Was it a coup or a revolution? This question became salient in public debates in Egypt after the ousting, on July 2, 2013, of Mohamed Morsi, President of Egypt and member of the Islamist Muslim Brotherhood, by field marshal Abdel Fattah El-Sisi, then minister of defense and head of the Supreme Council of Armed Forces. Supporters of the revolution camp would typically point at the dismal performance of the Morsi administration during his yearlong term in office and popular mass mobilization in support of the government’s removal, organized by the
Tamarod (Arabic “rebel”) movement in June 2013. Proponents of the coup-narrative would invoke the army’s de facto takeover of power, repressive treatment of political opponents, and El-Sisi’s capture of the presidential office on May 27, 2014. In this context, it should not come as a surprise that “[t]he combination of mass protests with leadership turnover has created divisive debates about whether an event should be called a coup or a popular revolution” (
Tansey, 2016, p. 153).
Political disagreements about what happened in Egypt reflect a larger problem in the scholarly research program on regime change. In this body of literature, coups and revolutions are usually thought of as distinct events driven by different causes.
Roessler (2011), for example, maintains that “coups and insurgencies represent alternative antiregime technologies” (p. 303),
Brownlee (2007) identifies “coups” and “opposition victories” as different authoritarian succession scenarios (p. 602), and Svolik (
2009) categorizes “coups d’état” and “popular uprisings” as distinct phenomena causing the fall of autocrats (p. 478).
The co-occurrence of military coups with popular uprisings has thus reinvigorated a debate on the effects of military intervention. To begin with, analyses of regime trajectories in the Arab Spring have emphasized the crucial role of military behavior (
Barany, 2016;
Bellin, 2012;
Holmes & Koehler, 2018;
Koehler, 2017). Only where military organizations defected from dictators did popular mass uprisings succeed in deposing chief executives. Based on this observation, some analysts have referred to “military-opposition alliances” (
Nepstad, 2011, p. 485) or to the degree to which different militaries were open to reform demands (
Lutterbeck, 2013). The suggestion is that not only did militaries defect from their commander in chief in the context of popular mobilization but that this behavior was driven by an alignment of preferences between military elites and popular protestors.
On a more general level, this perspective is linked to a debate on the notion of “democratic coups” (
Kuehn, 2017). Hence, while scholars of regime change more broadly tend to analyze coups and revolutions as discrete events, the literature on military coups has more recently accounted for a conjunctural perspective. If it is true that military defection is close to a necessary condition for revolutionary success, then military actions that allow (or even trigger) the fall of dictators might be seen as having prodemocratic mid- to long-term consequences.
Thyne and Powell (2016), for example, argue that coups against dictators could be considered a “window of opportunity for the promotion of democratization” (p. 194).
Collier (2009) has forcefully argued in support of this perspective, suggesting that “coups and the threat of coups can be a significant weapon in fostering democracy.” Some analysts have explicitly linked the democratic coup hypothesis to the removal of Egyptian presidents Mubarak in 2011 (
Varol, 2012) and Morsi in 2013.
Yet, the empirical evidence regarding the effects of coups on the level of political development is mixed at best. Some scholars find that coups increase the likelihood of democratization by removing recalcitrant dictators from power (
Collier, 2009;
Thyne & Powell, 2016) or that coups are less likely to result in the establishment of stable authoritarian regimes than they were during the Cold War (
Marinov & Goemans, 2014). Other scholars have demonstrated that coups more frequently lead to the establishment of authoritarian regimes than democracies and that postcoup regimes tend to be characterized by increased repression (
Derpanopoulos, Barbara Geddes, & Wright, 2016).
What explains these inconclusive findings? We argue that differences between distinct coup types have not been taken into account. Scholars have emphasized differences between the effects of coups before and after the Cold War (
Derpanopoulos et al., 2016;
Marinov & Goemans, 2014), between coups against autocratic or democratic rulers (
Derpanopoulos et al., 2016), as well as between successful and failed coups (
Thyne & Powell, 2016). Yet, they have not systematically accounted for the difference between coups in the context of regime crises, on the one hand, and other types of military intervention on the other hand. Systematic studies of democratic coups are typically based on large-
n data on episodes of contentious politics (
Derpanopoulos et al., 2016;
Marinov & Goemans, 2014;
Thyne & Powell, 2016), thereby ignoring revolutionary mass uprisings as a specific context factor.
This difference is crucial. Coups occurring in the context of antiauthoritarian mass uprisings—what we refer to as coups in authoritarian “endgame scenarios” (
Pion-Berlin & Trinkunas, 2010)—are systematically different from “regular” coups occurring in the absence of regime crises. Drawing on our original medium-
n data set of endgame coups, we suggest disaggregating specific coup types by distinguishing between different context factors (regular coups vs. endgame coups) and coup agents (senior vs. junior officers). In so doing, this article contributes to a growing research program differentiating between discrete types of coups, for instance, by emphasizing distinct agents within the military apparatus (
Albrecht & Eibl, 2018;
Bou Nassif, 2015;
De Bruin, 2019;
Singh, 2014) or the social and ethnic identities of coup plotters (
Harkness, 2018;
Kandeh, 2004;
Roessler, 2011).
