|Why Class Inequality Breeds Coups but Not Civil Wars
Michigan State University
Forthcoming in Journal of Peace Research
Does class inequality increase the risk of civil war? I posit that inequality between social classes affects civil wars through two pathways: (1) it heightens the risk of political violence by fueling distributive conflicts; and (2) it reduces structural coup-proofing, which, in turn, increases the capacity of the military to fight insurgents. Combining these effects implies that the net effect of class inequality on civil war is ambiguous. Although class inequality increases the propensity for violence, in unequal countries political violence rarely takes the form of wars because such countries have strong militaries. Class inequality, however, breeds other forms of political violence. In particular, it increases the likelihood of military coups. The two effects of class inequality reinforce each other in the case of coups: it simultaneously stirs distributional conflicts and increases the capacity of the military to mount coups by reducing coup-proofing. Using data on 128 developing countries between 1960 and 2008, I find that while class inequality fosters coups, it has no discernible effect on civil wars. I also provide evidence consistent with my causal mechanisms: (1) inequality creates greater threat to the rulers by fueling political instability; (2) inequality reduces structural coup-proofing; and (3) structural coup-proofing increases the likelihood of civil war.
Does inequality increase the risk of civil war? While inequality has been found to increase the likelihood that a country experiences diverse forms of political violence, such as coups, riots, demonstrations and revolutions (Alesina & Perotti 1996; Roe & Siegel 2011), empirical studies about the effect of inequality on civil war have been largely inconclusive (Collier & Hoeffler 2004; Fearon & Laitin 2003).1 The lack of consistent relationship between inequality and civil war challenges the expectations of a celebrated literature which predicts that inequality makes war more likely notably because it fuels distributive conflicts (Davies 1962; Gurr 1970).
One of the main explanations for theses ambiguous results is that ‘greed’ (or opportunity) is a better predictor of conflict than ‘grievances’ (or motivation) (Collier & Hoeffler 2004). However, inequality does not only capture grievances. For example, at any income level, increasing inequality decreases the opportunity cost of potential recruits – who are typically members of the lower class. This, in turn, reduces the cost that political entrepreneurs – who are often (albeit not always) members of the economic elites – must pay to finance a rebellion. Hence, inequality creates motives and opportunities for insurgency.
I provide and test an explanation that accounts both for the absence of a consistent relationship between inequality and civil war, and for the strong effect of inequality on other forms of political violence, particularly military coups. My argument focuses on the concept of interclass inequality, defined as inequality between the economic elites and the masses.2 I posit that class inequality affects civil wars and other forms of political violence through two pathways. First, it heightens the risks of political violence – including civil wars – by generating conflicts over the distribution of resources. I refer to this as its direct effect.
Second, class inequality has an indirect effect, which operates through its influence on the structure of the military. This effect builds on the important work of Svolik (2012), who argues that unequal countries tend to have stronger militaries. Rulers of such countries are more likely to face threats from the population, which forces them to delegate more power to the military in order to quell unrest. This logic further suggests that class inequality reduces the ability of rulers to use structural coup-proofing tactics, since such tactics reduce the capacity of the military to repress. Structural coup-proofing refers to changes to the structure of the military, for example by dividing the military into multiple units, which inhibit its capacity to stage coups (Powell 2012). Such tactics differ from non-structural coup-proofing strategies, such as increasing military expenditure, that do not reduce the capacity of the military to stage coups by altering its structure but instead increase the incentives of the army to support the regime.
The indirect effect has different implications on different forms of political violence. On the one hand, having a strong military reduces the likelihood of violence when, as for civil wars, the military is the main agent that fights against those that initiate violence. Militaries that are not weakened by structural coup-proofing are better at fighting wars, which deters potential insurgents. Thus, interclass inequality has two confounding effects on civil wars. It increases the propensity for conflict (direct effect) while simultaneously increasing the capacity of the military to fight insurgents (indirect effect); leading to an ambiguous net relationship.
