How Can You Tell You’re About to Have a Civil War?
The signs aren't what you might think
The Iraqi Kurd’s vote for independence in September 2017 sent shockwaves through the region. The day after the vote, the Iraqi military conducted a military exercise with Turkey. The government in Baghdad grounded international flights in and out of Kurdish Regional Government territory. And the Turkish government threatened to shut down the pipeline carrying oil from northern Iraq.
The Iraqi government recently launched a military offensive to retake the lucrative Kirkuk oil fields that the Kurds had won during their fight against Islamic State in 2014. The offensive largely has been bloodless. This was apparently due to political deals between the main Kurdish opposition party and the central government in Baghdad.
But the specter of yet another full-scale civil war hangs over Iraq. Sayyed Hassan Nasrallah, the leader of Lebanese Hezbollah, has been explicit in expressing worry about another war. “It will open the door to partition, partition, partition,” he said, referring to the referendum. “[P]artiton means taking the region to internal wars whose end and time frame is known only to God.”
The threat of civil war isn’t just a problem in the Middle East.
The union for the Spanish police said that the tensions in Barcelona following a vote for Catalan independence were nearing levels not seen since a terror campaign by Basque separatists in the 1980s. And Catalan separatist leader Charles Puigdemont’s public references to the Spanish civil war in the 1930s has prompted many Catalan protestors to hold signs referring to the episode.
And now Spanish government has taken the drastic step of beginning to process to revoke Catalonia’s autonomy.
It’s not just separatist movements that are causing many to worry about civil war. The massive unrest in Venezuela has given rise to several groups attempting to mount an armed rebellion against the government. A group of around 20 men even attempted to raid a military base.
Will Iraqi Kurdistan, Barcelona and Venezuela devolve into full-scale civil war?
It’s difficult to tell.
Most Americans see civil wars as the result of political and economic repression or ethnic and religious animosity. And with good reason. Most of the civil wars familiar to Americans do stem, at least in part, from these causes.
ISIS got its start in Iraq following political repression against Sunnis by the Shiite government, according to many analysts. A similar story of ethnic and political repression could explain the current civil war in Syria, as well.
But there are many countries that experience the same repression and animosities, but have never experienced a civil war.
In fact, there have been more than 200 civil wars since the end of World War II, with many complex causes.
The field of political science has amassed dozens of studies asking one simple question. “What explains the onset of civil war?” In fact, it’s probably the most researched question in political science.
At top — Kurdish peshmerga fighters. Matt Cetti-Roberts photo. Above — Venezuela’s ruling socialist party relies on bike-riding militias, or colectivos. Joka Madruga photo via Flickr
Economics, politics and ethnicity
The first major studies of civil war onset occurred in the early 2000s. Several of the most important studies during this period came to the rather jarring conclusion that neither political grievances, nor ethnic and religious animosities, are a significant cause of civil war.
Leading the charge on this thesis were two professors from Oxford University, Paul Collier and Anke Hoeffler, and two professors from Stanford University, James Fearon and David Latin.
Collier and Hoeffler’s main contention is that civil wars begin when rebels have the “economic opportunity” to actually wage war. They found little evidence in their work that ethnic-religious animosities or other political grievances significantly contribute to the onset of civil war.
Their 2004 study “Greed and Grievance in Civil War,” statistically tests a number of explanations for civil war onset using a dataset covering the period between 1960 to 1999 and consisting of 161 countries. In order to statistically test whether economics, politics and ethnicity affect civil war onset, Collier and Hoeffler had to come up with numeric representations for these factors.
At this point in history very few scholars had tried to come up with ways to numerically represent these factors. Just like with cancer research, when scholars first approach a question the data is usually not ideal.
So Collier and Hoeffler used what are called “proxy variables.” These are somewhat distant numeric representations of things like “ethnic exclusion,” that are difficult to measure directly.
For instance, a more direct way to measure “ethnic exclusion” would be to read historical accounts of such exclusion or consult experts, and come up with a way to numerically represent the level of exclusion over time.
