Abstract
Military coups have posed a persistent threat to political stability in Africa, undermining democratization efforts, igniting insurgencies, and leading to years of devastating military governance. Initial cross-national studies found little consistent evidence linking ethnicity to coups, leading recent formal and statistical work on coup risk and coup-proofing to largely ignore ethnic politics. This article, however, argues that in two important contexts of African political development—decolonization and democratization—ethnic politics are critical to understanding the occurrence of coups. Both case study evidence and statistical analysis of original data on African military history and ethnic politics reveal that practices of ethnic manipulation within security institutions have driven coup attempts. When leaders attempt to build ethnic armies, or dismantle those created by their predecessors, they provoke violent resistance from military officers.
On April 12, 2012, rebellious soldiers seized power in Guinea-Bissau. They quickly established control over the small capital city, captured the TV and radio stations, and detained both the interim president and the front-runner in the upcoming presidential elections. A week later, gun and mortar fire could still be heard in Bissau although power rested firmly in the hands of the new military junta. Promises were made of new elections and a return to constitutional rule. This is a familiar story for the citizens of Guinea-Bissau, whose military has time and again overthrown the government and then retreated to the barracks by holding elections—only to repeat the cycle. Since independence, no elected president has managed to serve out his full term.
Guinea-Bissau is not alone in its struggle to escape a pernicious cycle of military interventionism and build democratic institutions. Military coups pose a profound and persistent threat to political stability in Africa. While the last decade has witnessed an overall decline in coup incidences (Powell and Thyne 2011), they are still a far too common event in Africa. In the last five years alone, coups have been staged against the governments of Egypt, Eritrea, Guinea, Guinea-Bissau, Madagascar, Mali, Mauritania, and Niger. Since independence, over 215 coup attempts have been made across the continent, affecting forty-three countries. Nearly half of these have been successful, displacing civilian governments, undermining democratization efforts, and leading to years of devastating military governance. Compelling evidence exists that coups also ignite insurgencies by weakening the central government and thereby opening up opportunities for rebellion (Fearon 2003b, 280). In the midst of Mali’s March 2012 coup, for example, Tuareg rebels launched a powerful military offensive. They and Islamic rebel groups proceeded to capture much of the country before French intervention forces drove them back.
Given the impact that military coups have on political stability, civil war onset, and democratization, it is important to ask the following question: what causes African militaries to intervene so frequently in the domestic politics of their societies?
To answer this question, I depart from recent statistical and formal modeling works that have come to ignore ethnic politics as an important driver of military coups. Rather, I ground my theoretical claims in both the ethnic politics literature and the extensive case study literature on African coups that has long suggested the centrality of ethnic dynamics to this form of political violence. These theoretical claims are then evaluated using original data on African military history and original measurements of ethnic manipulation within security institutions.
I argue that when leaders attempt to build ethnic armies, or dismantle those created by their predecessors, they provoke violent resistance from military officers. At the critical juncture of independence, African leaders made important decisions over how they would ensure military loyalty. Where leaders conditioned military recruitment, promotion, and access to patronage on shared identity—despite existing diversity in the colonial officer corps—they tended to initiate violent cycles of ethnic coups and countercoups as officers resisted discrimination. I find that such countries experienced coups roughly four times as often as their counterparts in the postindependence period.
I then examine how these historical legacies have impacted democratization efforts in contemporary Africa. Many leaders did successfully build coethnic armies, either with the assistance of departing colonial powers or through a violent process of purges. Democratic elections in a multiethnic society threaten to empower new leaders who no longer ethnically match these existing armies and have strong incentives to dismantle them. When this occurs, the ethnically stacked officer corps fears the loss of their power, privilege, and patronage—and may react by seizing power. I find that when elections bring to power a new leader ethnically distinct from a previously constructed ethnic army, coup risk increases dramatically from under 20 percent to roughly 89 percent.
This article underscores the necessity of military reform in contemporary Africa. If democracy is to thrive in multiethnic societies, existing ethnic armies must be dismantled and national military institutions diversified. Yet, my work should also emphasize how difficult and dangerous this task will be. Ethnic armies do not stand idly by and allow for their own demise. Reformers should thus anticipate a struggle.
Previous Approaches
In the 1960s and 1970s, the spate of military coups across recently decolonized countries elicited the development of a statistical literature concerned with the structural risk factors underlying coup propensity. Such studies analyzed a wide range of potentially important variables, including the underdevelopment of political institutions, strength of civil society, degree of military professionalization, number of soldiers and their relative pay, economic development and growth rates, commodity dependency, past coup history, and ethnic diversity, among others. Yet, these works generated only two consistently statistically significant findings: that coups tend to occur more frequently in poor countries and where there is a pattern of past military coups (Jackman 1978; Jackman et al. 1986; Jenkins and Kposowa 1990; Johnson, Slater, and McGowan 1984; Kposowa and Jenkins 1993; Londregan and Poole 1990; Lunde 1991; McGowan and Johnson 1984; O’Kane 1981; Przeworski et al. 1996; more recently see Bazzi and Blattman 2011; Belkin and Shofer 2003; Collier and Hoeffler 2005; Galetovic and Sanhueza 2000).
These studies produced no systematic evidence, however, that ethnic politics matter to the production of military coups, even though many of them were deeply concerned with ethnic dynamics and included ethnic diversity and dominance measures in their regressions. Rather, they arrived at contradictory results with some finding significant correlations (Jenkins and Kposowa 1990; Kposowa and Jenkins 1993; Lunde 1991) and others no relationship (Collier and Hoeffler 2005; Jackman 1978; Jackman et al. 1986). This absence of systematic evidence linking ethnic politics to military behavior has led recent statistical and formal work on coups and coup-proofing to neglect ethnicity altogether, even as a control variable (see Biddle and Zirkle 1996; Galetovic and Sanhueza 2000; Lindberg and Clark 2008; Pilster and Böhmelt 2011; Powell 2012; Powell and Lasley 2012).
Yet, the case study literature has long suggested that ethnic dynamics are critical to understanding evolving military dynamics in many African states. For example, thick descriptions of the Ghanaian, Nigerian, and Sierra Leonean armies indicate that ethnic hostilities within these military institutions played an important role in inciting particular coup attempts (see, respectively, Hutchful 1985; Luckham 1971; Cox 1976). Based on accounts such as these, medium-N comparative scholars of African politics and ethnicity, such as Samuel Decalo and Donald Horowitz, have claimed that “ethnic coups” are an important and distinct phenomenon worthy of theoretical articulation (Decalo 1990, 1998; Horowitz 1985, 480-92).