On the other hand, class inequality should have a particularly strong positive effect on forms of violence that are instigated by the military, such as military coups. In such instances, the two effects of class inequality reinforce each other: it simultaneously stirs distributional conflicts (direct effect) and increases the capacity of the military to mount coups (indirect effect). This argument helps to explain why many unequal countries are more prone to coups than civil wars. For example, between 1950 and 2006, Bolivia – one of the world’s most unequal countries – experienced nine successful and three failed coups but not a single civil war.3
This paper contributes to the small but growing literature that treats the risks of coups and civil wars as related threats to the regime (e.g., Roessler 2011; Bodea, Elbadawi & Houle Forthcoming). As pointed out by Fearon (2004), ‘both coups and peripheral insurgencies are strategies for using violence to take power’ (p.289). Similarly, Roessler (2011) argues that ‘coups and insurgencies represent alternative antiregime technologies’ (p.303). Roessler (2011) finds that ethnic exclusion increases the likelihood of civil war but reduces that of a coup. Bodea, Elbadawi & Houle (Forthcoming), for their part, find that partial factional democracies are more likely to experience coups, civil wars and riots. This paper adds to this literature and suggests that as inequality increases, challengers become more likely to opt for coups as opposed to insurgencies.
I test this argument with a dataset covering 128 countries between 1960 and 2008. Using three indicators of inequality – two of which proxy for class inequality – I first show that while inequality increases the likelihood of coups, it has no discernible effect on war. Finally, I examine the causal mechanisms driving the aggregate relationships. Consistent with my argument, I find that: (1) inequality creates greater threats to rulers by fueling political instability; (2) inequality reduces the use of structural coup-proofing; and (3) structural coup-proofing increases the likelihood of civil war. These findings suggest that the weak effect of inequality on civil wars may at least partially be explained by its effect on structural coup-proofing.
Apart from providing an answer to the puzzle discussed above, I contribute to the literature in multiple ways. First, some recent studies have found that inequality breeds civil war (Bartusevi
ius 2014; Baten & Mumme 2013). My findings on the effect of inequality on civil wars contribute to this ongoing debate. My tests differ from previous ones notably because they use measures of inequality that proxy for class-based inequality.
Furthermore, few authors have tested the effect of inequality on coups. Svolik (2012) finds that the relationship is inverted U-shaped, but his analysis is restricted to autocracies. Collier & Hoeffler (2005) test the effect of inequality on coups but their sample is limited to sub-Saharan Africa. Some authors show that inequality increases the likelihood of democratic reversals, which usually (but not always) take the form of coups (e.g., Houle 2009). But, to my knowledge, this is the first paper to test the effect of inequality on coups in a world sample that contains both democracies and autocracies. While testing my causal mechanisms, I also provide evidence that inequality fuels many forms of political instability, such as riots and revolutions. These findings corroborate those of previous studies using different measures of inequality, such as capital shares (e.g., Alesina & Perotti 1996).
Finally, I contribute to the growing literature on the causes and consequences of structural coup-proofing. This is the first study to empirically demonstrate that inequality reduces structural coup-proofing. I also show that structural coup-proofing fosters war. Although Powell (2014b, 2014c) reports a similar finding, my analysis differs from his in a number of ways. First, Powell's (2014c) analysis is limited to sub-Saharan Africa. Second, Powell (2014b) finds that structural coup-proofing only fosters civil war in autocracies that are facing a large coup threat. My analysis generalizes this finding to all regions and finds that the relationship is not conditional on the risk of a coup.
Previous literature on inequality and civil wars
The idea that inequality stirs political violence dates back to at least Karl Marx, who argued that inequality between the owners of the means of production and the laborers fuels class warfare. A similar argument was later restated and refined by the deprivation theory (e.g., Davies 1962; Gurr 1970) and other authors (e.g., Midlarsky 1988; Muller 1985; Russett 1964). More recently, Boix (2008), whose argument explicitly focuses on class inequality, argued that inequality increases the likelihood that the masses initiate a civil war in order to install a regime in which it dominates.
