Parity or Peace: Varying Intensities Caused by Third Party Interventions in Civil War

Abstract: 

Existing literature shows that external interventions into civil wars intensify and prolong conflict. However, why some interventions result in brutal, enduring conflict, while other interventions end conflict in only months, has yet to be explained. We hypothesize that these varied outcomes are due to a combination of three factors: the level of military capacity of the intervening state, the resources committed by that intervening state, and the balance of power between the government and opposition forces. We divide conflicts into weak interventions—interventions by countries with low military capacity, committing weapons or material aid; and strong interventions—interventions by countries with high military capacities that involve the deployment of troops. Using OLS regression and duration models on all civil wars between 1975 and 2009, we find that strong interventions, rather than weak interventions, lead to longer and more severe conflicts. Furthermore, we find that countervailing interventions approaching parity result in more deaths and increased conflict duration. This finding implies that strong interventions may still decisively end conflict, but only when combating weaker opposition.

Table of Contents: 

    Introduction

    In cases of civil war, many states face threats of war contagion, loss of trade agreements and resources, influx of refugees, and loss of lives. These threats often result in third party intervention in efforts to influence the war outcome to their benefit. However, many interventions fail to impact the war in the intended way, instead exacerbating the conflict further. Though there remain questions of whether or not third parties intervene in civil conflicts that are already worsening (Regan 2002), substantial empirical evidence supports the notion that adding ammunition, troops, or non-lethal material support into a conflict will increase the casualties and duration of that conflict (Regan 2002; Elbadawi and Sambanis 2000). Existing research fails to explain why certain interventions intensify and prolong conflict more than others. Why do some interventions result in the conflict lasting a decade with large numbers of casualties, while others shorten conflict with fewer casualties? This study expands upon the existing research and proposes that the level of intensification and duration depend upon a combination of factors—namely, the level of intervention, the capacity of the intervening party, and the presence of countervailing interventions.

    Varying conflict intensities caused by third party interventions is well illustrated in Sierra Leone’s civil war, which took place between 1992 and 2002. What began as a small, Liberian-supported group, opposing the government, expanded to include numerous warring parties that adopted brutal recruitment and looting tactics.  As the armies ravaged villages, victims flocked to nearby countries, and by 1992, 120,000 refugees had spread to Guinea. Through the Economic Community of West African States—Military Observer Groups (ECOMOG) arrangement, Nigeria and Guinea provided military support to the Sierra Leone military, although this proved ineffective in reducing the violence incurred by the Liberian-supported rebels (Gbrerie 2005).  According to the UCDP/PRIO Dataset on External Support, the main conduit of support to Sierra Leone’s primary rebel party (RUF) was from Liberian president Charles Taylor, who between 1992 and1996 provided the RUF with arms, safe haven, ammunition, material communications, and intelligence. Between 1998 and 2000, Burkina Faso also contributed weapons to the RUF. In addition, the United States and the United Kingdom offered non-lethal assistance to the Sierra Leone government army, such as maps, maintenance equipment, medical, and military supplies. Only in 2001 did British forces intervene with troops, after which the war ended within the year (Högbladh, Pettersson, and Themnér, 2011). This example shows that low-level interventions such as weapons and safe havens can serve to prolong war, but a single, decisive intervention in the form of troops and airstrikes by a strong power, can be effective in ending conflict swiftly.

    In order to craft effective policy for managing civil conflict, it is critical to identify the causal link between interventions and prolonged and intensified conflict. Policymakers must be cognizant of how certain types of interventions influence conflict in given situations. Understanding why interventions into Sierra Leone affected the outcome in different manners will give better insight into the most effective forms of intervention. Furthermore, examining cases such as Sierra Leone can help to illustrate which conditions in target states, given the capacity of bordering states, require certain forms of intervention rather than others. For example, third parties could potentially understand whether to intervene in the form of deploying troops or by providing lethal weapons given the number of other states intervening.

    This research question is particularly relevant regarding the current civil war in Syria. As the international community deliberates on whether or not to take action to end the conflict or assist the opposing parties, it is important to know whether weapons said to have been transferred from Qatar are exacerbating the conflict (Khalaf and Smith 2013). If this is the case, it is important to know what additional support would benefit or worsen the existing situation. With U.S. President Barack Obama having recently announced his decision to support the opposition party in Syria with lethal arms, it would be beneficial to be able to make a calculated prediction as to how this decision will affect the conflict (Mazzetti et al. 2013). This study will seek to predict how interventions, such as that of weapons transfers, will affect conflict intensity. And, subsequently, these predictions will guide policy makers in avoiding interventions that worsen conflict.

    Literature Review

    This section will first outline the existing literature relating to the dependent variables of this study: civil war duration and conflict severity. The chosen literature will look at the effects of both interventions and domestic conditions on civil war outcomes. Second, this section will provide an overview of scholarly work on the following independent variables: state capacity, strong and weak interventions, and effects of regionalism.

    Civil War Duration

    Within the last fifteen years, empirical studies have drawn attention to the adverse effects of third party interventions on civil war duration (Regan 2000, 2002; Balch-Lindsay and Enterline 2000; Elbadawi and Sambanis 2000). These studies provide evidence that intervention, whether in the form of non-lethal support, economic support, or military support, will cause the conflict to be prolonged and intensified. Within this literature, the authors discuss how certain forms of intervention intensify conflict more than others. There is a disagreement regarding the effects of biased interventions—interventions supporting either the government or opposition party—versus neutral interventions—interventions aiming to cease conflict without supporting a particular side. Balch-Lindsay and Enterline (2008) argue that biased military interventions prolong conflict; Regan and Aydin (2006) claim that they have no effect on conflict duration; and Collier, Hoeffler, and S̈oderbom (2004) find that these biased forms of military interventions reduce conflict duration. Scholars have also focused on the effects of interventions that occur at different stages in the conflict. Regan (2000,2002) finds that interventions at the beginning of conflict instill more confidence in warring parties and cause the conflict to last longer. He also measures the effectiveness of multilateral interventions as opposed to unilateral interventions, and what conditions contribute to their effectiveness. These findings indicate that for a multilateral intervention to be advantageous, it must have the consent of all parties in the conflict and have an organized and implemented strategy (Regan 2003). The above literature on interventions expands upon the general finding that third party interventions cause civil wars to be more severe and prolonged. This literature provides some explanation for why certain interventions intensify conflict more than others, but fails to account for what combinations of intervention efforts can explain the varying intensities.