If ethnic dynamics are as integral to African military coups as contextually rich case studies suggest, why then has ethnicity failed to consistently attain significance in cross-national work that stresses generalizability? The answer, I contend, lies in two limitations of the existing statistical literature. First, typical studies of coup risk arguably violate modeling assumptions of unit homogeneity. In the weakest version of this assumption, the same change in an independent variable generates a constant expected effect across all units of observation. Coups, however, are tactics. Military officers may seize power to achieve diverse political ends, from capturing the reigns of government to protesting a devastating war. Coups as tactics may thus be deployed for systematically different reasons dependent on the larger political context.
For example, civil wars can place extreme pressures on civil–military relations. Battlefield losses, lack of supplies, and war fatigue have all contributed to military uprisings—including against France’s and Portugal’s colonial counterinsurgencies in Africa and Mali’s underfunded war in the Sahel. When a country is fighting a war, we might therefore expect the adequacy of arms supply to significantly effect coup propensity. This same variable is unlikely, however, to exert an identical effect in the absence of war.
Modeling techniques such as interaction effects can account for this type of unit variation by integrating the observed heterogeneity into the model. This requires, however, that each context first be properly theorized in its own right, revealing how independent variables are likely to behave across contexts. Some recent work has moved in this direction: toward theorizing and testing hypotheses within a particular political context, rather than developing universal models that explain coups across all conceivable contexts (an ambitious task for any tactic). For example, Galetovic and Sanhueza (2000) develop a formal model to explain coups against autocratic regimes while Clark (2007) and Lindberg and Clark (2008) analyze coups against liberal democratic African regimes. These context-specific works, however, have yet to integrate ethnic politics into their models.
A further problem lies in the measurements of ethnicity employed in existing cross-national studies. Broad social measures of ethnic diversity and dominance have increasingly been criticized by the ethnic politics and conflict literatures as “meaningful operationalization[s] of ethnic politics” since they cannot capture the often complex causal pathways between ethnicity and outcomes, including political violence (Chandra and Wilkinson 2008, 527-36; Lieberman and Singh 2012, 1-2). Indeed, there is no compelling theoretical story linking such measures to the actions of military officers.
This article seeks to overcome these limitations and contribute to our understanding of how ethnic politics condition military coups and compound the challenges facing liberalizing African states in three important ways. First, I too take a step back from more universal statistical approaches and instead build a middle-range theory specific to the contexts of decolonization and, later, democratization. At the same time, I attempt to transcend the limitations of standard case studies by generating theoretical insights that are testable cross-nationally, even if constrained by context-specific scope conditions. Through this approach, I bring back the recently neglected centrality of ethnic politics to African civil–military relations and to our understanding of when African militaries are likely to overthrow elected leaders. I leave to future research, however, other vital contexts—such as coup dynamics during civil wars or military responses to social uprisings such as the Arab Spring—and the ultimately desirable theoretical integration of diverse contexts into a more universal model.
Second, this article also contributes to the commendable effort by scholars of ethnic politics to better theorize causal mechanisms linking ethnicity to outcomes and to develop appropriate measures capturing them. For example, the Ethnic Power Relations (EPR) data set measures the exclusion of ethnic groups from executive power (Cederman, Wimmer, and Min 2010; Wimmer, Cederman, and Min 2009); the EVOTE project constructs measures for the ethnic basis of political parties and how they implicitly and explicitly appeal to identity groups (Chandra 2009); and Wilkinson is developing a measure of “ethnic balance” in postcolonial military institutions (Chandra and Wilkinson 2008, 537-45). In this article, I similarly develop original measures to capture how ethnic groups were incorporated into colonial officer corps, whether postindependence leaders recruited officers on an ethnic basis, whether past leaders had created ethnic armies, and whether a constitutional transfer of executive power also entailed a change in the ethnic identity of leadership.
Finally, I construct an original data set on African military history and ethnic manipulations within African security institutions. These data were culled from extensive sources including archival documents from Africa, Europe, and the United States, historical newspapers, secondary scholarly accounts, and tertiary reference books and qualitative data sets.
A Theory of Ethnic Coups
I make two related arguments, both grounded in an understanding of ethnic politics and competition, to explain the relative propensity of African militaries to seize state power in the contexts of decolonization and democratization. The first argument analyzes independence as a critical juncture for security institutions and posits the development of coup traps as a function of how early African leaders chose to build military loyalty. In broad strokes, I argue that where leaders chose to condition military recruitment and promotion on shared ethnic ties, despite a diverse colonial officer corps, they tended to set their countries on dangerous paths of violence. The second argument examines the legacies created by this earlier historical period and their effects on the new critical junctures brought about by the third wave of democratization. Many African leaders did successfully construct coethnic armies, either with the help of departing colonial powers or through a violent process of military purges. I argue that these ethnic armies view electoral liberalization (in a multiethnic context) as anathema to their corporate interests. Elections open up the possibility that leaders from new ethnic groups may obtain power and then seek to dismantle the ethnic armies they inherit. To protect their entrenched position of power and access to patronage, these armies then undermine processes of democratization by blocking elections or overturning their results.
Colonial Military Practices and Decolonization
Military loyalty plays a critical role in regime stability, particularly the loyalty of the officer corps. Rank-and-file soldiers typically do not command others (except non-commissioned officer [NCOs]). Thus, they cannot easily organize large-scale resistance and, when they do revolt, rarely accomplish more than a short-term mutiny. On the other hand, officers—including very junior officers and (rarely) NCOs—can use their position in the command hierarchy and ability to issue largely uncontestable orders to move troops and engage in combat, without revealing their antigovernment intentions to their subordinates. 1 Officers topple governments. 2
As African decolonization progressed, troubling events in the region made this general concern for loyalty acutely urgent. Almost immediately after Sudanese independence in 1956, civil war broke out as black Christian and animist southerners resisted incorporation into a discriminatory Arab north (who had already disbanded southern army units). Similar tensions along racial, ethnic, and religious fault lines divided most Sahelian states—divisions created during the fourteenth-century Arab conquest and long-standing trans-Saharan slave trade and reinforced by colonial policies that granted political autonomy to northern Arab regions while disproportionately developing infrastructure, education, and economic opportunities in southern districts. In 1959, with the tacit consent of Belgian officials, the “Rwandan Revolution” overthrew the Tutsi monarchy. The political upheaval was accompanied by pogroms in which several thousand Tutsis were massacred, highlighting the vulnerability of ethnic minorities in new states (Newbury 1998, 13). In August 1960, the Congolese government disintegrated after rank-and-file soldiers revolted against continued Belgian officership, spreading fears of state failure across the continent (Meditz and Merrill 1993). Facing potential ethnic violence, civil war, and state disintegration, African leaders were hard-pressed to find an effective means to tie soldiers to the state.