There are two main groups of empirical studies testing the effect of inequality on civil wars. The first focuses on individual and household inequality (i.e. the overall level of inequality in the society), usually measured with Gini coefficients. The findings of these studies have been inconclusive. While some authors find no relationship (Collier & Hoeffler 2004; Collier, Hoeffler & Rohne 2009; Fearon & Laitin 2003), others find a positive relationship (Bartusevi
ius 2014; Baten & Mumme 2013; Boix 2008; Russett 1964). Moreover, Besançon (2005) finds that inequality has a negative effect on ethnic wars but a positive one on revolutions. Most studies that find no relationship use income Gini indexes, while those that find a positive relationship usually use non-income Gini coefficients. For example, Bartusevi
ius (2014) looks at inequality in education attainment and Baten & Mumme (2013) at inequality in physical height. Bartusevičius (2014) also provides evidence based on income inequality, but the estimated relationship is usually only significant at the ten percent level.
Despite the differences in their findings, these empirical studies share several limitations. First, as explained above, most studies employ Gini indexes (whether based on income or not). One problem with using Gini coefficients is that they do not capture inequality between social classes – or between any particular groups – but individual and household inequality. The arguments of Marx, on which most of the literature (at least implicitly) builds, and many of the more recent studies (e.g., Boix 2008) unequivocally center on the role of inequality between the masses and the elites. Gini indexes, however, are opaque and do not map neatly into any specific cleavage. Different income distributions can lead to the same Gini coefficient values. A high Gini coefficient could be the result of a high level of inequality between, say, the 30th and 10th percentiles of the income distribution or between the 99th percentile and the rest. These two types of income distributions may have very different consequences on political stability.
Moreover, most authors using income Gini indexes rely on datasets that employ national surveys.4 The problem is that different countries use different methods and definitions in their surveys, which renders cross-national comparisons unreliable. For example, some surveys are based on net and others on gross income. Finally, most authors use datasets with many missing values, often around 60 percent of the full population, which creates a serious sample-selection bias.
In response to studies looking at individual and household inequality, a second group of authors have examined the effect of horizontal inequality – inequality between culturally defined groups (usually ethnic groups) – on civil wars (e.g., Cederman et al. 2011; Gubler & Selway 2012; Houle 2015; Østby 2008). These studies tend to show that horizontal inequality does breed conflict. Of course, an ethnic group is not the same as a social class and so these studies do not answer the question of the effect of interclass inequality on wars. As discussed below, my argument about why inequality between the elites and the masses has little effect on conflicts is not directly applicable to the question of horizontal inequality. Among other contributions, the analysis below tests the effect of inequality on civil war using measures that more closely proxy for class inequality.
Class inequality, civil wars and coups
I posit that class inequality has two effects on political violence: a direct and an indirect effect. Neither effect, on its own, is entirely new to the literature. Both draw heavily from existing studies. Taken together and applied to the question of the effect of interclass inequality on civil war, however, they offer an explanation for the absence of a consistent relationship between inequality and war. They also explain why inequality has a much stronger effect on other forms of political violence. Below, I explain these two effects and discuss inequality’s net effect (the combination of the two effects).
Class inequality fuels distributive conflicts
As implied by the arguments of Marx and others, class inequality fuels distributive conflicts. I argue that class inequality fosters all forms of political violence through this mechanism. Below, I first focus on coups and then on civil wars.