In our data, we find strong evidence for two interrelated contentions. First, endgame coups occur against the backdrop of relatively more consolidated political–military relations compared to regular coups. This is because popular mass uprisings happen in authoritarian regimes where incumbents have been in office for a long period of time, which has allowed them to apply coup-proofing measures. Second, only senior officers who are part of the military’s leadership have the incentives and the capacities to overcome such adverse opportunity structures for coup plots. Endgame coups therefore feature a specific type of military intervention where military elites take over power to safeguard their political positions. Often representing the conservative interests of military elites, they are likely to be rollback coups. This has significant consequences for the prospects of postcoup democratization. The positive effect of coups on postcoup democratization noted by some scholars is more likely to be observed following regular coups orchestrated against both the political elite and parts of the military leadership.
Endgame Coups: A Distinct Type of Military Take Over
Are coups triggered by revolutionary mass uprisings conceptually different from “regular” military power grabs? And how should endgame coups be theorized in the context of general debates on military intervention in civilian politics?
Endgame coups are indeed different in two theoretically intriguing ways. First, they occur against the backdrop of consolidated authoritarian political–military relations. Popular mass uprisings disproportionately challenge long-standing dictatorships. This has palpable consequences for the relations between incumbents and officers as well as coup risk: surviving an initial consolidation period creates “common knowledge” (
Little, 2017, p. 206) among individual officers about the regime’s strength, which impedes officer coordination for coup attempts. Military intervention is the most frequent form of leadership exit in authoritarian regimes. Based on
Svolik’s (2012) leadership change data, 82.5 percentage of cases in which an authoritarian incumbent lost power to elite competitors involved the military. Given this fact, the above-average time in office exhibited by incumbents who become targets of endgame coups must be seen at least partially as an effect of successful coup-proofing. Since successful coups require coordination within the military (
Geddes, 1999;
Singh, 2014), we should expect the coordination costs of endgame coups to be significantly higher than those of regular coups given the context of consolidated political–military relations.
Our second theoretical proposition concerns coup agency. We contend that endgame coups are primarily staged by senior officers and members of the military leadership. This is because only officers in the upper echelons of the military apparatus have the capacities to overcome obstacles for coup plots emanating from the consolidated political–military relations in endgame scenarios. Second, senior officers not only have the capacities to make a move, they also have greater incentives for an intervention in politics than junior officers. Since popular mass uprisings often threaten the regime itself—and hence elite officers’ stake in power and material resources—military coups during such uprisings can be characterized as rollback coups, that is, conservative interventions in politics by military elite s preventing popular mass uprisings to result in democratic transitions.
Capacities
Popular mass uprisings create insecurity and shorten time horizons in the decision-making process of any agent in the state apparatus. Confronted with a popular mass uprising, military leaders must decide within mere days, sometimes even hours, whether to continue to support the chief executive or to turn against their commander in chief (
Gallopin, 2019;
Koehler, 2017;
Singh, 2014). This shortened time horizon further increases coordination costs, which privileges senior officers over their more junior counterparts.
Singh (2014) found that, as potential coup plotters, senior officers coordinate among themselves
horizontally, that is, across larger military units. This is to accumulate what he calls “soft power,” including the ability to gather information about internal dynamics within the military organization but also about other elite factions of the regime and the interests of foreign powers. Coordination across senior military leaders is particularly useful to overcome an effective coup-proofing measure applied by authoritarian incumbents: institutional counterbalancing (
Albrecht & Eibl, 2018).
What does this mean for our situation of endgame scenarios? In the high-uncertainty and low-information environment of such revolutionary moments, it is the soft power of senior officers and their ability to coordinate within the military’s leadership that can overcome obstacles for plotting a coup rather than the junior officers’ “hard power,” that is, their access to weapons and direct command over ordinary soldiers. This is for various reasons. First, military leaders have an advantage over combat officers in lower ranks of the military hierarchy in that they have superior information about the extent and regional dynamics of the popular challenge, possible support for a takeover from civilian elite members, and the position of foreign powers with an interest in the country’s political trajectory. Most importantly, elite officers also have superior knowledge about the level of military cohesion itself and the loyalty norm among soldiers and low-ranking officers across the organization—knowledge of critical importance during revolutionary mass uprisings where the military might have to take sides in a conflict between the people and the political incumbency (
Albrecht & Ohl, 2016;
Bou Nassif, 2015;
Gallopin, 2019).
Second, while combat officers rely on the element of surprise in their coup attempts (
Singh, 2014), elite officers need to increase communication among each other—a condition met during revolutionary episodes where security personnel often meet on a continuous basis in the crisis room. In an ironic twist, revolutionary mass uprisings therefore come to the advantage of elite-officer coup plotters in that they help them overcome coordination obstacles.