Colonial military practices provided a ready model of ensuring loyalty through racial and ethnic manipulations. While native soldiers comprised their rank and file, colonial armies were officered almost exclusively by white Europeans. This remained the case on the eve of independence and, in many countries, even after the formal transfer of sovereignty. Throughout colonial Africa, race defined who could be trusted to command.
Conceptions of ethnic loyalty also shaped rank-and-file recruitment practices. British martial race doctrine, developed first in India, stipulated that some ethnic groups were naturally more politically reliable and suited to combat and military discipline than others. Frederick Lugard, the high commissioner (1900–1906) and then the governor-general (1912–1919) of Nigeria, further developed this doctrine in Africa. He proscribed that one ethnic group should dominate the colonial army, provided that that group was both politically reliable and unable to dominate civilian political life (i.e., the local civil service). It was a policy of “making the politically strong militarily weak and the politically weak militarily strong” (Adekson 1979, 154).
Lugard’s model was widely emulated. By such logic, the Kamba and the Kalenjiin came to dominate the Kenyan battalions of the King’s African Rifles, while the Kikuyu were purposely excluded (Parsons 1999, 58). In Uganda, northerners, and especially the northwestern Acholi and Teso ethnic groups, were targeted for military recruitment (Keegan 1983, 598-600). In addition to the British, Belgian colonial administrators also established ethnic recruitment quotas based on racialized perceptions of martial prowess (Adekson 1979, 160-161). Even the French, who initially avoided deterministic group-based evaluations of military fitness, were drawn to such practices over time. Tribes from the sparsely populated forest zones were eventually seen as undesirable soldiers because they lacked a strong martial tradition (Echenberg 1991, 63).
These colonial practices normalized the idea that race and ethnicity were linked to military loyalty. Colonial empires relied on their own ethnic kin to officer far-flung armies, despite the cost and high mortality rates. Lugard’s model of ethnic recruitment further embraced the notion that state security institutions were best left in the hands of loyal ethnic groups. When it came time for independence leaders to Africanize their officer corps, many turned to this established model of ethnically based loyalty. Relying on mechanisms of ethnic patronage and ethnic affinity to ensure political reliability, they conditioned military recruitment and promotion on shared ethnicity.
Before delving into how this choice to build stability on ethnic grounds conversely led to great instability, it is necessary to address an important endogeneity concern. Those leaders facing divisive ethnic violence may have recruited coethnics into the military to ensure the loyalty of soldiers in the event of ethnic war. If this were true, then ethnic violence would drive both the loyalty choice and subsequent political instability: military loyalty would be endogenous to existing cycles of ethnic violence. In order to address this concern, I collected data on two types of ethnic violence between the end of World War II and the year of independence for each country: ethnic riots and violence conducted by ethnic political parties against other ethnic groups. Full results are presented in the Online Appendix. The correlation between pre-independence acts of ethnic violence and the choice to build military loyalty on ethnic grounds is −0.14 with a χ 2 of 0.2768, which is statistically insignificant (p > .5988).
The Development of Ethnic Coup Traps
Choosing ethnicity as the foundation for military loyalty was, counterintuitively, dangerous for political stability—despite its very purpose of ensuring such stability. Building a coethnic army often required not just discriminatory recruitment, but the dismissal or marginalization of non-coethnic officers. Leaders employed two primary tactics to build such coethnic security institutions: (1) restructuring the officer corps of the existing army along coethnic lines or (2) constructing coethnic parallel military institutions, such as presidential guards and militias, and then disarming the regular army. Both present a similar threat to existing out-group officers, motivating many initial coup attempts and leading to cycles of violence wherein ethnic factions within the officer corps attempted to seize power and purge one another from the ranks. 3
Attempting to place a monopoly over the legitimate means of violence in the hands of a sole identity group inherently threatens excluded groups. They face both discrimination and declining relative power. Recent research in ethnic politics also suggests that conditioning access to a key state institution on ethnic identity heightens the salience of ethnic divides. By institutionalizing ethnicity and endowing it with imposed importance, governments prime individuals to interpret facts and events in terms of ethnicity. Indeed, the more identity shapes one’s life chances and access to resources, the more useful it becomes as a cognitive device for reducing uncertainty. Questionable actions and isolated incidents of violence, that may not be motivated by ethnic differences, are then more likely to be interpreted through an ethnic lens. This, in turn, escalates communal hostilities (Hale 2008, 33-40; Lieberman and Singh 2012, 2-6). We thus expect soldiers facing ethnic discrimination to feel threatened, primed to interpret events in ethnic terms, and faced with a shrinking window of opportunity in which to reverse their fortunes.
In these circumstances, violent resistance is likely. An initial attempt to restructure the military along ethnic lines, or perhaps even the anticipation of such restructuring, would prompt out-groups in the officer corps to protect themselves. Whether they succeed or fail, an ethnically motivated coup attempt—even if inspired by a defensive logic—aggravates and threatens officers of other groups, who may then attempt coups of their own. And, as officers are arrested, purged, and killed in the aftermath of each coup, the dynamics of fear and mistrust deepen. A cycle of tightly linked reactionary and counter reactionary violence is thus triggered within the military—an ethnic coup trap.
Initial resistance to the construction of an ethnic army, however, would only take the form of a military coup where the colonial state had previously built diversity into the officer corps. Departing colonial powers sometimes colluded with incoming governments to construct an ethnically homogenous army, matching the ethnicity of the nascent leadership, prior to decolonization. This was the case, for example, in both Mauritania (Maure) and Rwanda (Hutu). Here, the existing officers already benefited from a system of ethnic privilege within the military and were unlikely to protest its continuation.
Figure 1 summarizes the causal mechanisms that link military loyalty to instability within the context of African decolonization.
Causal mechanisms.