    Scholars have also looked at how domestic conditions affect conflict duration. Collier, Hoeffler, and Soderbom (2004) show that civil wars in countries with high levels of ethnic fractionalization last longer because organizations garner support through ethnic affiliations. These authors, as well as Fearon and Laitin (2003), emphasize the impact of geography and terrain on the likelihood of civil war duration. Forested areas provide more coverage for rebel groups to organize (Collier, Hoeffler, and Soderbom 2004). Salehyan and Heger (2007) find that a regime’s coalition size can influence the level of brutality it employs. These authors find that smaller coalitions will be more likely to use brutal repression upon citizens than larger coalitions, because smaller coalitions will have more patronage and therefore be more loyal to the rulers. Scholars have also outlined societal dynamics that may increase the likelihood of civil war recurrence, such as levels of political participation, economic factors, and quality of life (Walter 2004). Though this paper will not be expanding on the effects of domestic conditions, it is important to note the impact of these conditions on civil war outcomes.

    Civil War Severity

    Due to limitations in the availability of accurate data on conflict severity, there has been less research on the effects of foreign interventions on severity as it pertains to civil wars. One way of measuring conflict severity has been through using data on one-sided violence. Empirical evidence demonstrates that rebel groups receiving external support from democratic regimes, or from countries with a significant number of human rights lobbies, are less likely to engage in one-sided violence against civilians than those groups receiving support from autocracies or countries with weak human rights lobbies (Salehyan, forthcoming). Though measuring one-sided violence produces meaningful findings, this method does not capture the total number of deaths—or level of severity—that ensue from a civil war. Lacina (2006) directly measures severity by using battle-death estimates as her dependent variable. She finds that democratic regimes in the conflict state are more likely to result in fewer combat deaths, while ethnic homogeneity results in increased fatalities. Although it appears that duration and severity are affected by the same forces, results may differ in that decisive, large-scale interventions may end conflict quickly but incur more bloodshed in doing so. This study seeks to use both severity and duration as dependent variables in order to measure the full impact of external intervention.

    State Capacity

    Though the literature shows that a state’s lack of economic and military capacity can increase the severity of civil war in that country (Collier, Hoeffler and Soderbom 2004), it fails to consider how the state capacity of intervening states can impact civil war outcomes. Lacina (2006) examines how a state’s low military and organizational capacity, as well as insufficient access to implements such as tanks, aircraft, and artillery, intensify the civil war in that country. This paper seeks to analyze how conflict may be affected by state capacity both in terms of domestic conditions, and in capacities of intervening states. Regan (2000) shows that an intervention by a major power is 22 percent more likely to end a conflict than intervention by a minor power. However, Regan’s prediction does not look at how this likelihood was influenced by the type of intervention—such as troop deployment as differentiated from weapons transfers.  Regan’s findings may also be contingent on the evenness of distribution of external support across warring parties. 

    “Neighborhood Effect”

    This paper argues that a critical factor in the effectiveness of third party interventions in civil wars is the influence of military capacities in neighboring states. While there has been a decline in interstate conflict in the 21st century (Gleditsch and Beardsley 2004), there has not been a decline in states that have been involved as external actors in intrastate conflicts. Scholars have suggested that this is due to negative externalities of civil wars on nearby territories, such as the impact of spillover effects of refugees or contagion of violence (Siverson and Starr 1990; Ward and Gleditsch 2002). Studies have shown that these spillover effects increase the probabilities of the onset of interstate or civil war (Gleditshch, Salehyan and Schultz 2008). Few studies, however, explore the way in which civil war duration and severity may suffer as a result of states intervening into the war without sufficient means to end the war. The states that are pulled into the war as a result of the negative externalities civil wars inflict upon surrounding territories are the same states that are insufficiently capable of bringing the war to a rapid and less fatal conclusion.

     Scholars have found that conflict between states within the same region increases the likelihood of intrastate conflict, and show that 66 percent of states involved in militarized disputes are contiguous (Gochman 1990).  The focus of this finding, however, is to explain the way in which contiguous states’ involvement in conflict is a result of territorial disputes, rather than how contiguous interventions impact intrastate conflict outcome (Gochman 1990; Vasquez 1995). Collier et al. (2003) expand on these findings to assert theories relating to the “conflict trap,” arguing that regional conflict is likely to arise from economic and social spillover from civil wars, spreading to the whole “neighborhood” (Collier et al. 2003; Gleditsch 2007). Kathman (2010) brings particular attention to the impact that neighborhood effects have on interventions, finding that the likelihood of civil war diffusion is directly related to a state becoming involved in a neighboring civil war. This illustrates that geographic proximity, which can involve ethnic affinity, security ties, and political and economic relationships, can spur contiguous interventions in civil wars (Austvoll 2005; Kathman 2010; Salehyan 2009).

    Countervailing Interventions

    In addition to the influence of intervention characteristics on civil war duration and intensity, we posit that civil war outcomes may be affected by whether or not an intervention creates power preponderance between warring parties. Balch-Lindsay and Enterline (2000) conduct a study regarding the distribution of external support across warring parties, which is closest to the aims of this study. The authors looked at 152 cases of civil war between 1820 and 1992 to analyze the interdependent nature of interventions. Findings from this study showed significant support for the theory that interventions that are more evenly distributed across warring parties result in longer conflict (Balch-Lindsay and Enterline 2000). The operationalization of this study, however, does not capture a thorough measurement of parity or power preponderance. Their decision to measure balanced interventions with a dichotomous variable—capturing whether both sides had more than one external supporter—does not capture the strength of each intervention that may be influenced by military capacity and amount of resources committed. Using updated data and recent instances of civil war, this paper seeks to further analyze the relationship between power preponderance and intervention strength on civil war severity and duration.

    Studies performed regarding the effectiveness of third party interventions, as well as the motivations and capacity of intervening states, lead to questions of whether or not there is a relationship between the military capacity of intervening states and the effect these interventions have on the conflict. Are those countries with greater incentives to intervene also those with lesser capacity to support an armed intervention or provide enough weapons to impact the conflict? Understanding the manner in which combinations of components of interventions affect civil wars may explain the variation in conflict intensity that interventions influence.

    Though a general consensus exists regarding the notion that external interventions into civil wars intensify these conflicts, why these interventions lead to different levels of intensity has yet to be adequately explained.  Existing literature fails to consider the advantages and disadvantages of interventions sparked by the “neighborhood effect” on the outcome of civil war. Scholars have also neglected to investigate the impact of weak or strong levels of intervention on the severity and duration of conflicts, as well as the influence of countervailing interventions.