Albrecht and Bishara (2011) have shown that the 2011 ousting of Hosni Mubarak was ultimately made possible by continuing conversations and a shifting loyalty norm among officers in the Supreme Council of Armed Forces. Both the 2011 and 2013 coup episodes in Egypt also show that, once the military leadership has made a decision to move against the respective incumbent, a coup does not necessarily occur in secrecy but is rather made a fact by public announcement on national TV. Finally, officers in the upper echelons of the military hierarchy have direct access to the incumbent in power. They can signal to incumbents and civilian elites that staying in power is not an option owing to the breadth of opposition manifested in the uprising but also the lack of willingness to intervene on behalf of the embattled political leadership. While falling short of a military coup, the ousting of Tunisia’s Ben Ali in January 2011 was, at least in part, made possible by such signals from members of the Tunisian security establishment (
Gallopin, 2019).
Incentives
Revolutions include the call for regime change and therefore pose a challenge for incumbents and elite officers alike. The latter are members of the elite coalition and/or benefit exorbitantly from the authoritarian regime’s resource allocation (
Bueno de Mesquita & Smith, 2017;
Svolik, 2013). Regime change would either entail the establishment of democracy—with the military being ultimately sidelined from politics—or the resurrection of authoritarianism, most likely witnessing an alternative institutional framework and the political elite substantially reshuffled. Hence, elite officers stand to lose their personal positions or risk the complete sidelining of the military as an institution. While they would not necessarily hold on to fight for an individual political incumbent in the face of the popular insurrection, they certainly retain strong incentives to preserve the present authoritarian regime, its institutional infrastructure, and personal access to political influence and resources.
Since senior officers are often members of the political elite (
Acemoglu & Robinson, 2006), if they favor change, they most likely develop a preference for elite change—engineered through their very takeover of power—over regime change, the latter typically advocated by agents of revolutionary mass uprisings. Junior officers, by contrast, do not have a stake in political power and perceive military service primarily as professional employment. Since they are the direct superiors of rank-and-file soldiers, their perceptions and interests in the context of popular mass uprisings are also shaped by their subordinates, that is, military personnel who more likely associate themselves with the revolutionary spirit of ordinary people rather than those in the corridors of power (
Albrecht & Eibl, 2018). While junior officers and rank-and-file soldiers lack interest in coup attempts, they also do not want to be engaged in the suppression of popular uprisings, at least when they remain peaceful. Such threats of insubordination from the lower ranks, in turn, constrain the options of the military leadership (
Dragu & Lupu, 2018). Where a shooting order threatens cohesion in the military apparatus, removing the incumbent may quite well look like an easy way out of the conflict (
Lutscher, 2016).
The Empirical Universe of Popular Mass Uprisings
Our data set comprises the complete universe of endgame scenarios in nondemocratic regimes since 1945. Drawing on the work of
Pion-Berlin and Trinkunas, we understand an endgame scenario as a situation of inclusive, mass-based popular mobilization in which “a government has exhausted most of its political capital or will to find a peaceful resolution to a conflict” (Pion-Berlin & Trinkunas, 2010, p. 398; also see
Dragu & Lupu, 2018). While Pion-Berlin and Trinkunas are interested in endgame scenarios across different regime types, including consolidated democracy, we maintain that endgame scenarios in nondemocratic regimes are likely to be characterized by distinct causal dynamics.
1 Limiting ourselves to nondemocratic regimes and to the revolutionary situation itself (rather than the nature of a possible solution), we thus define our universe of endgame cases as all those situations in which
sustained, cross-sectoral episodes of popular mass mobilization create a crisis that exceeds a nondemocratic regime’s regular capacities of social control.
Operationally, we use four distinct criteria to identify endgame scenarios. These four criteria taken together discard instances of conventional contentious politics and make sure that we only capture protest episodes that threaten regime survival, irrespective of whether or not the regime survives the challenge. Four elements allow us to distinguish revolutionary mass mobilization from more mundane and regular forms of popular dissent, namely the power in numbers, social inclusion, and the sustenance of contentious mass action required to pose a significant threat to authoritarian power maintenance: (1) Protests are large in terms of
numbers, exceeding 50,000 participants, or 3% of the population; (2) protests are
sustained, lasting at least 1 week and possibly several months, unless they are cut short by the resignation of the state’s chief executive or violent repression; (3) protests are socially inclusive and
cross-sectoral, representing more than one particular social group (tribe, clan, religious community, class, interest group); (4) the
target of the protests is the regime incumbent in a sovereign, nondemocratic state.
2 We relied on existing data covering different aspects of protest events and domestic instability to compile a candidate list of potential endgames.
3 We then examined each candidate case individually and coded endgames on the basis of the four criteria outlined above. Through this procedure, we identified a total of 77 endgame scenarios in nondemocratic regimes between 1945 and 2014. Military coups d’états, in turn, are
rapid, irregular takeovers of power through military personnel.
4
In our universe of endgame scenarios, we observe a total of 22 coup attempts, all of which were successful (
Table 1). Two observations raise immediate attention: First, military intervention in endgame scenarios practically guarantees the removal of nondemocratic incumbents, which supports expectations that coup success increases with increasing size of protests (
Casper & Tyson, 2014).