Sierra Leone and Cameroon
A brief comparison of decolonization processes in Sierra Leone and Cameroon highlights these dynamics. Early independence leaders in both countries chose to build military loyalty on ethnic foundations, and yet they experienced different coup trajectories: while Sierra Leone underwent four coup attempts in the decade following independence, Cameroon escaped from any such military intervention for over twenty years. The key difference was the state of the officer corps prior to decolonization. In Sierra Leone, the British recruited officers in a broadly representative fashion across ethnic groups while in Cameroon, French officials assisted Ahmadou Ahidjo’s government in building a coethnic Fulani/Peuhl military prior to the formal transfer of sovereignty. Thus, Ahidjo faced no resistance to policies that continued Fulani/Peuhl dominance of the military while Sir Albert Margai’s attempt to construct a coethnic Mende army met with stiff resistance from non-Mende officers, resulting in a series of coups and countercoups.
Sierra Leone gained formal independence in 1961 under the leadership of Prime Minister Milton Margai, an ethnic Mende, and the Mende dominated Sierra Leone People’s Party (SLPP). Milton Margai, however, made no actual choices over military loyalty, leaving Sierra Leone’s new national army entirely in the hands of the British, who continued to control training, recruitment, and promotion (Keegan 1983, 517-18). 4 Africanization of the officer corps proceeded at a slow, but steady pace and the British recruited from across ethnic groups. In 1964, 26 percent of officers were Mende, 12 percent Temne, and 64 percent were drawn from other, mostly northern groups (Cox 1976, 54).
Things began to change with the death of Milton Margai in 1964 and the ascension of his brother Sir Albert to the prime ministership. In the face of rising opposition to government policies at home, and in the aftermath of the 1966 coup in nearby Ghana, in mid-1966 Sir Albert dismissed the British and assumed control over the military. He quickly began restructuring the army, principally recruiting his coethnic Mendes (Cox 1976, 73; Kandeh 1992, 93). Within a year, Mende representation had doubled to 52 percent of the officer corps (Cox 1976, 75). This move toward ethnic homogenization provoked immediate discontent among non-Mende officers. In January 1967, a serious coup plot was foiled only when the prospect of high-level assassinations gave one of the plotters, who was related to the minister of education, cold feet. 5 The following month, nine Temne officers were purged from the military, further increasing Mende domination (Cox 1976, 100-6).
Scheduled elections in March 1967, however, gave non-Mende officers hope that democratic politics could halt this trend toward Mende domination of the security sector. The opposition party, the All People’s Congress (APC)—who represented northern tribal groups including the Temne, the Susu, the Loko, and the Mandingo—was popular and nationally competitive. And although neither side played fair—the incumbent SLPP tampered with ballots and engaged in voter intimidation while the opposition’s leader, Siaka Stevens, had trained a paramilitary force over the border (Allen 1968, 218; Kposowa 2006, 37)—the APC eked out a win by the narrowest of margins.
Sir Albert, however, backed by Mende soldiers, refused to concede and re-seized power. The following month, Temne and Limba soldiers staged a successful countercoup and purged almost all of the existing Mende officers before handing over power to Siaka Stevens and the APC (Keesings: World News Archive 1968). Stevens then attempted to further narrow the ethnic base of the military, from a coalition of northern factions that had opposed Mende domination to his own coethnic Limba. In reaction, Temne officers attempted to seize power in 1971. They failed and Stevens’s regime was soon able, through additional purges, to achieve Limba homogeneity of the army and police forces. Stevens also created a paramilitary unit, known as the Internal Security Unit (ISU)—also predominantly Limba—that was used to inflict violence on domestic opponents and eventually grew to be more powerful and better equipped than the regular army (Cox 1976, 106-7; Davies 2000, 352-53; Horowitz 1985, 477-79; Kandeh 1992, 94).
The choice to ethnically homogenize the army thereby provoked a series of coups and countercoups which only ceased through the eventual creation of coethnic security institutions, combined with the use of external Guinean troops for added protection. Political stability was thus achieved under Stevens and the APC regime. The price of this stability, however, was high: the nearly total ethnicization of the state, reliance on paramilitary units, rampant corruption, erosion of civil liberties, and economic stagnation under Stevens’s and then Momoh’s rule set the stage for later civil war (Kposowa 2006).
In contrast to Sierra Leone, Cameroon experienced great political stability in the immediate postcolonial period: from independence until his retirement in 1982, Ahmadou Ahidjo’s government was never threatened by a coup attempt. And yet Ahidjo made broadly similar choices to Sir Albert. He too chose to ground military loyalty in shared ethnic identity while transforming the state into a personal autocracy. The key difference was that in the late 1950s, French colonial authorities helped Ahidjo to build a coethnic officer corps prior to decolonization and thus enabled him to continue ethnic recruitment practices without military protest.
Fears of communism played an important role in conditioning French policies in Cameroon. A resistance movement against French rule began in 1955, led by the Union des Populations de Cameroun (UPC), which adopted a broadly socialist ideology. The armed wing of this movement, the Armée de Liberation Nationale Kamerun (ALNK), turned to various communist governments, including both China and the Soviet Union for military assistance. The escalating Cold War context led both the French and the British—who governed part of colonial Cameroon—to perceive these ties as evidence of an extreme security threat to the region. 6 Rather than compromise with the UPC, French authorities excluded the movement and its political sympathizers, comprised mainly of southerners and particularly the Bamiléké and Bassa ethnic groups, from the civilian government that inherited power upon decolonization (Atangana 2010, 14; Minority Rights Group 2010).
Instead, the French preferred to work with a more “moderate” Cameroonian government that would shun communism and pledge to maintain strong ties with the French economic community. In 1958, the French high commissioner installed Ahmadou Ahidjo and his northern Fulani/Peuhl-based political party in power (Atangana 2010, 74). In the same year, France also devolved significant powers to local governments across Africa, including the construction of national armies. Although Cameroonian troops had previously long served within the regional French Equatorial Africa force, prior to this time no territorial army had existed. Moreover, the first cohort of native officers did not graduate from the newly established Yaounde cadet school until 1960–1961. 7 Ahidjo thus inherited a blank canvas from which to construct an officer corps while also benefiting from continued French protection. He chose to recruit these new officers extensively from the north, particularly from his own ethnic Peuhl group as well as from the allied Fulani (Minorities at Risk Project 2009; Minority Rights Group 2010).
Thus, decolonization entailed handing over power to a northern-dominated and Western-friendly government, backed by an ethnic army. Ahidjo continued to recruit Fulani and Peuhls extensively into both the civil service and the security forces until he left office over twenty years later (Minorities at Risk Project 2009). He never faced a coup attempt.