But at the early stages of research it’s difficult to justify spending a large amount of time and resources seeking out such evidence. Especially since, in Collier and Hoeffler’s case, they had to cover 161 countries over four decades.
So, as a proxy for “ethnic exclusion” the authors note whether one ethnic group makes up 45 to 90 percent of the population in a country. They assumed such an ethnic breakdown would mean the dominant ethnicity would exclude the minority.
Imperfect. But it’s a start. Other variables share a similar distance from what is actually being measured.
Remember that Collier and Hoeffler contend that rebellions are most likely to begin when rebels have the economic opportunities to actually fight the state. Their proxy variables for “opportunity” are the size of the diaspora, external government support, cost of recruitment and procuring military equipment, strength of the state military, social cohesion of rebels and access to natural resources.
Collier and Hoeffler assume that the diaspora population provide the financial means for rebels to contest the state. This does occur. For instance, Irish Americans funded the IRA in the 1990s. But Collier and Hoeffler measure diaspora size in the United States only, and they don’t bother to check whether or not the diaspora population actually contributes to rebels.
The level of external government support is inarguably important for rebel groups getting off the ground. Unfortunately, Collier and Hoeffler’s measure of “external support” simply notes whether the Cold War was going on during the period in question. Of course, even insurgents in Africa regularly received support from external states after the Cold War.
But what’s even more problematic is the fact that Collier and Hoeffler don’t actually account for whether external aid was available to a rebel movement before the onset of civil war.
The level of secondary schooling, and mean per-capita income are meant to proxy for the ease with which rebel leaders can conjure an army. The logic being that it should be easier to recruit from poor and uneducated communities.
The existence of previous wars is meant to proxy for cheap military equipment. However, Collier and Hoeffler don’t measure whether rebels can actually buy cheap military equipment left over from the last war.
Military strength of the state is measured by how mountainous a country is. “Rough terrain” is supposed to weaken any military. Collier and Hoeffler do not account for jungles or dense forests. And they don’t actually know whether rebels fight in the mountains.
Weak social cohesion is a proxy for a country’s level of ethnic and religious diversity. The higher the diversity, so the logic goes, the more difficult it is to start a rebellion—and vice-versa. Ethnic diversity is measured by the ethno-linguistic fractionalization index, which “measures the probability that two randomly drawn people will be from different ethnic groups.” The authors construct a similar measure for religion.
Finally, Collier and Hoeffler’s most important variable for “opportunity” is natural resource access. Quite simply, this measure is the ratio of primary commodity exports to GDP.
Four measures are used for political grievance. “Ethnic or religious hatred, political repression, political exclusion and economic inequality.” Ethnic or religious hatred is measured as a function of diversity and political polarization. Political repression relies on existing data from the Polity III database and Freedom House. Political exclusion has already been discussed.
The statistical tests reveal that the only “political grievance” variable that helps explain civil war onset is “political exclusion.” The authors argue that since the statistical tests show that ethnic diversity actually decreases the probability of rebellion, the “political exclusion” variable—measured by ethnic dominance—should actually proxy for “opportunity.”
Still it’s important to keep in mind that none of these grievance proxies actually measure whether those making up rebel organizations experienced any of the “political grievances” variables.
A different result comes from the “opportunity” variables. Low secondary schooling, per capita income, and economic growth rate, all increase the likelihood of civil war. These factors could logically be seen as “political grievance.” But since economic inequality doesn’t predict rebellion, the authors don’t think so.
Rough terrain—a proxy for weak state militaries—somewhat contributes to civil war onset. Diasporas have a significant impact on the likelihood of rebellion.
But the strongest finding comes from Collier and Hoeffler’s “natural resources” variable. “At peak danger (primary commodity exports being 33 percent of GDP), the risk of civil war is about 22 percent, while a country with no such exports has a risk of only one percent.” Disaggregating the data further, they find that oil exports has the strongest effect on likelihood of rebellion.
A year earlier Fearon and Latin had published an article they titled “Ethnicity, Insurgency and Rebellion.” They found that Collier and Hoeffler’s “natural resource” measure did not significantly explain the onset of rebellion.