Second, coups occur only in less than one third of endgame cases, indicating that substantial disincentives persist for officers to intervene in politics despite obvious popular displeasure with the incumbent, the latter’s vulnerability, and prospects for change in political leadership. Our observations thus do not support the assumption that endgame scenarios themselves are sufficient opportunities for power-hungry officers. While incumbents appear weak in the face of massive popular contestation, the execution of coups requires careful planning, organizational capacities, and communication networks among potential coup plotters. As Geddes (
1999) recalls, “coups are often preceded by extensive consultation among officers, delays until almost total consensus within the officer corps is achieved, and elaborate negotiations over power sharing and rotation in office” (p. 128).
5 Rather than stage an ill-prepared coup attempt, military officers often opt to wait out an uncertain revolutionary dynamic.
Empirical Evidence: Rollback Coups Amid Consolidated Political–Military Relations
In this section, we proceed to explain why coups do occur during endgame scenarios and present empirical evidence in support of our theoretical expectations. Comparing endgame coups with regular instances of military coups reveal that the former typically occur against the backdrop of consolidated political–military relations. Second, if coups do happen amid popular mass uprisings, it is primarily the military’s leadership making a move to oust incumbents in order to preserve their own position in the embattled regime.
Endgame Coups Versus Regular Coups
To what extent do popular mass uprisings serve as a context factor for military intervention in politics? If endgame coups are indeed distinct from regular coups along the lines suggested above, we expect to find systematic differences between these two types of military interventions along theoretically significant dimensions. We first draw on the substantial literature on coups d’état to outline three categories of theoretically relevant variables to show that endgame coups are indeed different from regular coups in ways that are consistent with our theoretical argument. Existing accounts for the causal determinants of military intervention in civilian politics can usefully be broken down into three categories: grievances, opportunities, and organizational features.
Personal or corporate grievances are among the most frequently cited factors said to motivate military officers to enter the fray of domestic politics. Indeed, Nordlinger (
1977) maintains that “the great majority of coups are partly, primarily, or entirely motivated by the defense or enhancement of the military’s corporate interests” (p. 78). Large-
n studies have usually focused on factors such as (changes in) military size or spending as operational measures (
Leon, 2014;
Powell, 2012). Most fundamentally, low levels of economic development and low or negative economic growth are “close to being a necessary condition for coups” (
Londregan & Poole, 1990, p. 151).
Grievances alone, however, will not lead to military intervention. Irrespective of the strength of grievances, a coup will only occur if an opportunity presents itself (
Finer, 1962). Numerous factors are thought to constitute such pull factors. This includes low legitimacy of the target regime (
Lindberg & Clark, 2008), domestic political crises and civil war (
Bell & Koga Sudduth, 2017), and institutional weakness or decay (
Huntington, 1968). Most coup incident models feature gross domestic product (GDP) per capita and GDP growth as independent or control variables to account for the effects of economic development (
Londregan & Poole, 1990), and the use of different regime type measures, such as the Polity IV Scale, is a common choice in operationalizing the opportunity structure of military intervention (
Belkin & Schofer, 2003;
Powell, 2012). It has also been argued that military regimes are particularly vulnerable to military coups (
Geddes, 1999).
Finally, determinants of military intervention have been sought in characteristics of military organizations themselves. Military size has been discussed as a variable influencing coordination costs (
Powell, 2012), and the degree of organizational counterbalancing figures prominently in debates on the effects of coup-proofing (
Boehmelt & Pilster, 2015;
De Bruin, 2018). Moreover, experiences with intervention in the past increase the likelihood that coups will recur. Military intervention in civilian politics might contribute to the establishment of an interventionist ethos within the armed forces, leaving countries in “coup traps” (
Lehoucq & Perez-Linan, 2014;
Londregan & Poole, 1990). Empirically such factors have been operationalized by relying on some measure of past coup incidents (
Belkin & Schofer, 2003;
Powell, 2012). More recently, military behavior in the Arab Spring has given rise to an influential interpretation focusing on the degree of patrimonialism in the armed forces. The argument here is that communal militaries would refrain from replacing incumbents when facing threats from among ethnically divided societies (
Bellin, 2012;
Lutterbeck, 2013;
Roessler, 2011).
Table 2 summarizes our comparison between endgame coups and regular coups along the dimensions discussed above. Concerning corporate military grievances, regular and endgame coups do not exhibit major differences. Military expenditure as a percentage of GDP in the years preceding both regular and endgame coups is relatively low but not significantly different between different coup types. When looking at military funding in terms of expenditure per soldier, the relationship turns around with endgame coup armies appearing to be slightly (and not significantly) better funded. GDP per capita is also higher in endgame coup years (although the difference is again insignificant). In terms of corporate grievances, our data thus do not suggest major differences between regular and endgame coups.
The situation looks quite different when we turn to opportunities and organizational features. In both categories, endgame coups differ significantly from regular coups. What is particularly intriguing here is that all of these differences point in a direction that would suggest substantially higher risks for coup plotters in endgame scenarios compared to normal times—corroborating or theoretical claim that endgame coups occur in settings of consolidated political–military relations. One of the most striking observations about endgame coups is that they are typically directed against incumbents with above-average tenure in office. While regular coups target incumbents who have on average spent about 5 years in office, those chief executives targeted by endgame coups show almost twice that value. If we understand time in office as a measure of authoritarian regime consolidation (
Bienen & van de Walle, 1991;
Gandhi, 2008;
Goemans, 2008), endgame coups occur in regimes where elite coalitions are significantly more consolidated. Endgame coups therefore show a dilemma in protracted authoritarian power maintenance: While regime consolidation keeps incumbents in power for an extended period of time, they also appear to wear out their welcome, which increases the risk of revolutionary uprisings and hence endgame coups (
Albrecht, 2015).