Ethnic Armies and Democratization
The ethnic coup traps that developed from decolonization processes could not sustain themselves indefinitely. Some ended when rebels overthrew a military paralyzed by factional infighting. Uganda, for example, experienced a remarkable nine coup attempts during the Idi Amin and second Obote regimes, many of which resisted ethnic manipulations by those leaders (Horowitz 1985, 486-92). The coups ended when rebel forces under Yoweri Museveni took over the state. Some rebel groups, such as the Tutsi-based Rwandan Patriotic Front, were drawn from particular ethnic constituencies, leading to new ethnic armies when they gained power. Other coup traps came to an end through Pyrrhic victory: as in Sierra Leone, when one ethnic faction finally managed to purge all others from the officer corps. Thus, more countries joined the ranks of those who had, with colonial collaboration, built ethnically loyal militaries from the beginning. Much of Africa was thus left with a legacy of ethnic armies (see also Enloe 1975, 1980; N’Diaye 2001).
This historic tendency to base military loyalty on ethnic foundations is critical to understanding contemporary struggles over African democratization.
8
In Africa, most states are highly diverse with no single ethnic group constituting a majority. The opening of political competition and establishment of electoral processes—regardless of the extent or depth of other democratic reforms—increases the likelihood that executive power will change hands between leaders from different ethnic backgrounds. This otherwise normatively desirable state of affairs poses a danger to political stability where past leaders had stacked the military with their coethnics. Under a new leader, who no longer shares their identity, these military officers will fear a decline in their power and privilege. The new leader may wish to emulate his or her predecessor and create his or her own ethnic army, purging the existing one and recruiting coethnics in their place. Or the new leader could be committed to nonethnic, inclusive state institutions and military diversification. Either set of policies requires dismantling the inherited ethnic army. Such restructuring threatens existing officers, who now face strong incentives to defend their positions of privilege. They may do so by blocking democratization or by violently reversing the outcome of elections, deposing new leaders, and restoring their coethnics to power. Figure 1 also summarizes the causal mechanisms by which electoral transfers of power can lead to coup attempts.
Cameroon: After Ahidjo
Returning to Cameroon, the aftermath of Ahmadou Ahidjo’s retirement illustrates the danger to political stability posed by existing ethnic armies during electoral leadership transitions, even ones of questionable democratic character. 9 In 1982, facing severe ill health, Ahidjo retired from office and was succeeded by Paul Biya, the current prime minister and a Christian southerner from the ethnic Bulu group. In January 1984, Biya secured his presidency by winning his first full term in a general election. Immediately thereafter, he sentenced Ahidjo to death in abstentia while simultaneously announcing his decision to transfer Fulani/Peuhl soldiers from the Republican Guard, an elite paramilitary force still dominated by Ahidjo’s coethnics, to other military units (Keesings: World News Archive 1984).
The very next day, threatened soldiers of the Republican Guard mounted a coup attempt against Biya. The rebels took control of the radio station, attacked the presidential palace in Yaoundè with artillery, seized the airport, and severed communication links. It took nearly four days of intense fighting for loyal troops to put down the rebellion. Subsequently, the Republican Guard was immediately disbanded, its loyal members placed under the command of the chief of the National Gendarmerie, and its disloyal members purged (Keesings: World News Archive 1984).
Using the failed coup attempt to his advantage, Biya then moved forward with discriminatory hiring and promotion policies of his own, within both the civilian government and the military. Over the course of his reign, which continues today, southerners have come to dominate both politics and the military (Minority Rights Group 2010). In particular, members of Biya’s own southern Bulu group, as well as members of the closely related Beti group, disproportionately hold key positions in the military (Minorities at Risk Project 2009). Stability was thus reachieved through renewed ethnic matching policies.
Data Collection and Coding Procedures
To evaluate whether the general historical experiences of African countries support these hypotheses, I compiled existing data on Military coup attempts and coded original cross-national data for the following, theoretically significant independent variables: Prior ethnic violence, Loyalty choice, Ethnically matched officer corps, Ethnic change in leadership transitions, and Prior ethnic stacking. The data were culled from a wide variety of sources including archival documents, 10 newspaper articles, 11 reference books and tertiary qualitative data sets, 12 and secondary scholarly accounts.
Military Coup Attempts
To construct the dependent variables, I cross-referenced McGowan’s (2003) data set on African coups with the Archigos database on political leaders (Goemans, Gleditsch, and Chiozza 2009) and expanded that data to cover the years 1952 to 1955 and 2002 to 2012. A minimum of two reports citing evidence of actual military violence, military occupation of a government building, or military occupation of a communication or transportation center count as a coup attempt.
Loyalty Choice
If the initial leadership purposefully recruited officers and/or created a presidential guard or personal militia from members of his own ethnic group—and, in some cases, their known allies—then loyalty choice was coded as Ethnic. The creation of a coethnic presidential guard, in particular, indicates that the leader is basing military loyalty on ethnic foundations since this tactic later enables the ethnic restructuring or disarmament of the regular army. On the other hand, if the first leader recruited officers from a diverse cross section of society, loyalty choice was coded as Inclusive.
Ethnically Unmatched (or Matched) Officer Corps
This variable builds on codings of both the ethnic heterogeneity of the officer corps and the ethnic identity of the independence era leader. If an ethnic group dominated the officer corps, and the postindependence leader shared that identity, then this variable is coded as Ethnically matched. If there was no preexisting officer corps, then no officers could rebel against its restructuring along ethnic lines. Since this situation entails the same theoretical prediction (no rebellion), it is also coded as matched. In all other cases, the variable is coded as Ethnically unmatched.
Coding individual identity presents a number of challenges. Individuals can emphasize aspects of their identity in strategic and flexible ways (Posner 2005). Leaders can also marry outside of their own ethnic group and utilize the identity of their spouses to foster political loyalty (Londregan, Bienen, and Walle 1995, 6). Nonetheless, leaders usually self-identified—and were identified by their contemporaries—as having a single-ethnic identity. And while spousal affiliations could play a role in the political game writ large, those who built coethnic armies chose their own group and usually exclusively so (with some exceptions for traditional ethnic allies or neutrals). Where mixed identity did occur, then if any group dominated the officer corps and that group matched part of the leader’s identity, the variable was coded as matched.
Defining when a particular ethnic group dominates the officer corps is difficult. Dominance should imply such a majority that no other group has a realistic chance of successfully seizing power. There is no clear numerical line, however, above which all cases would meet the qualitative standard. Dominance, as defined previously, depends on many factors, including how many identity groups have representation in the military and how they are distributed across the branches of military service. I have thus stayed with more qualitative understandings: did credible observers at the time think that the officer corps was dominated by a particular group?