    Theory

    This paper proposes that the strength of intervention, combined with a state’s military capacity to support an intervention, significantly impacts the severity and duration of conflict. By analyzing the influence of interventions of varied strength on civil conflict, as well as the effects of varying military capacities on conflict, this paper draws a causal link between the level of intervention and the proclivity of certain types of states to undertake effective or ineffective interventions. Additionally, this section will outline the effects of countervailing interventions on conflict duration and severity. I will argue that in addition to the strength of an intervention, the ratio of support given to rebel and government forces will impact the balance of power between the two combatants, potentially making the conflict increasingly difficult to resolve. The ten-year civil war in Sierra Leone will be used to show how states’ capacities to use weak or strong forms of interventions, as well as the motivations for proximate states to intervene, help to account for why interventions result in varied levels of conflict severity and duration.

    Strong and Weak Interventions

    Contrary to the implications posed by scholars of third party interventions, not all interventions increase conflict duration and severity to the same extent. As the civil war in Sierra Leone illustrates, the British intervention had a much different effect on the outcome of the civil war than the Liberian or ECOMOG interventions did. More specifically, the types of resources committed by the United Kingdom (UK), Liberia, and ECOMOG had different impacts on the conflict outcome. This example provides support for the theory that “weak interventions” increase the length and intensity of conflict. We define weak interventions as operations that provide support short of troops— such as non-lethal support, weapons, ammunition transfers, and material support (secondary non-warring support). Strong interventions are defined as those that involve troop deployment, air strikes, or sea power (secondary warring support). The provision of lower levels of support such as weaponry may provide combatants with enough support to continue fighting, but may be insufficient to enhance prospects for winning the war. These efforts to end the conflict thus work adversely to aggravate them further.

    Conversely, concrete surges of troops and strategic airstrikes can decisively expedite the end of a civil war.  As seen by the UK intervention into Sierra Leone, a significant amount of resources committed by a third party, such as sending troops or using advanced technology, can have a determined impact on the war. With a strong military intervention, there is less opportunity for armed insurgencies and external support to develop, and war will likely end in a more expedient manner relative to weak intervention efforts. The varied impacts created by weak and strong interventions culminate in the first of three hypotheses in this paper:

    Hypothesis 1: Weak interventions, such as providing arms transfers, will prolong and intensify the conflict more than strong interventions, such as the deployment of troops.

    Military Capacity

    Research on regional conflict dynamics exposes an important element of third party interventions: countries will often intervene into conflicts, despite lacking the means to decisively affect the outcome. As civil wars pose threats to neighboring states, due to such factors as the spillover of displaced persons or the diffusion of violence across borders, neighboring countries are motivated to affect the outcome of nearby conflicts. For countries with low military capabilities and the incapacity to provide significant, meaningful material support to the warring parties, efforts to contribute will often come in the form of an assortment of weaker forms of lethal and non-lethal support to the warring armies. Though these efforts on behalf of contiguous third parties to affect the outcome of the conflict may seem like a rational option given these countries’ capabilities, and as a result of the negative externalities they experience, these interventions may significantly increase the severity and duration of that war rather than end it. They essentially provide enough support to affect the war but not enough to decide the outcome.

    Sierra Leone’s civil war is an example of war occurring in a region comprised of states with lesser military capacities than those of the major military powers. When one country in the region has a civil war, there are numerous ramifications felt by surrounding states, such as the influx of refugees requiring food and shelter and threatening the spread of disease through refugee encampments (Collier et al. 2003). In addition to displacement, there is the threat of policies a new neighboring regime may impose, such as a change in vital trade agreements or policies prompting segments of the population to emigrate. There is also the threat that the violence will cross into nearby regions or threaten the governments of contiguous states (Kathman 2010). For these reasons, even countries with very poor military capacities will often be motivated to affect the outcome of the neighboring war by intervening in any way possible. As seen in the case of Sierra Leone, the government army was predominantly comprised of young recruits who were poorly trained (Gbrerie 2005). Though unable to contribute large numbers of well-trained troops, Nigeria and Guinea aided these untrained troops by supplying them with weapons. Nearby countries supported the opposing parties to a degree that did not end the conflict but instead increased its severity. The conflict persisted for a decade while incurring vast numbers of fatalities.

    Despite evidence of weak interventions aggravating conflict, external interventions by international actors or states with high levels of military capacity can also prolong and exacerbate conflicts. Significant empirical evidence shows that strong interventions do increase the severity and length of conflict (Regan 2000, 2002; Balch-Lindsay and Enterline 2000; Elbadawi and Sambanis 2000). These interventions, however, impact conflict to varying extents. Unlike rebel and government forces in intrastate conflict that struggle to coordinate strategies amongst troops and train civilian recruits to follow these strategies, international actors or states have high levels of military competence and can provide large militaries with ample resources at hand. These resources include sea and air power and technologically advanced equipment including communication and tracking devices. These states, however, have less incentive to intervene than those states proximate to conflict.  States farther away from the conflict do not experience direct impacts of displacement or violence, their regimes are less at risk, and their economies less interwoven. Though less-proximate states may be concerned over the potential for a broken trade agreement or reduced access to a critical resource, the impact of civil war on neighboring countries is far more salient to these than on those countries that are farther away. Nevertheless, frequently less-proximate states do intervene to affect conflict outcome, in the interest of alliances, trade agreements, or humanitarian concerns. This paper proposes that external interventions by third-parties with high military capacity relative to the target state, who commit greater resources, will end the conflict sooner than will interventions with low military capacities, who commit fewer resources, despite less motivation to become involved.

    Hypothesis 2: Interventions from countries with lower military capacities will lead to longer and more severe conflict.

    Countervailing Interventions

    This paper hypothesizes that warring parties receiving relatively equal levels of external support will be more balanced, thereby rendered incapable of decisively defeating each other—thus lengthening and intensifying the conflict. Although a single intervening state could affect the outcome by means of the strength of its intervention, as outlined above, this potential impact is affected by the strength of the opposing force. The influence of the UK’s strong military intervention on the outcome of Sierra Leone’s war  may have been different had the RUF been supported to a comparable degree by another external actor. In the nine years of the civil war in Sierra Leone prior to the British intervention, a relatively equal amount of external support aided both the government and rebel armies. This resulted in a balance of power between the warring parties too close to parity for either side to achieve decisive victory. This paper uses the international relations concepts of parity and preponderance. When forces reach parity they are balanced in strength, and when one party has power preponderance, that party is stronger than its opposition (Moul 2003). This paper argues that conflicts will end sooner with fewer battle deaths when one party receives significantly more support than the opponent than if support for the two parties approaches parity.