Another set of intriguing differences relates to organizational factors. Powell (
2012) has shown that troop size has a negative effect on both the likelihood of a coup attempt and its chances of success, arguing that military size presents major challenges to organizational cohesion (p. 1032). As can be seen in
Table 2, however, militaries involved in endgame coups are larger in terms of troop size, both absolutely and relative to population size. Again, this suggests that—all other factors being equal—the risks associated with staging a coup should be higher in cases of endgame coups than in regular coups. A last organizational factor pertains to the contrast between volunteer and conscript forces, which can be used to account for a military’s communal organization.
Cohn and Toronto (2017) have pointed at “widespread agreement that volunteer forces will tend to be more socially isolated and possibly less representative than conscript forces” (p. 437). All other things being equal, general conscription is likely to give rise to military organizations whose rank and file are broadly representative of social heterogeneity, while volunteer forces facilitate ethnic stacking (
Bellin, 2012;
Harkness, 2018). The observed overrepresentation of conscription forces among militaries intervening in endgame situations implies that endgame coups are more likely in institutionalized armies than in patrimonial militaries.
What do these differences mean in substantive terms? To begin with, there is strong evidence for treating endgame coups as a distinct category. While military leaders involved in endgame coups do not seem to have stronger corporate grievances than coup plotters in normal times, they face comparatively more adverse opportunity structures. This might seem, at first sight, a counterintuitive finding given that endgames themselves are situations typically created by grievance-driven popular contention. However, agents of coup plots are military officers, and the measures used to explore their grievances typically treat them as agents in an institution with a specific organizational ethos producing interests and grievances different from those of ordinary people.
Beyond the obvious point that endgame coups occur against the backdrop of strong popular mobilization, they also happen in situations of adverse opportunities and involve militaries with specific organizational features. These patterns corroborate our specific theoretical angle. Endgames typically occur in regimes featuring consolidated elite coalitions, including a relatively stable political–military balance. This suggests two things. First, under the conditions of consolidated political–military relations, senior officers are more likely to achieve the coordination necessary to stage a successful coup—all the more so if time is a major constraint as in revolutionary situations. Second, while there are few incentives for senior officers to turn against the incumbent on their own accord, if they do so in the context of an endgame scenario this intervention is likely to be restorative, rather than revolutionary, in nature. The next section explores these implications in more detail by examining the empirical effects of endgame coups compared to regular military interventions.
Agency in Endgame Coups
There is strong empirical evidence in support of our second core assumption that endgame coups represent conservative interventions of military leaderships. Among our 22 endgame coups, 18 have been executed by elite officers (81.8%).
6 Only four military interventions during revolutions came from low-ranking officers, which we conceptionalize as combat officer coups.
7 This observation also reveals a significant difference compared with regular coups, that is, those military interventions in politics that did not go hand in hand with popular uprisings.
According to our global coup data, a majority of coups have been attempted by officers outside of the military’s leadership (see
Table 3), which highlights our emphasis of endgame coups as discrete events where military leaderships step in to preserve the political status quo. Of a total of 475 coup attempts reported between 1950 and 2018, 58% witnessed junior coup plotters—an observation in striking contrast to the vast majority of endgame coups attempted by elite officers. This is in line with our expectation that only the military’s leadership possesses the means to organize for a takeover of power in the context of endgame scenarios. While the success rate of the aggregated coup attempts stands at 50%, substantial differences exist when looking at coup agency, with elite officers significantly more successful than junior plotters. Recent scholarship found that elite officers have superior capacities to coordinate for coup plots compared to their junior colleagues (
Albrecht & Eibl, 2018, p. 319;
De Bruin, 2019, p. 5;
Singh, 2014)—a factor of particular importance amid uncertain political dynamics of revolutionary mass uprisings.
Since they are executed primarily by senior officers from the military’s leadership, endgame coups are textbook examples of power struggles within the political elite (
Acemoglu & Robinson, 2006;
Bueno de Mesquita & Smith, 2017;
Svolik, 2013). They are thus best understood as rollback coups, exercised by existing political elite members and based on their control over the coercive institution of the state. In the face of popular pressure for revolutionary change, the ultimate aim of the coup plotters is to safeguard their own position in politics as well as the authoritarian nature of politics writ large.
Some examples illustrate this interpretation. Apart from the 2013 episode mentioned in this article’s introduction, Egypt witnessed an earlier intervention, in 2011, by the Supreme Council of Armed Forces under the leadership of minister of defense Mohamed Hussein Tantawi and chief of staff Sami Annan. The ultimate aim of the coup, apart from the sidelining of longtime President Hosni Mubarak, was the preservation of the military’s control over the Egyptian state and the subsequent transition process (
Albrecht & Bishara, 2011). Algeria witnessed its own rollback coup in 1992. The military leadership not only replaced then-president Chadli Bendjedid but also ended a period of limited political opening. An Islamist movement, the Islamic Salvation Front, capitalized on political liberalization, succeeded in elections, and hence mounted a substantial challenge to the Benjedid-regime (
Cook 2007, pp. 32–62). In Myanmar, “(t)he military (…) managed to reconsolidate its power after it cracked down on prodemocracy demonstrations in August 1988, killing several thousand protestors” (
Bünte, 2014; p. 744).