Interestingly, Fearon and Latin use almost the same proxy variables for political grievances and ethnic animosities as Collier and Hoeffler in their statistical tests. And even many of the same economic variables.
Just like in the case of Collier and Hoeffler, Fearon and Latin didn’t find any evidence that ethnic-religious animosities or political grievances predict rebellion.
However, Fearon and Latin did find that the best predictor of civil war onset is the presence of a weak state military. The main proxies for “weak military” are “rough terrain,” “political instability,” low per capita income and population size.
Fearon and Latin’s study uses the same measure for “rough terrain” as Collier and Hoeffler’s. “Political instability” is essentially defined as having a significant shift in regime type over a short period of time. For example, going from a democracy to an autocracy in two years.
To be clear, the main difference between Collier and Hoeffler’s conclusion and Fearon and Latin’s is that the latter didn’t find significant evidence that “natural resources” predict the onset of civil war.
Fearon and Latin agree that economic variables matter. For instance, both studies found that low per capita income is a strong predictor of rebellion. However, the authors argue “that economic variables such as per capita income matter primarily because they proxy for state administrative, military and police capabilities.”
Specifically, they say that poor states have a hard time funding their militaries. And adding political instability and rough terrain to the equation creates fruitful grounds for a rebellion.
This is because an unstable state has a difficult time implementing a coherent counterinsurgency plan, and mountainous terrain makes finding the rebels more difficult. Large population size also helps the rebels by providing a place to hide and the ability to acquire resources.
As may be clear to some readers, Fearon and Latin are listing off the basic conditions necessary to mount a guerilla war. Guerrilla wars are a specific type of war characterized by a weak insurgent fighting a militarily superior state. Guerillas usually avoid direct battle, and instead carry out sabotage and ambushes against state forces.
Reliance on a population for shelter, food and other resources is also common for guerillas. Particularly rural populations where rough terrain makes it difficult for a state military to fight.
Fearon and Latin are quite explicit in their assertion that all civil wars are guerilla wars.
However, some civil wars are fought between symmetrically powerful armies with conventional capabilities. For instance, many of the rebels in the current Syrian civil war maintain armor, artillery, etc. And a small number of civil wars feature symmetrically powerful groups fighting each other using mainly light weapons. For instance, the Liberian civil war.
Thus, it’s unlikely “rough terrain” causes civil war onset because it facilitates guerilla war. In fact, a 2010 study distinguished between the types of civil wars and found that “rough terrain” didn’t help predict what type of civil war was fought.
The surprising findings in the Fearon and Latin and Collier and Hoeffler studies led many academics to rigorously test their conclusions.
In fact, the Fearon and Latin study became the “standard model” for statistically-based studies on civil war onset in political science. This means that many of the studies that came after Fearon and Latin’s used many of the same political, economic and ethnic variables in their studies.
But future studies re-defined variables or added new ones with the aim of finding better explanations for civil war onset.
Many of these studies sought to answer two questions in particular. “Does reliance on natural resources in a state really increase the likelihood of civil war onset?” Also … “do political and/or ethnic grievances really not matter in explaining rebellion?”
Afghan Hazara troops of the Iranian Fatimioun Brigade entering the Shi’a enclave of Nubol and Zahra in western Aleppo in Syria on Feb. 4, 2016. Photo via Tom Cooper
Collier and Hoeffler’s thesis has been a target for many academic studies
UCLA Professor Michael Ross’ 2006 article titled “A Closer Look at Oil, Diamonds, and Civil War,” argues that Collier and Hoeffler’s measure of “natural resources”—the ratio of commodity exports to GDP—is poor.
Ross says “that conflict, or the anticipation of conflict,” causes much of the manufacturing sector to slow, shut down or move oversees. Resources extraction, however, is less likely to change substantially because “extractive industries often function in enclaves or remote regions, making them easier to secure” by the state. And these industries obviously can’t move overseas.
Thus, “the anticipation of conflict should reduce a country’s GDP (the denominator) more than its resource exports (the numerator)—thus producing a higher resource-exports-to-GDP ratio,” according to Ross.