Imagine a situation in which a regime is dominated by either the elites or the masses. For the sake of the argument, let us assume that the regime represents the interests of the elites (although the logic of the argument is unaltered if it represents those of the masses). Moreover, the regime can either be democratic or authoritarian. For simplicity, I assume that it is authoritarian. A political entrepreneur can mount a coup. In order to successfully establish him/herself as the new ruler after a (successful) coup, he/she is assumed to have to form a political coalition that supports the new regime. That is, a new regime is more easily established and maintained when a political entrepreneur can rely on a solid source of support. No political regime, even if authoritarian, can be instated and maintained without a coalition (Bueno de Mesquita et al. 2003; Geddes 1999).
Inequality between social classes widens the gap between the preferences over economic policies of members of the masses and the elites (Acemoğlu & Robinson 2006; Boix 2003, 2008). For example, the masses may want to increase redistribution or to adopt poor-friendly policies, such as public education and health, while the elites may oppose such policies. In the present example (in which the elites are politically dominant and the regime is authoritarian), interclass inequality creates incentives for members of the masses to support potential coup plotters in exchange for policies more in line with their own preferences. Class inequality can thus produce political coalitions on which coup leaders can rely. This, in turn, facilitates the establishment and maintenance of the new regime; hence increasing the likelihood of a coup.5 Notice that the relationship is driven by the level of inequality between the class in power (here, the elites) and the one that is not (masses), not by the country's level of household/individual inequality, which also reflects inequality within each class.
Although the example above pertains to left-wing coups – staged to represent the interests of the masses – waged against autocracies, the argument extends to right-wing coups and coups staged against democracies. In fact, Boix (2003) and Acemoğlu & Robinson (2006) theorize that inequality between social classes induces right-wing coups against democracies. In a democracy the median voter is a member of the masses, and so the elites can stage a right-wing coup to reduce redistribution. A similar argument could be made that democracies that fail to redistribute sufficiently could fall victims to left-wing coups. Finally, although autocracies are usually ruled by the economic elites (Acemoğlu & Robinson 2006; Boix 2003), one could imagine a situation in which the dictator represents the interests of the poor. Interclass inequality increases the likelihood that such regimes get overthrown by right-wing coups aimed at reducing redistribution. Examples of left-wing coups carried out against autocracies include Burkina Faso (1983), left-wing coups staged against democracies Peru (1968), right-wing coups waged against autocracies Peru (1948), and right-wing coups staged against democracies Chile (1973) and Brazil (1964).
This is not to say that coups are always driven by class interests and that all coups are alike. Many coups are the results of purely intra-elites conflicts. In fact, the indirect effect discussed below pertains to how class inequality affects the strictly intra-elites dimension of coups. Instead, class inequality can produce conditions conducive to coups, notably by providing a source of support to coup plotters. Moreover, coup leaders need not be from the same social class as their supporters. They simply need to enact policies that are consistent with their preferences.
I identify three reasons why interclass inequality may have a direct positive effect on civil wars. The first is based on the argument of Boix (2008) and is closely related to the discussion on the effect of class inequality on coups above. Class inequality increases the gap between the preferences over economic policies between social classes, which increases the incentives of the class that is not in power to topple the regime through an insurgency and install a regime which would adopt policies more in line with its preferences. In other words, in the scenario above the class that does not hold office may simply resort to a rebellion rather than a coup. This reasoning is consistent with the quotes of Fearon (2004) and Roessler (2011) above, according to which coups and insurgencies are alternative strategies that political entrepreneurs can employ to usurp power.
My second argument is based on the mechanisms proposed by the deprivation theory and other authors discussed above (e.g., Russett 1964) that theorize that inequality breeds grievances. In short, according to this argument, inequality creates resentment among the masses, rendering them more willing to join insurgencies.
Finally, not only does class-based inequality produce grievances and increase the gap between the preferences of different social classes, but it also creates opportunities for insurgency. At any income level, increasing interclass inequality decreases the opportunity cost of potential recruits (which are typically members of the masses). This creates a large pool of young people willing to join a conflict in exchange for food, a roof, a meager salary and access to spoils. Keeping everything else constant, class inequality thus reduces the cost of hiring recruits, which decreases the cost of mounting a rebellion for political entrepreneurs, who are themselves usually members of the elites (Esteban & Ray 2011). According to this argument, class inequality can fuel war even if the aim of the rebellion is not to install a more equitable regime, and if the regime in place already adopts poor-friendly policies.
