Grievance: Economic Agendas in Civil Wars. Boulder, CO: Lynne Rienner. Kreutz, Joakim (2010) How and When Armed Conflicts End: Introducing the UCDP Conflict Termination Dataset. Journal of Peace Research 47(2): 243-250. Le Billon, Philippe (2001) The Political Ecology of War: Natural Resources and Armed Conflict. Political Geography 20(5): 561-584. Lujala, Päivi (2009) Deadly Combat over Natural Resources: Gems, Petroleum, Drugs, and the Severity of Armed Civil Conflict. Journal of Conflict Resolution 53(1): 50-71. Lujala, Päivi (2010) The Spoils of Nature: Armed Civil Conflict and Rebel Access to Natural Resources. Journal of Peace Research 47(1): 1-15. Lujala, Päivi, Jan Ketil Rød, and Nadja Thieme (2007) Fighting over Oil: Introducing a New Data. Conflict Management and Peace Science 24(3): 239-256. Lujala, Päivi; Nils Petter Gleditsch, and Elisabeth Gilmore (2005) A Diamond Course?: Civil War and a Lootable Resource. Journal of Conflict Resolution 49(4): 538-562. Mason, T. David and Patrick Fett (1996) How Civil Wars End: A Rational Choice Approach. Journal of Conflict Resolution 40(4): 546-568. Mason, T. David; Joseph Weingarten Jr. and Patrick Fett (1999) Win, Lose, or Draw: Predicting the Outcomes of Civil Wars. Political Research Quarterly 52(2): 239-268. Regan, Patrick M., and Aysegul Aydin (2006) Diplomacy and Other Forms of Intervention in Civil Wars. Journal of Conflict Resolution 50 (5): 736-756. 26