    Hypothesis 3: Civil wars in which the opposition party and government party are supported by similar strength of intervening force, causing forces to approach parity, will be longer and more severe than civil wars in which the opposition party and government party are supported at significantly different levels.

    Data and Methods

    The dataset for this project includes all instances of civil war between 1975 and 2009. Ideally, analysis would include wars prior to this period to include the conflicts occurring between 1945 and 1975. However, the “UCDP External Support in Armed Conflict” dataset most suited to meet the needs of this project excludes conflict during that time (Högbladh, Pettersson and Themnér, 2011). The dataset constructed here includes all civil wars, without focusing on wars in specific regions, in order to compare the effects of interventions across regions. The unit of analysis is conflict-year—encompassing one observation per year for every conflict.

    This paper captures the severity of conflict by the number of annual deaths occurring during a given conflict. This dependent variable is derived from Bethany Lacina and Nils Petter Gleditsch’s battle-deaths data (2005), which compiles battle-death estimates from the Uppsala/PRIO Armed Conflict Dataset between the years of 1945 to 2008. Lacina and Gleditsch’s dataset provides a low estimate, high estimate, and a “best guess”—this paper will utilize the best guess estimate. This battle-death measure is appropriate for this paper because it includes conflicts with as few as 25 battle-related deaths per year. This low threshold, as opposed to Correlates of War “Intra-Sate War Data” (Sarkees, Reid and Wayman 2010) which only includes conflicts over 1,000 battle-deaths, is advantageous for research on interventions because it includes conflicts that may have been stunted early on by interventions. This dataset also only includes deaths directly inflicted by an armed force as opposed to including deaths due to non-lethal forces. This is necessary to be able to determine the direct impact from armed forces and intervention forces on conflict severity. Another advantage of this measure is that it includes the number of deaths per year, allowing us to link annual death totals to particular interventions occurring in a given year. One edit made to this data was to remove several years of conflict said to have taken place in the United States after 2001. The bulk of this conflict, we argue, occurred in Afghanistan, and therefore corresponds to the respective controls for that country.

    In addition to using conflict severity as a dependent variable, this paper examines the duration of the conflict as a second dependent variable. The length of civil conflict is derived from the “UCDP External Support in Armed Conflict,” which includes all civil wars between 1975 and 2009, and provides annual data for these conflicts (Högbladh, Pettersson and Themnér, 2011). Though other data sets such as the “Non-State Actor Data” (Gleditsch, Cunningham and Salehyan 2012) provide data on the length of conflicts, it is necessary for this project to be able to disaggregate an intervention’s impact on conflict duration after the intervention occurs, as opposed to the entire length of conflict. To accurately measure how an intervention impacts conflict duration, interventions must be measured annually.

    The two primary independent variables in this study are the intervening states’ military capacity and the type of intervention used. Examining the type of intervention illustrates the degree to which certain forms of intervention such as troop deployment or arms transfers affect the severity and duration of the conflict. Considering the military capacity of the intervening countries provides insight into the influence that a state’s ability to afford powerful intervention strategies has upon the severity and duration of the conflict. However, one must consider whether a country with a high military capacity is choosing to commit fewer resources than it potentially could.  Conversely, it is important to consider situations in which countries with low military capacities choose to contribute a large percentage of resources. Thus, it is important to consider the interaction between military capacity and resources committed.

    We do this by creating a unique variable based on the sum of the intervening state’s military capacity and the amount of resources it commits. We assign a numerical value of 0, 1, or 2 to the form of intervention used, and a value of 1, 2, or 3 to the state committing those resources, based on that state’s level of military and economic capacity. The sum of these two values becomes the independent variable representing the interaction between these two factors. For intervention type, the highest level of intervention (2) is given to interventions involving troops—the highest level of intervention recorded in the dataset. The next type of intervention is weapons transfers, given a value of (1). The final value, (0), is assigned to interventions that have not provided support in the form of weapons transfers or troops. This implies that these countries have contributed by providing safe havens, material aid, training, or intelligence. In order to assign a value to the military capacity, we divide intervening states according to high, middle, and low military capacity. Although possible methods for determining a country’s capacity to afford military interventions include measuring Gross Domestic Product (GDP) or incorporating data on military capacities, these operationalizations do not capture a country’s full capacity to support a military intervention. These measures do not account for potential alliances existing between high-capacity states or states with economic and human assets. Instead, we give the highest value (3) to countries that are either members of the Organization of Economic Cooperation Development (OECD) or a permanent member of the United Nations Security Council. These 36 countries, we argue, have the institutional support and the economic backing to support high levels of interventions based on their positions in the international community, economic viability, and military capabilities. The lowest value for military capacity (1) is assigned to the 47 “Least Developed Countries” as defined by the United Nation (United Nations Conference on Trade and Development). Based on analyses of income, human assets, and economic vulnerability, these countries are identified as the world’s poorest countries, and as those that are least capable of supporting a robust military intervention. All countries other than those identified as having high or low military capacities are given an intermediate value of 2. These countries—Turkey and Egypt are examples—are less capable then the most developed countries to stage thorough interventions, yet are not as handicapped as the least developed countries when they do so.

    Because this paper does not focus on the impact of non-state actors, and due to these actors’ military capacities being difficult to capture, we do not assign a specific value for them. Instead, a dummy variable is included to represent whether or not non-state actors intervened in a conflict in a given year. We also ran a test using a variable that combined states with mid-level and low-level military capacities. Permanent members of the UN Security Council and members of OECD were assigned a value of 2, and all other intervening parties, including non-state actors, were given a value of 1. Results with this variable were similar to the largest intervention variable shown in Table 1 and Table 2. (See Appendix).

    For every year of each conflict, the value of the intervention with the greatest strength is used as the independent variable. This tests whether or not conflicts involving large-scale interventions by countries with high military capacities are more likely to terminate and be less severe than conflicts that only involve external support from countries with low military capacities and lower forms of intervention.

    The third independent variable is the level of power preponderance between opposing forces. We do this by calculating the largest strength variable, as previously outlined, for both the government and rebel parties. We then calculate the absolute value of the difference between the strength of the intervener for the government forces and the strength of the intervener for the opposition forces. For each warring party, we create a value based on the military capacity of the largest intervener, combined with the resources committed. We then take the absolute value of the difference between the strongest interveners on each side, creating a variable measuring the unevenness of support each side in the conflict receives. 

