Even Generals Need Friends

Domestic Responses and Regime Tenure

Defined as the “degree of consensus among citizens, elites, and organizations about the state’s right to make rules” perception of a regime’s legitimacy has been a commonly noted predictor of coups (Belkin and Schofer 2003, 607). Economic performance has been argued to proxy for a government’s legitimacy, though variables accounting for overt antipathy against a regime, such as protests and riots, seem to be more robustly associated with coup activity (Thyne 2010; Casper and Tyson 2014). Finer (1962) claimed that militaries will intervene when the public overtly displays its dissatisfaction with the regime, while Pion-Berlin and Trinkunas (2010) more recently argued that Latin American armies have chosen to remain quartered during constitutional crises due to worry about the public’s response to their actions.

Putschists are ultimately dependent on constituents accepting their claim to legitimacy, but often miscalculate what the public reaction will be. Many failed coups, including Germany’s infamous Kapp putsch in 1920 and the Soviet attempt to oust President Gorbachev in 1991, collapsed after public support anticipated by the instigators never materialized. Roberts (1975, 31) concluded that “even token civilian opposition can have a disproportionate effect” on an effort’s outcome, while Sutter (1999) equates unpopular coups with overtaking a ship: armed men can overtake the bridge, but the passengers can still knock out the engines. Indeed, civilians have the potential to make a polity completely ungovernable in a postcoup environment. These leaders would seem to have three options: (1) repress dissent in an attempt to consolidate power, (2) stay at the helm of Sutter’s disabled ship and risk being overthrown by force, or (3) voluntarily step down from the bridge.

We expect leaders to be less willing to repress dissent or make other moves to retain power that might risk their elite status altogether. Research has shown that the costs of leadership turnover are dangerous following the costs of conflict (coups do, after all, have costs of their own), and that frequent leadership turnover leads to commitment problems for the state (Wolford 2007, 2012). Far from being the end of political turmoil, the putschists’ rise to power could merely represent the beginning of a much longer political crisis. To avoid further deterioration, coup-born leaders are tasked with building a regime that is seen as legitimate by those they mean to rule. These leaders consequently justify their efforts as acting in public good, including ending repression utilized by the previous regime. Few actions will undermine the new regime’s efforts quicker than the use of repression, and repression itself has severe limitations and can undermine the interests of the military (Danopoulos 1988; Wintrobe 1998; Enterline and Gleditsch 2000; Pion-Berlin and Trinkunas 2010).

To avoid risking their elite status altogether, coup leaders faced with protests are likely to cede power or open up the political process. For example, Thyne and Powell (2016) suggest that coup-born governments will attempt to increase their popularity by permitting democratic reforms. Although such reforms could lead to one’s ouster through electoral loss, this is preferable to the consequences of being ousted through either countercoups or mass uprisings. Leaders removed through irregular means are over twice as likely to be exiled as to go unpunished and are equally likely to be killed (Goemans 2008). ¹ Maintaining power in the face of domestic resentment can thus be risky, especially when one considers the disproportionate increase in coup likelihood shortly after one has been attempted (Londregan and Poole 1990).

While opening the political process is not always a viable option, coup-born regimes do have the ability to shift hostile opinions by stepping down from office even when they are unable to handpick their successors. This might seem contrary to the incentives of attempting a coup in the first place, but such a strategy has its advantages, especially if the plotters are met with unexpected resistance. First, the plotters do not run the risk of being overthrown, and successful putschists can negotiate an exit by securing an amnesty guarantee (e.g., 2003 São Tomé e Princípe). Second, stepping down can earn rewards from the ensuing regime, including improved organizational benefits or promotions for officers. Even Captain Amadou Sanogo, whose 2012 coup in Mali can be described as nothing short of calamitous, was promoted to the rank of lieutenant general shortly after ceding power. Third, stepping down most often leaves officers eligible for a future office. The target of Captain Sanogo’s coup in Mali, for example, originally came to power following the 1991 ouster of Moussa Traoré. Newfound leader Amadou Toumani Touré quickly stepped down, was promoted to general by the following government and later ascended to the presidency after testing the 2002 presidential ballot.

Public responses to coup-born governments are important for three broad reasons. First, protests can undermine the regime’s legitimacy and make it difficult to consolidate rule. Second, the existence of a coup in the first place indicates that the regime is already weak and potentially vulnerable to a subsequent coup. Such a scenario grows increasingly tenuous with an increase in antipathy toward the leader. Third, though the public could react with hostility toward a coup, all is not lost for its participants. Voluntarily stepping aside helps avoid larger, revolution-like crises that risk the coup leaders’ elite statuses altogether and can earn the plotters amnesty or allow them to contest later elections. This discussion leads us to our first hypothesis:

Hypothesis 1: As negative postcoup reactions from domestic actors increase, the tenure in office for leaders who came to power via a coup d’état should decrease.

International Responses

International reactions to coups have similar effects as domestic responses. While coups have not previously been examined individually in the leadership turnover literature, work shows that the perceptions of the international community play a strong role in leadership durability (Bueno de Mesquita and Siverson 1995; Colaresi 2004; Wolford 2007; Goemans 2008; Debs and Goemans 2010; Shannon et al. 2015; Escribá-Folch and Wright 2010). In the light of the irregular manner of removal from office, these conditions should be amplified with coups. Such a viewpoint is consistent with Thyne’s (2010) investigation of coups in Latin America, which found attempts to be more likely when regimes receive hostile signals from the United States. Such a view is also prevalent in the literature on coups in Africa, which has argued that a lack of international condemnation of independence-era coups increased the willingness of militaries to intervene (e.g., First 1971).