Bias in source materials then becomes an important concern. Perhaps observers had a vested interest in portraying colonial military recruitment practices in a certain light—as discriminating against their group or, conversely, in downplaying their group’s representation. Such tendencies could introduce significant distortion into the data. Careful attention to sources and their potential biases can alleviate some of this concern. Most countries were coded based on official military intelligence reports from the French and British archives. When such documents were unavailable, I turned to the accounts of sociologists and historians writing about these militaries in close temporal proximity to decolonization.
Ethnic Change (or Continuity)
For the democratization data, it is important to differentiate between elections and electoral transfers of power. Autocratic regimes routinely hold uncompetitive elections, known as electoral authoritarianism, that merely rubber stamp a longtime leader’s hold on power. In order not to confuse autocratic elections with at least partially democratic ones, I thus limit the universe of cases to elections that produced a turnover in executive power as well as other constitutional transfers of power (e.g., after the natural death of a leader).
The ethnicity of each leader before and after the constitutional or electoral transfer of power was then coded at the highest level of ethnic aggregation. 13 If the ethnic identity of the leader prior to the transition was different from that of the leader assuming power, the power transfer was coded as an Ethnic change. Otherwise, it was coded as Ethnic continuity.
In no case was regional identity used as the basis of coding, even though region plays an important role in the politics of many African countries—particularly in Sahelian states such as Benin, Nigeria, and the Sudan where north–south divides are arguably more important than ethnic differences in shaping the struggle for power. While such contextual awareness in coding is desirable, it also increases the risk of fitting theory to the evidence rather than evaluating hypotheses according to the data. For this reason, I follow the stricter coding procedure.
Moreover, excluding regional considerations likely biases the results against the proposed theory. In Africa, the ethnic homelands of powerful groups do not tend to cross important administrative regional divides—which were largely constructed by colonial practices that treated the groups separated by these boundaries differently. Thus, recoding by regional identity would rarely alter a categorization of ethnic continuity into one of ethnic change. Rather, we would only expect to find the reverse—codings of ethnic change that do not involve a difference in region and thus, when coded by region, would become instances of ethnic continuity. We do not expect to observe coup attempts in such cases where regional identity is highly salient and power transfers between leaders from the same region. Thus, coding by region would eliminate several “false negatives” as cases originally incorrectly predicted to have coups would instead be successfully predicted as stable. Therefore, the more stringent coding rule, which excludes regional identity, biases the results in a downward direction—increasing our confidence in positive findings.
Prior Ethnic Stacking
This variable was coded 1 if the leader prior to the constitutional transfer of power had stacked the military with his coethnics and 0 otherwise. Similar concerns over coding identity, defining dominance, and identifying source material bias arose as with the Ethnically unmatched officer corps and Ethnic change variables and were handled in the same manner.
Empirical Analysis
In this section, I test the two primary implications of my theoretical framework using different data sets—one covering the postindependence period and the other inclusive of all electoral and other constitutional transfers of executive power. Following independence, choosing ethnic loyalty despite a diverse officer corps should result in an ethnic coup trap and thus greater instability. Later, during a constitutional transfer of power, if there is an ethnic change in leadership—and the previous leader had stacked the military with his coethnics—we also expect a coup attempt.
Descriptive Data
Table 1 presents the postindependence frequency of coup attempts broken down by loyalty choice and the diversity of the officer corps created under colonial rule. For each cell, I calculated the average number of coup attempts, across countries, in the twenty years following independence. 14 The results are compelling: countries with an unmatched officer corps, whose leaders chose to build ethnic armies, experienced on average four times as many coups as other countries did.
Average Number of Coups within Twenty Years of Independence.
Note: Bivariate negative binomial (coup count ∼ unmatched × ethnic cell) = 1.4319 (p ≤ .001).
Ordinary least squares (OLS) = 2.9812 (p ≤ .001). Missing Cases: Burkina Faso, Cape Verde, Central African Republic, Comoros, Democratic Republic of the Congo, Gambia, Guinea, Mauritius, São Tomé and Príncipe, and the Sudan.
Turning to the context of democratization, from 1950 to 2007, there were seventy-nine cases of power changing hands from one African leader to another by constitutional means. The majority of transitions (fifty-nine) were the result of electoral processes, usually routine elections within the context of an established democracy or as part of a planned democratization. The twenty nonelectoral transitions stemmed from a variety of circumstances, mainly the natural death (eleven) or retirement (four) of the previous leader.
Many of these power transfers were followed closely by violent military reactions: 33 percent experienced a coup attempt within four years (41 percent of electoral transitions). 15 Periodizing the data by decade illustrates an important temporal pattern: as the third wave of democratization swept across Africa in the 1990s, the number of constitutional leadership transitions skyrocketed, from eight (three electoral) in the 1980s to twenty-nine (twenty-seven electoral) in the 1990s. Yet, so too did the number of coup attempts undermining those transitions, from 2 or 25 percent (2 or 67 percent of electoral) in the 1980s to 11 or 38 percent (10 or 37 percent of electoral) in the 1990s. And while the record has improved in the 2000s, the problem has by no means disappeared. The Online Appendix contains a full breakdown of the reasons for each constitutional transfer of power as well as their periodization over time with resulting coup attempts.
Approximately 57 percent of these constitutional power transfers involved a change in the ethnicity of leadership. Table 2 depicts the relationship between ethnic changes in leadership, the tendency of previous leaders to build coethnic armies, and coup attempts. Where leaders had successfully constructed coethnic armies and constitutional processes then brought to power new leaders from different ethnic backgrounds, those ethnic armies reacted by seizing power—75 percent of the time.
Coups after Constitutional Transfers of Power (percentage of cases).
Note: Bivariate logit (coup ∼ ethnic change × prior ethnic army) = 2.8904 (p > .0001). Missing observations: Cape Verde (91, 01), Comoros (96, 98, 06), Liberia (97), Mozambique (86), São Tomé and Príncipe (91, 01), and the Sudan (86).
Regression Analysis
These cross-national patterns lend preliminary support to the theories developed in this article. They do not, however, control for potentially confounding variables, test alternative explanations, or rigorously estimate uncertainty. I thus turn to regression analysis. 16 Two distinct models are leveraged to further test the hypotheses.