    This research uses the “External Support in Armed Conflict 1975–2009” to capture the variables representing intervening country, type of intervention, and intervention year (Högbladh, Pettersson and Themnér 2011). This dataset differentiates external support for government and opposition forces, which is necessary for measuring the impact of countervailing interventions. Though there are datasets available that include external support for civil war combatants, most other datasets lack the disaggregated data regarding support for countervailing interventions that this project requires.

    This paper also includes a number of control variables. First, we control for Gross Domestic Product (GDP), derived from the World Bank, to determine whether conflict duration and severity is a function of countries’ levels of wealth. Second, we control for the level of ethnic fractionalization, using Fearon and Laiton’s (2003) measure. This variable is included to test whether a high level of ethnic fractionalization in a country contributes to a civil war’s severity and duration. Third, we use Polity IV to measure how the regime type of the host country may affect the repression in that conflict, as well as that regime’s ability to control and end the violence (Marshall and Jaggers 2005).

    We also control for whether a country’s population affects the number of fatalities or the length of conflict, with data from the World Bank. We predict that increased population size will result in more fatalities and a longer conflict. Finally, we control for the presence of former British and French colonies, using Fearon and Laitin’s (2003) data. This captures whether or not certain interventions are a result of previous colonial ties to those territories.

    Because battle deaths are a continuous variable, we use an Ordinary Least Squares (OLS) regression test to measure severity.  To measure duration, we use a Cox Proportional Hazards model. This calculates the likelihood that conflict will end in a given year.

    Results

    In Model 1, Table 1, we estimate the impact of intervention strength on the number of annual battle deaths. Our findings show a positive relationship between intervener strength and battle deaths—indicating that interventions by states with high military capacities that commit a significant amount of resources to the conflict cause more fatalities in a given year. The significance of another variable—power preponderance—provides insight, however, into why this relationship may exist. Findings show a negative relationship between power preponderance and number of battle-deaths, providing support for the hypotheses that conflicts in which government and opposition forces receive similar levels of support result in more fatalities. Furthermore, this finding indicates that when strong interventions oppose forces with relatively equal levels of strength, the combined interventions will cause the conflict to intensify and lengthen.

    Upon closer inspection of the data, we find that these results may be swayed by the number of conflicts in which warring parties are supported at similar levels. Table 3 indicates that only 12 out of 1,176 observations have a high level of power preponderance, while the remaining conflicts are divided between having intermediate levels of strength imbalance and complete parity. These findings imply that the majority of conflicts will be unlikely to terminate given the level of parity between forces. This shows that while strong interventions may be positively related with battle deaths, the majority of strong interventions into civil conflict are combating an opposition supported by nearly equal levels of support. This indicates that while strong interventions appear to cause longer and bloodier conflicts, this may be more a function of parity of power between interveners than intervention strength.

    The findings in Table 2 are similar to those in the aforementioned Table 1.  The results illustrate a positive relationship between the strength of intervention and conflict duration. This indicates that the stronger the intervention becomes, the less likely the conflict is to end. This model also signifies a relationship between greater degrees of power preponderance and an increased likelihood of conflict termination. This indicates that when one side is supported at a substantially greater level than its opponent, the conflict will be shorter than if forces are closer to parity.

    Model 2 and Model 4 in both the OLS regression model and the Cox Proportional Hazards model unpack the largest intervention variable that was tested in the first models. These models attempts to capture the impact of the strongest intervention separate from the influence of opposing forces with balanced strength. Because testing the respective parties’ strength tests a similar relationship to that of power preponderance, the variable power preponderance is excluded from these models. Within this model, we see largest intervention strength behaving more closely to what our theory predicts: the stronger the intervention, the fewer annual battle deaths and the shorter the conflict duration.

    Several of the control variables were also found to be significant. In Models 1 and 2, interventions involving non-state actors were found to have increased severity. This finding indicates that even if interventions by non-state actors are not the largest intervention in a given year, these efforts will likely lead to increased severity and duration. The analyses also illustrate significant findings for gross domestic product, level of democracy, size of population, and ethnic fractionalization. Higher gross domestic product is found to reduce number of battle deaths, as seen in Table 1-Model 2, as opposed to an increase in ethnic fractionalization, which is found to increase annual fatalities—as is seen in both models in Table 1. Though population size and level of democracy were insignificant in the OLS regression test, they produced counterintuitive results in both models in Table 2. The findings indicate that democracy has a positive relationship with conflict duration; greater population is shown to increase likelihood of conflict termination.

    Controlling for whether or not conflicts occurred in previous British or French colonies showed highly significant results. Interventions into former British or French colonies resulted in fewer battle deaths. Yet, findings also indicate that interventions in former British colonies are likely to increase the duration of conflict.

    Conclusion

    In seeking to determine which intervention characteristics best lend themselves to conflict termination, we find that the power of intervention type to decide conflict outcome is secondary to a broader context. An intervention’s impact on conflict outcome is highly dependent on how much external support is given to the opposition. This finding implies that when an intervention employs enough strength to significantly outweigh the opposing party, the conflict will be more likely to have a shorter duration and incur fewer fatalities.

    This dynamic of power preponderance can be seen in the conflict between Georgia and South Ossetia. In August, 2008, Georgia staged a military offensive on South Ossetia, in its capital, Tskhinvali. The next day, Russia intervened on behalf of South Ossetia by providing army units, naval forces, and air forces. The conflict continued for five days, until Georgian forces retreated and a peace settlement was quickly negotiated (Golliday 2009). This case demonstrates the impact of a large-scale intervention by a country with high military capacity on a recently ignited war. The sudden shift in power preponderance between South Ossetia and Georgia led to timely conflict termination.

    Although the conflict in Georgia is an example of the way that a strong intervention can effectively terminate conflict, the analyses in this study indicate that strong interventions usually have the opposite effect. We posit that this finding is partly a result of selection bias for two reasons: first, interventions involving states with high military capacities who choose to commit a high amount of resources are more likely to occur in conflicts that are already intensifying, so ending these conflicts will be inherently bloodier and of longer duration; second, results showing a relationship between strong interventions and conflict intensification may be due to states selecting conflicts in which the opposition receives commensurate support. Therefore, strong interventions are more often seen in conflicts in which warring parties are balanced in strength. As the findings show, conflicts in which warring parties are supported at similar levels last longer and incur more battle deaths.