In basic terms, since the conflict effects the ratio of GDP to commodity exports, the measure can’t independently tell us much about why civil wars start.
Instead, Ross measures “natural resources” by the country specific per capita rents from minerals, and diamond production per capita.
He chooses to focus on minerals and diamonds for two reasons.
First, these resources are said to have a substantial impact on civil war onset by other scholars. Second, several other studies had already constructed fairly accurate measures of mineral and diamond production. This gave Ross the ability to improve further on the measures instead of having to start from square one.
The measure for mineral rents comes from World Bank data. Ross distinguishes between non-fuel minerals — bauxite, copper, iron, lead nickel, phosphate, tin, zinc — and minerals that produce fuel such as oil, gas, hard coal and lignite. He also distinguishes between onshore and offshore oil rents using the PETRODATA dataset.
Ross uses trade journals to estimate diamond production per capita in each country in his dataset. He then distinguishes further between kimberlite and alluvial diamonds using the Diamond Resources dataset.
Kimberlite diamonds require fairly advanced mining equipment to extract, while alluvial diamonds can easily be extracted using primitive instruments, such as shovels. When you think of “conflict diamonds” you’re thinking of alluvial diamonds.
Ross’ statistical test includes many of the variables for politics, economics, and ethnicity from the Fearon and Latin study. Ross finds “partial support for the proposition that resource wealth facilitates civil wars by funding rebels.” As expected, rents from fuel producing minerals found onshore is a fairly strong predictor of civil war onset, but rents from offshore fuel is not.
However, kimberlite diamonds helps explain civil war onset, but alluvial diamonds do not. Ross says that this is a fairly damning result for Collier and Hoeffler’s argument, since alluvial diamonds are easily extracted, and hence should be the most cost-effective natural resource for emerging rebel groups.
But Ross even cautions against interpreting too strongly his results on onshore fuel-based minerals and kimberlite diamonds. He notes that the statistical results are based off of a small number of civil war onsets. This means that the results are extremely sensitive to different measures of variables.
For instance, Ross explains that if the civil wars in oil-rich Russia or diamond rich Democratic Republic of Congo did not exist, kimberlite diamonds and fuel-based minerals would no longer be a good predictor of civil war onset.
In fact, at least two of the civil war onsets in Democratic Republic of Congo in the 1990s should be dropped from the dataset. Those civil wars got off the ground because various states in the region desired regime change, and thus funded and fought alongside Congolese rebel groups.
U.N. and Congolese troops celebrate gains made against the M23 rebel group in 2013. U.N./Sylvain Liechti photo
Politics and ethnicity
Gauging how economics effects civil war onset clearly isn’t easy.
But gauging how politics and ethnicity effect civil war onset is even more difficult. The main variable used to measure ethnic diversity by Fearon and Latin and others—the ethno-linguistic fractionalization index—can’t tell researchers which ethnicities are actually salient in state politics or a rebellion.
Freedom House and the Polity database do provide a way to gauge the level of political repression in a country. But these measures don’t tell us whether the population that makes up a rebel group actually endured political repression.
Since Fearon and Latin and Collier and Hoeffler made their controversial claims about politics and ethnicity in civil war, a few studies have tested their argument.
One of the most substantial of these studies came from professors Lars-Erik Cederman, Andreas Wimmer and Brian Min in their 2010 article “Why Do Ethnic Groups Rebel?”
Cederman and colleges spent nearly two years building a massive dataset of every politically relevant ethnic group in 155 countries between 1945 and 2005.
In order to feasibly carry out this task they “relied on the expert input of nearly one hundred students of ethnic politics to assess formal and informal degrees of political participation and exclusion along ethnic lines.”
Ethnic groups are classified “as politically relevant if at least one political organization claims to represent it in national politics or if its members are subjected to state-led political discrimination.” In gauging the level of political inclusion or exclusion, the authors focus on the executive branch of countries. This includes “the presidency, the cabinet and senior posts in the administration, including the army.”