Class inequality reduces structural coup-proofing
Class inequality has also an indirect effect on political violence that operates through its effect on the structure of the military. Here, I build on Svolik (2012). It is important to note, however, that Svolik (2012) does not discuss the effect of inequality on civil war (which is the focus of this paper). Also, he does not consider the direct effect of inequality on coups, discussed above.
Svolik (2012) argues that, in autocracies, inequality increases the level of mass threat faced by dictators. When inequality between the elites – which are usually in power in autocracies (Acemoğlu & Robinson 2006; Boix 2003) – and the masses is high, the latter have more incentives to revolt and are more costly to coop, for example, through a political party (since their preferences differ more widely from those of the elites). Moreover, as implied by the seminal work of Meltzer & Richard (1981), inequality increases the benefits (in terms of redistribution) associated with democracy for the masses. Consequently, in autocracies, inequality between the masses and the elites increases pressure in favor of democratization by the former (Acemoğlu & Robinson 2006; Houle 2009). Such threats to the rulers are not limited to (and, in fact, rarely involve) civil wars. Pressure for democratization, for example, almost never takes the form of a full-blown civil war.
Threats from below, in turn, push rulers to grant enough power to the military to quell mass opposition (Svolik 2012). The same logic implies that class inequality precludes rulers from adopting structural coup-proofing measures – such as dividing the military into multiple units – because such tactics reduce the capacity of the military to repress. Structural coup-proofing has been demonstrated to substantially diminish the effectiveness of the military (Powell 2014a, 2014b, 2014c; Pilster & Böhmelt 2011, 2012). For example, a partitioned army is not as effective at repressing mass opposition as a united one. Paramilitaries can repress small-scale opposition. However, because of the sheer size of the masses, only the military has the manpower necessary to suppress the mass-based opposition. In the words of Svolik (2012), ‘when the opposition to a regime is mass based, organized, and potentially violent, the military is the only force capable of defeating it’ (p.127).6 Therefore, structural coup-proofing, by weakening the military, reduces the capacity of the regime to suppress mass threats. This argument pertains to inequality between the elites and the masses, because smaller groups can, for example, be put down by paramilitaries. Moreover, individual and household inequality – which also captures within-class inequality – may not have the same implications.
Although these forms of threats are often short-lived, they have lasting implications for the role of the military in a country’s domestic politics. Once a ruler chooses to rely on the military to eliminate a threat, he and his successors become ‘trapped’: they cannot undermine its power without putting themselves at risk of a coup. While it is possible for a ruler to reduce structural coup-proofing to counter mass-based threats, in the short-term it is much riskier (albeit possible) to increase it. Such policies necessarily threaten the interests of top military officials. A military that has the capacity to stage a coup may opt to do so if a ruler attempts to undermine its power. Consequently, the capacity of the military to intervene domestically and the extent to which a country relies on structural coup-proofing is largely (though not completely) exogenous to any particular ruler. Countries that are unequal are likely to have a tradition of militaries that are involved domestically and that are not weakened by structural coup-proofing. Rulers can always adjust the strength of the military and structural coup-proofing in a manner that reduces the level of mass threat they face. However, making the equivalent adjustments for coups is more difficult, since the very measures that reduce the capacity of the military to stage coups may themselves trigger a coup in the short-term.
Although, according to this argument, strong militaries originate while countries are authoritarian, even after democratization the army is likely to retain its capacity to intervene domestically (Cheibub 2007; Svolik 2012).7 Attempts at weakening the military may trigger coups no matter whether the ruler is democratically elected or not. As argued by Svolik (2013, p.770), ‘not only dictatorships but also new democracies that inherit such empowered militaries will therefore be at a higher risk of military intervention.’ Cheibub (2007) also demonstrates that democracies with strong military legacies are more likely to experience coups because armies retain their strong position after democratization.