Burkina Faso is one of the few cases where combat officers plotted for a coup amid a popular uprising. In 2014, Isaac Zida, the deputy commander of the presidential guard, sidelined the country’s military leadership around General Honore Traoré and took power from longtime dictator Blaise Campaoré. While this might first appear as a challenging case for our theoretical expectations, the political dynamics in the country corroborate our premise about the main condition for endgame coups: preparation and coordination prior to the event. In fact, the endgame coup by combat officers did not occur in a vacuum but rather was preceded by a whole string of mutinies among soldiers and low-ranking officers just a few years prior to the 2014 uprising. Popular unrest in 2011 “included a series of often violent mutinies by uniformed soldiers and police. Those involved were generally very young, from the lower ranks. They had no known organization apart from informal networks of age-mates, and communicated largely by word of mouth or cell phone” (
Harsch, 2016, p. 234). Yet, the 2011 mutinies served as a test run for would-be junior coup plotters. As Dwyer (
2017) revealed, the most severe threat in 2011 emerged from a revolt of the presidential guard (p. 225)—the very unit from which the 2014-coup leader Zida would emerge.
Regime Trajectories After Endgame Coups
Our data present strong support for our theoretical assumptions. The evidence discussed so far suggests that endgame coups differ from regular coups mainly in that they occur under circumstances of consolidated elite coalitions. Since we compare endgame coups to military coups that occur without the presence of prior popular mass mobilization, we cannot draw inferences about the causal determinants of endgame coups as yet. Methodologically speaking, we are confronted with a sample selection problem since we only observe endgame coups in cases where an endgame actually occurred. The variation between regular and endgame coups is indeed to a large extent endogenous to revolutionary mass uprisings. In other words, the increased likelihood of military intervention in consolidated regimes during endgame scenarios largely reflects the fact that such regimes are more likely to experience mass uprisings.
We proceed toward further systematic tests by measuring observable implications about the main theoretical puzzle established in the previous sections. While most existing accounts of endgame coups highlight the revolutionary nature of the event and its prospects for democratization, our findings so far suggest a very different interpretation: Endgame coups are conservative rollback coups of power-driven elite officers trying to prevent regime change. Our data on endgame coups allow us to systematically observe the effects of coups on the level of regime trajectories. If coup plotters would associate themselves with the popular demands of revolutionaries, we would expect to witness a postcoup period of reforms, social development, and democratic transition. If, however, endgame coups are indeed conservative rollback coups meant to preserve the existing political order, we would expect them to install a new authoritarian regime or to trigger a protracted phase of political instability.
This is because the outcome of the revolutionary situation (conservative rollback coups) stands in stark contrast to the objectives of the people accepting significant personal risks for demanding substantial political, social, and economic change. We expect successful coup plotters to face a politicized public that—perhaps after a brief honeymoon period between the revolutionaries and the officers—will come to understand that the military’s intervention was not intended to implement the goals of the revolutionary movement but rather to preserve the political status quo. The ongoing struggle between civilian activists and conservative military officers following the April 2019 coup in Sudan illustrates this dynamic (
Lynch, 2019).
We would therefore expect such postcoup trajectories to be characterized by protracted periods of political conflict pitting disappointed revolutionaries against the coup plotters. While military regimes are generally less stable than other authoritarian regime types (
Geddes, 1999), we expect a revolutionary trajectory in endgame coups to result in even greater degrees of political instability than developments after coups that come about without prior mass mobilization. Our first two hypotheses therefore read as follows:
Hypothesis 1a: Endgame coups are less likely to result in stable political orders than other types of leader exit.
Hypothesis 1b: Endgame coups are less likely to result in stable political orders than regular coups.
Second, if our interpretation of conservative rollback coups is correct, we should expect that they do not result in democratic transitions. Our understanding of endgame coups therefore is in direct opposition to the prominent perception of such military interventions as “democratic coups” (
Degaut, 2017;
Marinov & Goemans, 2014;
Thyne & Powell, 2016).
Hypothesis 2a: Endgame coups are less likely to result in postexit liberalization than other types of transitions.
Hypothesis 2b: Endgame coups are less likely to result in postexit liberalization than regular coups.
Empirical Analysis: Consequences of Endgame Coups
We test these hypotheses in two different ways. First, we specify proportional hazard models to estimate the risk that a postcoup regime will end in political instability or political liberalization. We draw on data on all cases of irregular leader exit in authoritarian regimes between 1945 and 2015 as recorded in the Archigos data (
Goemans, Gleditsch, & Chiozza, 2009). There are a total of 257 such irregular exits, though the actual number of cases is smaller in most of our analytical applications due to missing data. Second, we test the robustness of our findings by reproducing the models specified by
Clayton Thyne and Jonathan Powell (2016), which estimate the effect of coups on democratization. These models are based on all authoritarian regime years between 1952 and 2008 and model the likelihood that a regime will democratize within the 3 years following a coup event.