Cederman and colleagues asked experts to place the level of political participation for ethnic groups into three categories of power. Whether the ethnic group holds absolute control of political power. Whether an ethnic group is part of a power-sharing regime. Finally, whether an ethnic group is excluded from political power.
They account for variation in these categories over time. This is important, because they expect that a sudden loss of power will make an ethnic group more likely to rebel.
Their conclusion is actually fairly simple. The more excluded a group is from power, the more likely it is to rebel. Further, low per capita GDP increases the probability that excluded groups will rebel.
The authors explain, “that in poor countries governments lack the resources necessary to co-opt leaders of protest movements that fuel resentment because of their exclusion from central government. Since the central state is all-decisive in poor countries, groups that lack representation and are marginalized in the distribution of state resources, government jobs and public goods may have greater motives to take up arms.”
Ethnic groups that endured a recent downgrade in political power are also more likely to rebel. For example, if a group goes from sharing power to being excluded.
The existence of a prior conflict in recent years increases the likelihood of rebellion. And the larger the excluded ethnic group’s population size, the more likely it is to rebel. The authors explain that this is because larger ethnic groups have greater military capacity to contest the state.
Interestingly, whether a country is a democracy or an autocracy doesn’t affect the authors’ conclusions.
It is important to note that study focuses on ethnic civil wars only. This means that combatants “explicitly pursue ethnonationalist aims and recruit fighters and forge alliances on the basis of ethnic affiliations.” Focusing on ethnic wars to judge the role of ethnicity in civil war onset is somewhat problematic. You’re more likely to find that ethnicity does matter in civil war onset.
However, restricting a study to ethnic civil wars allows researchers to be more accurate in their conclusions about that type of war. But their findings are less likely to explain other types of civil wars.
Cederman and colleagues admit that those “who are familiar with the qualitative scholarship on ethnic conflict may not be surprised that representatives of ethnic groups with less access to power are more likely to challenge the government.”
However, they point out that the most influential studies on civil war onset—Fearon and Latin, and Collier and Hoeffler—”contend that ethnic diversity is unrelated to the likelihood of conflict and conclude that ethnicity thus does not matter for understanding war and peace.”
This is an important point, since the Fearon and Latin and Collier and Hoeffler studies are taken seriously by policy makers.
Peshmerga fighters in Khanaqin in Iraq. Matt Cetti-Roberts photo
Policy implications for Iraq
If anything, this article should show the difficulties of conducting statistical studies on civil war onset.
The most consistent finding across much of the academic literature is that civil wars are likely to occur in poor states with large populations.
Most of the other explanations for civil war onset, including ethnic grievance, political instability, economic inequality, natural resources, etc., are sensitive to alternate measures.
This sensitivity means that any academic would be wary of making any overarching prescription to a policymaker based on this research. Especially on specific situations, such as the current tensions between the Iraqi Kurds and the Iraqi government.
But the research does, at the very least, provide us with tools to analyze the situation.
For instance, Vice News recently reported that a lucrative oil deal with a Russian oil giant “gave Kurdish leaders confidence that they could weather the post-referendum storm, according to sources closely involved in the region’s politics and oil sector.”
This means that the Kudish Regional Government’s decision to seek independence largely falls in line with Collier and Hoeffler’s “natural resources” thesis. In fact, the KRG did plan to vote on independence in 2014, but backed down in the face of U.S. pressure. One of the main differences between 2014 and September of 2017 was the Russian oil deal.
Thus, if Collier and Hoeffler are right, then the KRG’s loss of the oil wells during the recent Iraqi offensive in Kirkuk should keep the Kurds from mounting a rebellion.
On the other hand, if Fearon and Latin are correct, then the military dominance of the Iraqi state should stave off civil war. This means it would be bad idea for the U.S. military to stop arming and training the Iraqi military.
And if Cederman and colleagues are correct, then a downgrade in political power of the Kurds in the Iraqi government might be the tipping point. The fact that Iraq has been at war for years, and the Kurdish population is quite large, increases the probability of civil war, according to the Cederman study. But further political inclusion of the Kurds in the Iraqi central government may be the best defense against conflict. At least for now.