The indirect effect has different implications on different forms of political violence. On the one hand, when the military is the main actor fighting against those that initiate violence, as in the case of civil wars, having a strong military reduces the likelihood of violence. Militaries that are not undermined by structural coup-proofing are also better at fighting civil wars, which decreases the likelihood that they will be instigated. As discussed above, structural coup-proofing has been shown to reduce the capacity of the armed forces to fight against international and domestic opponents. For example, Pilster & Böhmelt (2011) demonstrate that structural coup-proofing increases causalities in the domestic army during conflicts. Moreover, Powell (2014b, 2014c) shows that structural coup-proofing increases the probability of civil war, albeit, as pointed out above, only under some conditions.
On the other hand, when the military is the one that instigates violence, as in the case of military coups, interclass inequality increases the likelihood of violence through its indirect effect. The military is more likely to stage a coup if it is not weakened by structural coup-proofing. For example, Powell (2012) demonstrates that coup-proofing reduces the probability of coups. The indirect effect is thus about how class inequality affects the intra-elites dimension of coups. Importantly, class inequality may even affect coups that are not directly motivated by class interests, though only indirectly by enhancing the capacity of the military to carry them out.
Although the indirect effect builds on Svolik (2012), my prediction about how inequality’s impact on the structure of the military affects coups differs from his: he expects the relationship to be inverted U-shaped, whereas I expect it to be monotonic and increasing. It seems unlikely that countries with middling structural coup-proofing levels are more at risk of coups than those with no or few coup-proofing measures. Moreover, my predictions about inequality's net effect on coups are the product of the combination of the direct and indirect effects, whereas Svolik (2012) only considers the indirect effect.
But, why would rulers ever empower the military if it means that they become more vulnerable to coups, especially since coups are the most common threat they face? Although for most rulers military structures are largely inherited, some do reduce coup-proofing and strengthen the military. I argue that rulers are likely to do so under two sets of conditions. The first is when mass opposition imperils the survival of the regime in the very short-run. Second, a ruler may also opt to empower the military if it is a prerequisite to the establishment of the regime. Establishing an autocracy in spite of mass opposition may simply be impossible without the support of the military. Repressing mass opposition is a prerequisite to establishing a dictatorship, in which a few rule over the majority. In short, rulers may be willing to trade off the risk of being eventually ousted by the military in exchange for being able to establish their rule in the first place and/or remain in power when faced by immediate mass threats.
Combining the direct and indirect effects helps explain why results on the effect of inequality on civil war have been largely inconclusive. While class inequality increases the overall propensity for violence by breeding distributive conflicts (direct effect), it simultaneously deters potential insurgents by increasing the capacity of the military to fight (indirect effect); meaning that its net effect is ambiguous. The two effects, however, reinforce each other in the case of coups: class inequality, by accentuating grievances, creates coalitions on which coup plotters can rely (direct effect), and produces opportunities for coups by reducing structural coup-proofing (indirect effect). The net effect of class inequality on coups should thus be positive.
My argument does not necessarily contradict the findings of authors working on horizontal inequality and conflicts. While the direct effect is also relevant to horizontal inequality, the indirect effect is less directly applicable. It is not because a specific ethnic group is much poorer or much richer than the rest of the country that rulers will necessarily be forced to allocate extensive power to the military. This is particularly clear in the cases of small ethnic groups that can, for example, be repressed by paramilitary forces. The masses, however, because of its sheer size does impact the need to use the military. The indirect effect may be relevant in special cases, such as under ethnocracies in which a large group that is excluded from power is much poorer or much richer than the one in power. But, more generally, we should expect horizontal inequality to affect civil wars primarily through the direct effect; explaining why it has been found to foster war.