Postexit regime trajectories
We use two different dependent variables in our first test of the hypotheses outlined above. The first dependent variable captures the notion of instability after irregular leader exit. It is coded 1 if, within a spell initiated by an irregular leader exit, significant political instability is observed. We code this variable based on the SPEED Civil Unrest Data (
Hayes & Nardulli, 2011). If an exit spell witnesses significant political instability operationalized as a value in the SPEED data of at least 1 standard deviation above the mean, the instability variable is coded one and the spell exits the observation set. Otherwise, the spell continues until there is another irregular exit (in which case a separate spell is initiated) or the observation period ends (right censoring). Our second dependent variable, in turn, captures postexit deliberalization. It is coded 1 if within a spell initiated by irregular exit, the Polity IV score decreases by at least 3 points compared to the year in which the exit occurred (
Marshall, Gurr, & Jaggers, 2017). Again exit spells experiencing deliberalization exit the observation set. The exit spells are sliced in years, which allows us to accommodate time-varying covariates.
As a first step, we compare the frequency of our two outcomes (instability and deliberalization) across different exit types.
Table 4 contains information on the overall distribution of postexit trajectories in our universe of cases, differentiated by different exit types. The rows labeled change display the change in the frequency of instability or deliberalization if exits by endgame coup are compared to regular coups.
This simple tabulation of frequencies is suggestive: exit by endgame coups ends in instability in 24% of the cases, while the proportion for regular coups and other types of exits is 12% and 20%, respectively. In other words, endgame coups are more likely to lead to large-scale political instability than both, regular coups and other types of irregular leadership exit. Second, 41% of all cases of exit by endgame coups result in significant deliberalization, compared to 34% of regular coups and 48% of other exit types. This suggests that for the outcome of deliberalization, the effects of endgame coups might significantly differ from regular coups but not from other types of irregular exit.
In order to examine the robustness of these empirical patterns, we use the same dependent variables described above in a multivariate setting. We base our analysis on spells initiated by an irregular leadership exit and terminated by instability or deliberalization, or by another irregular exit (initiating a separate spell). Some spells are right censored in that they do not end in instability or deliberalization or are terminated by another irregular exit before the end of our observation period (2015). Given the structure of our data, we specify Cox proportional hazard models to estimate the risk of failure by instability (or deliberalization) following irregular leadership exit.
The main variables of theoretical interest capture different forms of irregular leader exit, namely exit by endgame coup, exit by regular coup, or exit through other forms. The construction of the endgame coup variable has been described above, while the coup variable captures those successful coups recorded by
Powell and Thyne (2011). All other forms of irregular leader exit (assassination, revolution, etc.) are captured as a residual category. In Models 1a and 2a, we compare leader exits triggered by endgame coups to all other irregular exits to test Hypotheses 1a and 2a. Models 1b and 2b, in turn, include the exit by endgame coup and other exit variables while the exit by coup variable is the excluded category. This allows us to directly compare the effect of exit by endgame coup to that of exit by regular coup and thus to test Hypotheses 1b and 2b.
We include a set of different control variables that might independently influence stability and liberalization, including regime type. The absence of institutionalized procedures of political competition that is characteristic of personalist regimes, for example, might trigger more intense political conflict after leadership exit compared to regimes that allowed for some political institutionalization. Similarly, military influence on a political regime has been shown to increase the likelihood of repression (
Davenport, 1995). We control for these factors by including variables for the most frequently used distinction between authoritarian regimes, namely that between party-based, personalist, and military regimes (
Geddes, 1999). The data are from
Geddes, Wright, and Frantz (2014), and party regimes are the excluded category. Potential effects of the included regime type must therefore be evaluated against the baseline probabilities for party regimes.
Second, violent forms of change are less likely to lead to stability and democracy (
Geddes, Wright, & Frantz 2014;
Teorell, 2010). We capture these dimensions by including the level of instability in a version lagged by 1 year (based on the SPEED data). Moreover, levels of economic development determine the likelihood of democratization (
Przeworski & Limongi, 1997).
Miller (2012) most forcefully held that economic development leads to democratization when political regimes experience a violent turn-over of power. To capture this possibility, we include a measure of GDP per capita as provided by
K. S. Gleditsch (2002). We lag all control variables by 1 year to make sure that we do not capture the effects of regime breakdown rather than its potential causes.
8
We specify Cox proportional hazard models to estimate the risk of failure by instability or deliberalization. Using survival models is an appropriate approach for two reasons. Conceptually, these models are appropriate since they explicitly model the risk of an event (instability or deliberalization) dependent on the passage of time. Moreover, slicing the data into exit spells allows us to also capture potential midterm effects by taking into account the entire duration of our spells, while alternative approaches would require the imposition of some (arbitrary) threshold. To make sure that our results are not driven by modeling choices, we also estimate logit models with controls for time dependence following the approach in
Thyne and Powell (2016).