First, I have argued that independence leaders choosing to build ethnic armies despite diverse officer corps should experience greater relative instability. To test this implication of the theory, I use the total count of military coup attempts in the twenty years following independence. Since the variance-to-mean ratio of this dependent variable (2.7) suggests over-dispersion, a negative binomial model is employed. 17
Second, I have argued that countries whose leaders have employed ethnic stacking policies in the past are vulnerable to military reactivity when processes of democratization bring new leaders to power from different ethnic groups. I analyze this hypothesis using logit regression on the same coup variable constructed for the descriptive democratization data: whether a coup attempt occurred in the four years following a constitutional transfer of executive power. This analysis is run on the following two versions of the data: (1) all constitutional changes in leadership and (2) only electoral successions of power.
Given the small size of each data set and its low power, only a limited number of variables and interaction effects could be included. I thus use category dummies to test the theory (based on Tables 1 and 2). For the decolonization model, I constructed indicators for a choice of ethnic loyalty given an unmatched officer corps (Ethnic and unmatched), ethnic loyalty with a matched officer corps (Ethnic and matched), and inclusive loyalty with an unmatched officer corps (Inclusive and unmatched). For the democratization models, I constructed similar indicator variables for the following categories: an ethnic change in leadership given an existing ethnic army (Ethnic change and stacking), an ethnic change in the absence of a prior ethnic army (Ethnic change and no stacking), and ethnic continuity in leadership given an ethnic army (Ethnic continuity and stacking). This leaves those countries with a matched colonial officer corps who chose to construct an inclusive military (decolonization data) and those countries who experienced ethnic continuity given an inclusive officer corps (democratization data) as the residual categories.
The control variables were also carefully selected based on theoretical significance. Complete descriptions and summary statistics are contained in the Online Appendix. Based on extant studies, controls were included for relative wealth (Ln GDP/k), 18 Prior coups, and ethnolinguistic fractionalization (ELF; Ethnic diversity; Roeder 2001). Pre-independence Ethnic violence was also included in the decolonization models to control for any effects that such past violence may have continued to exert on political stability, even though loyalty choices were not determined by it.
It is also reasonable to think that economic shocks, understood as sharp downturns in a country’s economic well-being, may increase coup risk. As public support for the incumbent government declines, recessions can create opportunities for many types of violence, including coups (Alesina et al. 1996; Galetovic and Sanhueza 2000). Systematic preindependence economic data were unavailable and thus could not be included in the count model where it would be appropriate. For the democratization models, an Economic shock variable was coded 1 if, in any year during the four-year period following the leadership transition, the country experienced a negative growth rate of 1 percent or more and 0 otherwise. 19 Three percent, 5 percent, and 10 percent economic shock variables were also coded in a parallel fashion to capture increasing levels of crisis severity. An alternative measure using the country’s lowest (i.e., worst) current account balance (CAB) as a percentage of gross domestic product (GDP) was also constructed. Large current account balance deficits tend to cause (or at least reflect) greater financial struggles as well as a government’s inability to pay its employees—including the military.
Clark (2007) and Lindberg and Clark (2008) have argued that the legitimacy accrued through processes of political liberalization may serve to inoculate regimes against military overthrow. This insight is drawn from liberal democratic theory, which posits that democratic regimes enjoy legitimacy in both the eyes of the general population and among important elites such as military actors. Thus, as a state travels from authoritarianism through the various stages of democratic transition and liberalization toward consolidated democracy, its legitimacy should correspondingly grow, decreasing its susceptibility to military overthrow (Lindberg and Clark 2008, 89). Trajectories of increasing liberalization should thus correlate to reduced coup risk while trajectories toward authoritarian regression may have the opposite effect.
Measuring legitimacy, however, is inherently difficult as legitimacy is at once a subjective and contextually dependent phenomenon (Clark 2007, 143). Survey data, such as that provided by Afrobarometer, may ultimately provide the best measure but are currently unavailable across enough countries and years to systematically measure trajectories of legitimacy. Alternatively, I employ the following imperfect proxy in the democratization models: if theories of democratic legitimacy are correct, then general trajectories of state legitimacy should track changes in the rights and freedoms afforded to society. I thus include a measure, Change in Civil Liberties, that is coded as the difference in a state’s Freedom House civil liberties score between the year in which a constitutional power transfer occurred and the end of the observation window. Such data are, unfortunately, unavailable for the immediate postcolonial years.
Many scholars have also articulated a potential relationship between natural resources and conflict, including coups. Natural resource rents could stabilize governments by providing fiscal revenue, independent of citizens, that can be used to increase the repressive capabilities of the state (Ross 2004)—including coup proofing. Other scholars propose an alternative mechanism that revenues from the sale of natural resources increase the potential rewards of state control, thereby encouraging attempts to seize power (Collier and Hoeffler 2005, 4-5; Humphreys 2005, 519). We thus have two mechanisms, pulling in opposite directions. For the count model, I include a dummy variable, Mineral exporter, for whether oil, gas, or other minerals constituted a principal export at independence (Mitchell 1998, 637-50; United Nations 1965, 37). For the logit models on democratization, I include Humphreys's variable for Oil production—the average daily per capita barrel production for the year of the transition.
For the logit models on electoral transitions only, three additional control variables are considered. First, closely contested elections may increase the probability of a violent military response by those officers sympathetic to the losing side. Drawn from Lindberg’s (2009) data on African elections, Margin of victory is measured by subtracting the vote share of the second place candidate or party from that of the winner. Second, continuity in leadership may exist despite a change in the chief executive. If the new leader held a significant post in the old regime, or if the same political party maintains power, then military officers may not fear restructuring despite the transition. Following Lindberg again, Regime continuity is coded 1 if the pre- and post-transition leaders belong to the same political party or if the post-transition leader held an integral post in the prior administration, such as the vice-president, the foreign minister, or the secretary of state. Finally, past success with peaceful, electoral transfers of power could encourage the military to refrain from intervention. Prior successful transfer is coded 1 if at any point in the past—no matter how long ago—executive power was transferred constitutionally/electorally without a subsequent successful coup attempt and 0 otherwise.
Results
The regression analysis confirms the findings of the descriptive statistics. Countries whose leaders chose to build ethnic armies despite facing an existing, diverse officer corps tended to face substantially more coups than their counterparts. In the count model, the Ethnic and unmatched category is positive and statistically significant (see Table 3). Considering that coups are relatively rare events, the substantive effect is quite important. Attempting to construct an ethnic army despite existing diversity in the officer corps is predicted to result in four times as many coups (see Figure 2).
Count (negative binomial) Model.
Note: GDP = gross domestic product. DV = number of coup attempts in first twenty years after independence standard errors, n = 43.