    The findings of this study imply policy considerations regarding the methods by which interventions can effectively terminate conflict.  We attempt to account for selection bias by testing only the conflicts that have reached a threshold of 1,000 deaths as opposed to 25. Tests were run on data from the same UCDP External Support dataset, including only conflicts that had one or more observations in which 1,000 battle deaths were recorded for a given year. Results were similar to tests with lower battle death threshold. (See appendix). The level of the intervention—strong or weak—may contribute to the level of parity or preponderance between warring parties. If the opposition party is not receiving external support, then an intervention of greater strength is likely to contribute to one side outweighing its opposition. Conflicts end faster and cause fewer fatalities if an intervention creates a high degree of power preponderance between warring parties, assuming the other side does not provide a strong response. Interventions in the form of troops, with possible addition of other forms of lethal aid, are more likely to create this preponderance, whereas support in the form of non-lethal materials such as safe havens and intelligence, or lethal weapons, are less likely to make one party significantly stronger than the other. When one party receives a low level of support, it is easy for the opposition to seek out the same level of support. Interventions should occur early on in the course of the conflict. As a conflict persists, both sides of the conflict are likely to acquire degrees of external support. A decisive intervention near the beginning of conflict is more likely to end the war, provided one party is not allowed time to build up strength in opposition. This relationship warrants further investigation into whether or not warring parties become more balanced as conflicts persist.

    Several findings in this project prompt a need for further analysis. This study focuses on interventions by state actors, and in doing so, fails to explore the full impact of interventions by non-state actors. Findings of the analyses reveal a significant relationship between interventions involving non-state actors and longer and more intense conflicts. Since non-state actors often have weaker military capacity, this finding may indirectly support the hypothesis that lower levels of intervention intensify conflict. However, this proposition requires further investigation.

    Future research should also unpack the relationship between intervention strength and power preponderance. Although this project has shown the pacifying influence of power preponderance in conflict, and implies that intervention strength is a mechanism to create this imbalance, it does not account for the spectrum of potential interactions between these variables. Further research could differentiate the impact on conflict of balanced forces as a result of strong interventions from those that are a result of weak interventions. This would further isolate the relationships between the variables that this project has attempted to measure and the severity and duration of conflict. Another avenue for further analysis would be to determine the impact of the number of intervening parties on conflict severity. One hypothesis could be that as the number of intervening parties increases, the more disjointed intervention efforts become, thereby lengthening or worsening the conflict. In terms of operationalizing the variables in these projects, future researchers should be aware of the limitations of using the best estimate in Bethany Lacina and Nils Petter Gleditsch’s battle-deaths data (2005), since it is missing a large number of observations.

    This study has begun to explore the complex interaction between characteristics of intervening states, characteristics of host states, and the conflict context—especially in terms of conflict outcome. We find that while the strength of the intervention is a primary component of intervention effectiveness, it is significant within these parameters only to the extent that it defines the level of parity or preponderance between warring parties. While the relationship between strength imbalances and strength of interventions offers insight into intervention strategy, further exploration into the implications this interaction could have on intervention policy is merited.

    Appendix

    Table 4, Table 5, Table 6, and Table 7 show results testing intervention strength—an ordinal variable that separates military capacity according to whether or not countries are permanent members of the UN Security Council or OECD countries. This variable includes the presence of non-state actors. Table 8, Table 9, Table 10, and Table 11 tests conflicts that reach 1,000 battle deaths. These models use the original ordinal variable largest intervention strength from Table 1 and Table 2.