Our models are based on a total number of 257 exit spells, 39 of which end in significant instability and 102 of which witness significant deliberalization.
Figure 1 is a coefficient plot of the model displayed in
Table 5. As can be seen from
Figure 1 and
Table 5, both Hypotheses 1a and 1b are confirmed: Exit spells initiated by endgame coups are less likely to lead to stability than spells initiated by other types of irregular exits (confirming H1a) or regular coups (confirming H1b). In effect, the risk of instability is 5.4 times higher for exit spells initiated by endgame coups compared to all other irregular exit spells and 6 times higher compared to regular coups only. The situation is somewhat different for the case of postexit deliberalization. There is no statistically significant effect of exit by endgame coup on the risk of deliberalization compared to all other irregular exits. When compared to exit spells initiated by regular coups, the risk of postexit deliberalization increases by a factor of 2.5. We thus reject Hypothesis 2a. Hypothesis 2b, by contrast, is corroborated by our analysis.
Postcoup democratization
To make sure that our findings regarding the effects of endgame coups on postcoup regime trajectories are not a mere artifact of our estimation strategy, we test the robustness of our results by adopting the alternative strategy employed in
Thyne’s and Powell’s (2016) analysis of democratic coups.
9 Thyne and Powell evaluate the effects of coups on democratization through a dependent variable coded one if a country receives a score of six or higher on the Polity Scale. Their universe of cases is delimited by all authoritarian country-years, operationalized as countries with a Polity score of below 6. If a country receives a higher score, the dependent variable is coded 1 and the country leaves the observation set. If a country’s Polity score subsequently slips below 6, the case reenters the observation set.
The main independent variable of theoretical interest in the original models is the occurrence of a coup. This variable is coded one for each year in which a coup occurred as well as for the 2 following years in order to capture potential midterm effects.
Thyne and Powell (2016) run separate models for successful, failed, and all coups. Given that endgame coups are always successful, we only estimate models with successful and all coups here. Control variables include a measure of prior democratization, variables capturing aspects of colonial history, a dummy variable controlling for whether the coup occurred during the cold war, GDP per capita and change in GDP per capita, as well as time controls to account for temporal dependence (see
Thyne & Powell, 2016). Our variable of theoretical interests captures the occurrence of endgame coups as defined above. In contrast to the previous models, however, we follow Thyne and Powell and also code the variable 1 for the 2 years following the event.
As can be seen from
Table 6, the results of this robustness test further corroborate Hypothesis 2b. While the democratizing effect of regular coups noted by
Thyne and Powell (2016) is maintained, endgame coups exert a significant negative effect on the likelihood of postcoup democratization in both models. In other words, while regular coups might open up avenues of democratization, coups staged in the context of popular mass uprisings promote autocracy.
Conclusions
Military coups triggered by popular mass uprisings comprise a unique category of military intervention in politics in that they occur against the backdrop of consolidated political–military relations. Compared to regular coups without popular mobilization, coup plotters face long-sitting incumbents—a strong indicator for successfully applied coup-proofing measures. Such coups during authoritarian endgame scenarios are rollback coups, executed by the regimes’ military leadership in order to preserve the status quo. We find that such coups trigger more instability compared to both other irregular exit types in general and regular coups in particular. They also reduce the likelihood of postexit liberalization compared to regular coups. These findings have significant substantive, theoretical, and methodological implications.
Substantively, we find that coups following popular mass uprisings lead to significantly lower levels of stability compared to any other postexit trajectory. Rather than securing political stability—a claim often used by officers to justify their interventions—military coups actually produce further conflict. Our findings are counterintuitive to those interpretations emphasizing that revolutionaries and coup plotters would move hand in hand to replace autocrats. Rather, these findings support the assumption that the rationale for those taking to the streets is quite different from the considerations of military agents actually replacing embattled regime incumbents. Third, military involvement in leader exit is also associated with illiberal regime outcomes. Endgame coups reproduce autocracy rather than bringing about democracy.
Methodologically, our analysis lends support to Charles Tilly’s advice to “split,” rather than “lump,” our empirical subject field (
Tilly, 1984) by isolating a medium-
n data set of endgame coups. There is significant empirical and theoretical value in such an exercise in that we introduce an important context factor—revolutionary mass uprisings—in order to more accurately operationalize our universe of cases. Scholars routinely emphasize the interplay of popular mobilization and military intervention in politics yet largely neglecting this context factor in systematic accounts when they use exhaustive, large-
n coup data. By focusing on rare events, we offer a point of departure for further research on revolutions, a unique type of military intervention in politics, and trajectories after authoritarian regime breakdown.
Acknowledgments
The authors thank Sarrah Kassem and Austin Schutz for their assistance in the data collection. We are grateful for insightful comments by Zoltan Barany, Jason Brownlee, Matt Buehler, Ferdinand Eibl, David Pion-Berlin, Kurt Weyland, Krista Wiegand, and the anonymous reviewers.
Authors’ Note
Previous versions of the article have been presented at the 2014 annual meeting of the American Political Science Association, Washington, DC, August 28–31; at the University of Texas, Austin, November 6, 2017; and at the University of Tennessee, Knoxville, February 23, 2018. All errors are those of the authors.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.