***p ≤ .001, **p ≤ .01, *p ≤ .05.
Decolonization and predicted coup count.
Likewise, the multivariate results for the democratization data confirm the descriptive findings. Table 4 presents the regression results for both the full set of constitutional power transfers and the elections only data. While both the Ethnic continuity and stacking and Ethnic change and stacking categories are significantly associated with increased coup risk across models, the magnitude of the effect is comparatively small: 12 to 25 percent risk compared to the 0 percent risk of the baseline category (ethnic continuity and prior inclusive army). The Ethnic change and stacking category, on the other hand, has a drastically higher predicted coup risk: 66 percent for all constitutional power transitions and 89 percent when elections are involved, all else being equal (see Figure 3).
Determinants of Military Coups after Constitutional Transfers of Power.
Note: GDP = gross domestic product. Logit, DV = coup attempt within four years robust errors, n = 69 for all transfers and 51 for elections only.
***p ≤ .001, **p ≤ .01, *p ≤ .05.
Ethnicity, power transfers, and predicted coup probabilities.
Few of the control variables attained statistical significance. Prior success with democracy, no matter how long ago, seems to decrease coup risk. While statistically significant, however, the effect is marginal: decreasing coup risk by under 1 percentage point. Economic shocks also increase the risk of a coup when closely following an election. Here, the substantive effect is strong: a 1 percent contraction of the economy leads to a 14.1 percentage point increase in coup risk while a 3 percent contraction increases that risk by an additional 34.6 percentage points. Similarly, in the all constitutional transfers data, a 3 percent economic shock was statistically significant (even though the 1 percent contraction was not), resulting in an increase of 40.8 percentage points in coup risk. No other variables were consistently statistically significant, including prior Ethnic violence, Ln GDPk, Prior coups, Ethnic Diversity, and Change in civil liberties.
Robustness checks
Due to substantial variation in the types of constitutional changes in leadership, unit heterogeneity may undermine the comparability of observations. A separate analysis was thus conducted on the most homogeneous subset of electoral transitions—those regularly scheduled or occurring through planned democratization efforts (n = 35). Both the Ethnic change and stacking category and the Economic shock variable retained their statistical significance and direction and showed a slight increase in magnitude (for full results, see the Online Appendix).
Further robustness checks were conducted by substituting alternative measures for several of the control variables as well as by introducing further controls for other potentially important factors. The Online Appendix contains descriptions of each of these additional variables as well as the full statistical results. Across all models, the competing ethnic fractionalization measures of Alesina et al. (2003), Fearon (2003a), and Posner (2004) were substituted for ELF. Two measures for ethnic dominance were also constructed from Fearon’s data: percentage share of the largest group in the total population and whether any group constituted more than 50 percent of the population. Indicator variables were included for ongoing civil wars, which could place independent strains on civil–military relations, as well as for Herbst’s (2000) measure of difficult geography.
For the democratization models, oil reserves and diamond production data were substituted for oil production. Alternative measures for prior coups were also constructed: the count of successful coups in the ten years prior to the power transfer and the number of years since the last attempt. Polity IV scores, Freedom House civil liberty scores, and regime type (following Lindberg and Clark 2008) were also added as substitute measures for legitimacy. Additionally, Powell (2012) finds that characteristics of military organizations have an important influence on general coup risk and Powell and Lasley (2012) argue that the anti-coup censure and sanctions imposed by the African Union (AU) have had a deterrent effect on coup attempts. Thus, their measures for military expenditures per soldier, counterbalancing paramilitary forces, the log of military personnel, and AU membership are also included.
None of these robustness checks meaningfully changes the results for the key independent variables. During the decolonization period, choosing ethnic loyalty despite an unmatched officer corps remains statistically significant, in the same direction, and of the same magnitude. None of the additional control variables were significant in the decolonization models. Likewise, with the democratization data, the category variable indicating an ethnic change in leadership combined with prior ethnic stacking remained statistically significant, positive, and of the same general magnitude across robustness checks. The majority of the control variables were insignificant including the military organization variables, civil liberty score, Polity measures, AU membership, ongoing civil war, and Herbst’s (2000) difficult geography. Alternative measures for prior coups, however, were statistically significant. Indeed, leaders seemed particularly vulnerable to military overthrow when a recent coup had led to the election that brought them to power.
Conclusion
I have argued throughout this article that understanding Africa’s susceptibility to military coups requires a contextually sensitive and nuanced appreciation for ethnic politics and the manipulation of ethnicity within state security institutions. Where early African leaders attempted to build coethnic armies, despite existing ethnic diversity in their officer corps, they often launched their countries into severe coup traps. Many of these traps could only end in Pyrrhic victory, as one group established homogeneity through violent purges of the ranks. Such countries then joined those who, with the help of departing colonizers, had successfully built ethnic armies from the beginning. This historical legacy of stacking military institutions with coethnics has had grave consequences for contemporary democratization efforts. When elections bring to power new leaders from different ethnic groups, these entrenched ethnic armies react by seizing power.
Given the highly multiethnic character of most African societies, we should embrace the normative ideal of constitutional politics that allow for the rotation of leadership between individuals and groups of different identities. The alternative is an ethnic state that indefinitely excludes much of its population from power. Such ethnic states fail to meet important standards of justice and equity. Recent research also suggests that exclusionary states are especially prone to insurgency and the devastation of civil war (see Cederman, Wimmer, and Min 2010; Roessler 2011; and Wimmer, Cederman, and Min 2009). Multiethnic democracy, and indeed long-term peace and stability, thus require dismantling ethnic armies.
Yet, this article underscores the difficulty and danger involved in restructuring military institutions—in dissembling ethnic patronage networks and building merit-based recruitment and promotion systems in their place. Ethnic armies will not sit idly by and allow for their own demise. The international community can play a limited role in helping willing governments to reform security institutions, particularly where peacekeepers or other monitors have already been deployed. External military personnel can serve as neutral arbiters, offer training and educational assistance in the mechanics of running a merit-based army, and detect coup plots where they are physically present. Yet, outsiders may have little sway over the long-term entrenchment of merit-based security institutions. Once foreign missions depart, they can no longer ensure that civilian leaders will not revert to ethnic recruitment practices. Nor can they prevent disenchanted or fearful officers from taking violent action. Lasting military reform is an ordeal that must be grappled with domestically.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The author would like to acknowledge the Bobst Center for Peace and Justice, the Princeton Institute for International and Regional Studies, and the Kroc Institute for International Peace Studies for financial support during the research and writing of this article.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