    References

    • Austvoll, Martin. 2005. “Transnational Ethnic Dimensions of Third-Party Interventions in Civil Conflicts.” MA thesis, Department of Political Science, University of Oslo.
    • 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.
    • ―.2008. “Third Parties & the Civil War Process”. Journal of Peace Research. 45 (3): 345-363.
    • Collier, Paul, V.L. Ellitio, Harvard Hegre, Anke Hoeffler, Marta Reynal-Querol, Marta, and Nicholas Sambanis. 2003. Breaking the Conflict Trap: Civil War and Development Policy. Washington, DC: World Bank and Oxford University Press.
    • Collier, Paul, Anke Hoeffler, and Mans Soderbom. 2004. “On the Duration of Civil War.” Journal of Peace Research 41(3): 253-273.
    • Correlates of War. 2010. “Intra-State War Data.” Sarkees, Meredith Reid and Frank Wayman Resort to War: 1816 - 2007. CQ Press.
    • Cunningham, David E., Kristian Skrede Gleditsch, and Idean Salehyan. 2009. “Non-State Actor Data.”
    • Elbadawi, Ibrahim A. and Sambanis, Nicholas. 2000. “External Interventions and the Duration of Civil Wars” presented at the World Bank’s Development Economic Research Group (DECRG) conference on “The Economics and Politics of Civil Conflicts,” Princeton University, NJ.
    • Fearon, James D., and David D. Laitin. 2003. “Ethnicity, Insurgency, and Civil War.” American Political Science Review  97(1) 75-90.
    • Fearon, James D. 2004. “Why do Some Civil Wars Last so Much Longer than Others?” Journal of Peace Research 41(3): 275-301.
    • Fortna, Virginia. 2004. “Does Peacekeeping Keep Peace? International Intervention and the Duration of Peace After Civil War”. International Studies Quarterly 48(2): 269–292
    • Gbrerie, Lansana. 2005. A Dirty War in Africa: The RUF and the Destruction of Sierra Leone. Bloomington, Indiana: Indiana University Press.
    • Gleditsch, Krisitan S. and Kyle Beardsley. 2004. “Nosy Neighbors: Third Party Actors in Central American Conflicts.” Journal of Conflict Resolution 48(3): 379-402.
    • Gleditsch, Kristian. 2007. “Transnational Dimensions of Civil War.” Journal of Peace Research 44 (3): 293–309.
    • Gleditsch, Kristian, Idean Salehyan, and Kenneth Schultz. 2008. “Fighting at Home, Fighting Abroad: How Civil Wars Lead to International Disputes.” Journal of Conflict Resolution 52 (4): 479-506.
    • Gochman, Charles S. 1990. “The Geography of Conflict: Militarized Interstate Disputes since 1816.” Paper presented to the annual meeting of the International Studies Association, Washington, D.C., 10-14 April.
    • Golliday, Jessica. 2009. “Who’s to Blame? Western Responses to the 2008 Georgia – Russia – South Ossetia Conflict.” Thesis submitted at University of North Carolina at Chapel Hill.
    • Greig, Michael J. and Paul F. Diehl. 2005. “The Peacekeeping-Peacemaking Dilemma.” International Studies Quarterly 49: 621-645.
    • Heger, Lindsay and Idean Salehyan. 2007. “Ruthless Rulers: Coalition Size and the Severity of Civil Conflict.” International Studies Quarterly 51(2): 385–403.
    • Kathman, Jacob D. 2010. “Civil War Contagion and Neighboring Interventions.” International Studies Quarterly  34(2): 989-1012.
    • Lacina, Bethany and Nils Petter Gleditsch, 2005. “Monitoring Trends in Global Combat: A New Dataset of Battle Deaths.” European Journal of Population 21(2–3): 145–166.
    • Lacina, Bethany. 2006. “Explaining the Severity of Civil Wars.” Journal of Conflict Resolution 50(2): 276–89.
    • Marshall, Monty G., and Keith Jaggers. 2005. Polity IV Project: Political Regime Characteristics and Transitions, 1800-2004, Dataset Users’ Manual. Arlington, VA: George Mason University, Center for Global Policy.
    • Mazzetti, Mark, Michael R. Gorden, and Mark Landler. “U.S. Is Said to Plan to Send Weapons to Syrian Rebels.” New York Times. 13 June 2013. Retrieved 2013-06-13.
    • Moul, William. 2003. “Power, Parity, and War between Great Powers, 1869-1989.” Journal of Conflict Resolution 47(4): 468-489.
    • Organization for Economic Cooperation and Development (OECD). 2013. List of OECD Member countries - Ratification of the Convention on the OECD. (Accessed July 7, 2013).
    • Quinn, Michael J., David T. Mason, and Mehmet Gurses. 2007. “Sustaining the Peace: Determinants of Civil War Recurrence.” International Interactions 33(2): 167-193.
    • Regan, Patrick M. 2000. Civil Wars and Foreign Powers: Outside Interventions in Intrastate Conflict. Michigan: University of Michigan Press.
    • ―.2002. “Third Party Intervention and the Duration of Intrastate Conflicts.” Journal of Conflict Resolution 46(1): 55-73.
    • ―.2003. “The Short-Term Effects of Military Interventions in Civil Conflict.” International Negotiation 7(3): 299-311.
    • Regan, Patrick M. and Aysegul Aydin. 2006. “Diplomacy and Other Forms of Intervention in Civil Wars.” Journal of Conflict Resolution 50 (5): 736–756.
    • Khalaf, Roula and Abigail Fielding Smith (16 May 2013). "Qatar bankrolls Syrian revolt with cash and arms.” Financial Times. Retrieved 3 June 2013.
    • Salehyan, Idean, David Siroky, and Reed M. Wood. “External Rebel Sponsorship and Civilian Abuse: A Principal-Agent Analysis of Wartime Atrocities.” (Forthcoming).
    • Salehyan, Idean. 2009. Rebels Without Borders: Transnational Insurgencies in World Politics. Ithaca, NY: Cornell University Press.
    • Siverson, Randolph M. and Harvey Starr. 1991. The Diffusion of War: A Study of Opportunity and Willingness. Ann Arbor: University of Michigan Press.
    • United Nations Conference on Trade and Development (UNCTAD). 2012. UN list of Least Developed Countries. http://unctad.org. (Accessed July 7, 2013).
    • Uppsala Conflict Data Program (UCDP). “External Support Project-Disaggregated/Supporter Dataset.” Högbladh, Stina and Therése Pettersson and Lotta Themnér. Version 1.0-2011.
    • Vasquez, John A. 1995. “Why Do Neighbors Fight? Proximity, Interaction, or Territoriality.” Journal of Peace Research. 32(3): 277-293.
    • Walter, Barbara, 2002. Committing to Peace: The Successful Settlement of Civil Wars. Princeton, NJ: Princeton University Press.
    • ―.2004. “Does Conflict Beget Conflict? Explaining Recurring Civil War.” Journal of Peace Research 41(3): 371-388.
    • Werner, Suzanne and Amy Yuen. 2005. “Making and Keeping Peace.” International Organization  59(2): 261-292.
    • World Bank. 2012. World Development Indicators. http://worldbankdata.org. (Accessed July 7, 2013.)

    Table 1: OLS Regression Model, Results for Battle Deaths

      Model 1
    Largest intervener strength
      Model 2
    Strength side A and side B
     
    Variables Coefficient (robust s.e.) P-Value Coefficient (robust s.e.) P-Value
    Strength side A 951.11
    (471.44)
    0.044**
    Strength side B 640.76
    (398.40)
    0.108 
    Largest intervention strength 797.64
    (268.13)
    0.022** -833.82
    (530.46)
    0.116 
    Power preponderance -786.61
    (314.58)
    0.061*
    Intervention by non-state actor 1931.94
    (1124.15)
    0.086* 2090.33
    (1208.52)
    0.084*
    Gross domestic product (log) 113.63
    (178.64)
    0.525 -525.86
    (299.79)
    0.079*
    Polity2 43.32
    (68.41)
    0.527 49.42
    (70.45)
    0.483 
    Ethnic fractionalization 4955.04
    (2009.76)
    0.014** 4941.43
    (2031.15)
    0.015**
    Population (log) 113.63
    (178.64)
    0.525 104.66
    (181.10)
    0.563  
    Former British colony -1841.43
    (875.85)
    0.026** -1605.25
    (844.11)
    0.057*
    Former French colony -3369.88
    (950.27)
    0.000*** -3454.01
    (981.11)
    0.000 *** 
    N 485   485  

    Wald Chi2

    24.33   24.68  
    Prob > chi2 0.0038   0.0060  

    ***p<.01, **p<.05, *p<.10; two-tailed test.

    Table 2: Cox Proportional Hazards Model, Results for Conflict Duration

      Model 1
    Largest intervener strength
      Model 2
    Strength side A and side B
     
    Variables Hazard (robust s.e.) P-Value Hazard (robust s.e.) P-Value
    Strength side A .737
    (.080)
    0.005***
    Strength side B .813
    (.095)
    0.077*
    Largest intervention strength .753
    (.076)
    0.005*** 1.282
    (.179)
    0.076*
    Power preponderance 1.283
    (.133)
    0.017**
    Intervention by non-state actor .533
    (.158)
    0.034** .499
    (.149)
    0.020**
    Gross domestic product (log) .968
    (.051)
    0.543 .975
    (.053)
    0.647
    Polity2 .949
    (.017)
    0.005*** .951
    (.018)
    0.010***
    Population (log) 1.097
    (.040)
    0.012* 1.103
    (.042)
    0.010**
    Ethnic fractionalization 1.518
    (.594)
    0.287 1.572
    (.620)
    0.252
    Former British colony .645
    (.149)
    0.046** .618
    (.144)
    0.040**
    Former French colony .886
    (.222)
    0.631 .945
    (.249)
    0.832
    N 685   685  

    Wald Chi2

    45.45   50.54  
    Prob > chi2 0.0000   0.0000  
    Log pseudolikelihood -691.20035   -690.33566  

    ***p<.01, **p<.05, *p<.10; two-tailed test.

    Table 3: Distribution of Levels of Parity/Preponderance in Civil Conflict Data

    Power Preponderance

    Frequency

    0

    448

    1

    175

    2

    170

    3

    176

     4

    195

    5

    12

    Total

    1,176

    Table 4: Cox Proportional Hazards Duration Model

      Haz. Ratio Std. Err. (Robust) P>z
    Intervention strength 1.45 .225 0.022**
    Power preponderance .733 .069 0.001***
    Population (log) .823 .080 0.047**
    Intervention by non-state actor .436 .137 0.008***
    GDP (log) .960 .054 0.479
    Former French colony .888 .234 0.655
    Former British colony 599 .145 0.035**
    Ethnic fractionalization 1.576 .634 0.258
    Level of democracy .952 .018 0.013
    N 685    
    Wald Chi2 50.35    
    Prob > chi2 0.0000    
    Log pseudolikelihood -689.46077    

    ***p<.01, **p<.05, *p<.10; two-tailed test.

    Table 5: Cox Proportional Hazards Duration Model

      Haz. Ratio Std. Err. (Robust) P>z
    Intervention strength .779 .092 0.036**
    Power preponderance 1.173 .108 0.084*
    Population (log) 1.09 .040 0.015**
    Intervention by non-state actor .529 .160 0.036**
    GDP (log) 1.001 .050 0.970
    Former French colony .985 .243 0.952
    Former British colony .643 .146 0.052*
    Ethnic fractionalization 1.504 .582 0.291
    Level of democracy .950 .017 0.005***
    N 685    
    Wald Chi2 43.07    
    Prob > chi2 0.0000    
    Log pseudolikelihood -693.61093    

    ***p<.01, **p<.05, *p<.10; two-tailed test.

    Table 6: OLS Regression Results for Battle Deaths

      Coef. Std. Err. (Robust) P>z
    Intervention strength -1223.641 707.820 0.084*
    Strength Side A 1004.069 482.604 0.037**
    Strength Side B 584.511 305.188 0.055*
    Population (log) 103.323 180.755 0.568
    Intervention by non-state actor 2341.294 1347.983 0.082*
    GDP (log) -497.8663 278.275 0.075*
    Former French colony -3309.803 943.608 0.000***
    Former British colony -1620.978 879.377 0.065*
    Ethnic fractionalization 5209.787 2144.404 0.015**
    Level of democracy 47.922 64.493 0.457
    N 485    
    Wald Chi2 24.38    
    Prob > chi2 0.0066    

    ***p<.01, **p<.05, *p<.10; two-tailed test.

    Table 7: OLS Regression Results for Battle Deaths

      Coef. Std. Err. (Robust) P>z
    Intervention strength 781.880 411.202 0.057*
    Power preponderance -593.290 362.215 0.101
    Population (log) | 99.593 179.302 0.597
    Intervention by non-state actor 1856.893 1129.392 0.100
    GDP (log) -575.7668 308.942 0.062*
    Former French colony -3559.191 1013.408 0.000***
    Former British colony -1803.37 819.680 0.028**
    Ethnic fractionalization 4634.31 1931.644 0.016**
    Level of democracy 45.131 69.639 0.517
    N 485    
    Wald Chi2 24.08    
    Prob > chi2 0.0042    

    ***p<.01, **p<.05, *p<.10; two-tailed test.

    Table 8: Cox Proportional Hazards Duration Model

      Haz. Ratio Std. Err. (Robust) P>z
    Intervention strength 1.918 .391 0.001***
    Strength Side A .618 .097 0.002***
    Strength Side B .696 .128 0.050*
    Population (log) 1.140 .065 0.021**
    Intervention by non-state actor .565 .283 0.256
    GDP (log) 1.206 .115 0.050*
    Former French colony 1.156 .578 0.772
    Former British colony 1.125 .457 0.771
    Ethnic Fractionalization .204 .168 0.055*
    Level of democracy .895 .028 0.001***
    N 417    
    Wald Chi2 53.50    
    Prob > chi2 0.0000    
    Log pseudolikelihood -270.18219    

    ***p<.01, **p<.05, *p<.10; two-tailed test.

    Table 9: Cox Proportional Hazards Duration Model

      Haz. Ratio Std. Err. (Robust) P>z
    Intervention strength .794 .130 0.161
    Power preponderance 1.540 .242 0.006***
    Population (log) 1.141 .064 0.019**
    Intervention by non-state actor .598 .304 0.314
    GDP (log) 1.186 .109 0.063*
    Former French colony 1.046 .498 0.923
    Former British colony 1.274 .477 0.518
    Ethnic fractionalization .172 .138 0.029**
    Level of democracy .891 .027 0.000***
    N 417    
    Wald Chi2 48.56    
    Prob > chi2 0.0000    
    Log pseudolikelihood 270.54922    

    ***p<.01, **p<.05, *p<.10; two-tailed test.

    Table 10: OLS Regression Results for Battle Deaths

      Coef. Std. Err. (Robust) P>z
    Intervention strength -877 .5101 0.277
    Strength Side A 1012.469 713.727 0.156
    Strength Side B 597.817 477.037 0.210
    Intervention by non-state actor 2739.983 1738.512 0.115
    GDP (log) -710.294 471.037 0.132
    Former French colony -3344.552 1339.842 0.013**
    Former British colony -1836.469 1210.534 0.129
    Ethnic fractionalization 7370.877 3630.607 0.042**
    Level of democracy 119.549 100.959 0.236
    N 347    
    Wald Chi2 16.78    
    Prob > chi2 0.0794    

    ***p<.01, **p<.05, *p<.10; two-tailed test.

    Table 11: OLS Regression Results for Battle Deaths

      Coef. Std. Err. (Robust) P>z
    Intervention strength 845.612 446.056 0.058*
    Power preponderance 780.145 544.644 0.152
    Population (log) 98.072 221.650 0.658
    Intervention by non-state actor 2645.762 1696.71 0.119
    GDP (log) -653.687 440.717 0.138
    Former French colony -3350.763 1329.86 0.012**
    Former British colony -2220.899 1214.964 0.068
    Ethnic fractionalization 7368.643 3612.119 0.041**
    Level of democracy 116.001 98.191 0.237
    N 347    
    Wald Chi2 16.02    
    Prob > chi2 0.0665    

    ***p<.01, **p<.05, *p<.10; two-tailed test.