Based on this literature, we suspect most coup-born regimes facing international backlash to be unwilling to risk the consequences of a foreign intervention, choosing instead to cede power before such consequences arise. Two of Latin America’s more recent coups help illustrate this expectation. Following the overthrow of Haitian President Aristide in 1991, the Organization of American States (OAS) quickly instituted a trade embargo, and sanctions went from hemispheric to global after the approval of a “universal” oil and arms embargo with United Nations (UN) Security Council Resolution 841. Although the junta initially refused to step down, it was ultimately convinced to do so following the approved deployment of 20,000 US marines. Raoul Cédras left for Panama soon afterward, leaving behind his position as commander in chief of the army.

Based on the Haitian example, Honduran soldiers who overthrew President Zelaya in 2009 likely knew what to expect from international actors, particularly after the anticoup framework was established from the OAS (specifically Resolution 1080 [1991] and the Inter-American Democratic Charter, 2001). They chose the coup route anyway and were predictably suspended from the OAS. As Thompson and Lacey (2009) explain, however, this was a strategic set of coup plotters. They predicted punishment but also predicted that they could withstand international backlash for a few months before ceding power following already scheduled elections. They were right. Power returned to democratic rule five months after the coup, and neither the civilian nor military coup leaders suffered consequences for their actions.

These examples illustrate how negative international reactions to coups can shorten the tenure of coup leaders. However, we might see dramatic variation in reactions from international actors. Following the 2008 Mauritania coup, for example, almost all reactions were negative, including official responses from states like Nigeria, the United States, and South Africa, and similar condemnation from IOs such as the UN, European Union, and African Union. International actors rarely speak with such a consistent voice. Focusing on the 2011 overthrow of Mubarak in Egypt, for example, Morey and his colleagues (2012) reveal that while states like France and the United States expressed concern fairly consistently, these negative reactions were balanced by early and consistent support from Russia and China.

While the Haitian and Honduran cases suggest that coup-born leadership should be short lived when international actors condemn the coup, the Egyptian example indicates that it would be unwise to assume the international community speaks with a unified voice. We expect opposing responses among international actors to weaken the effect that negative responses have on the coup-born government. The effects of sanctioning behavior are weakened dramatically when the coup regime can find support elsewhere, and the likelihood of direct interventions decreases if potential interveners sense that their actions could precipitate a much larger international crisis. This discussion yields our second hypothesis:

Hypothesis 2: As agreement in negative postcoup reactions from international actors increase, the tenure in office for leaders who came to power via a coup d’état will decrease.

Mixed Responses

We begin by considering cases where coup-born regimes receive conflicting reactions from domestic and international actors. Two scenarios are possible: the public could support a coup that is condemned by international actors or they could rage against a coup that is supported by international actors. In both cases, we expect the consequences of the response sent from any one actor to be tempered by contrasting reactions from another.

Considering the first scenario, international actors have fewer options to punish juntas that have domestic support. Sanctions can be applied and states can be ousted from IOs, of course, though such actions may draw a backlash by strengthening support via a rally effect and by giving the junta a scapegoat for domestic troubles (Galtung 1967). At more extreme levels, military invasions to oust even unpopular leaders will likely meet resistance, and resistance should be strongest when ousting a leader who enjoys broad public support (Edelstein 2008). Just as domestic support for coup-born leaders can offset international condemnation, so too can international support offset domestic protests. This is consistent with the more general concept of “black knights,” which are external actors that support authoritarianism (von Soest 2015; Tolstrup 2015). A general strike can derail a country’s economy and pressure the leadership to step down, for example, but this could be offset if the regime receives a lifeline such as economic aid from abroad.

The growing anticoup framework of the African Union (AU) illustrates these dynamics. The organization has seen some successes, as seen with both the declining frequency of coups and in influencing coup-born regimes (Souaré 2014; Powell, Lasley, and Schiel 2016). Following the 2003 coup in São Tomé e Princípe, for example, the new regime was immediately threatened with sanctions, including an oil embargo from Angola and a military intervention from then-AU Chairman Obasanjo’s Nigerian military (Ikome 2007). With little public support, the new regime collapsed in less than two weeks (Seibert 2003). In other cases, condemnation from international actors seems to have been offset by domestic support for the new government. The coup that overthrew Mauritanian President Taya in 2005, for example, was harshly condemned by a variety of international actors. These reactions stood in stark contrast to the flood of people who took to the streets to celebrate the strongman’s ouster (British Broadcasting Corporation [BBC] 2005). Able to withstand international condemnation, the junta leaders held onto power until elections in 2007, which saw the junta-supported Sidi Ould Cheikh Abdallahi come to power (N’Diaye 2009). These cases fit with Omorogbe’s (2011) more general overview of the AU’s anticoup efforts, which suggest that the positive influences associated with public support for the removal of dictators is being undermined by hostile international reactions. This leads us to our third hypothesis:

Hypothesis 3: The influence of reactions from domestic/international actors on the tenure of coup-born regimes should weaken if they are inconsistent with responses from international/domestic actors.

States, IOs, and Anticoup Norms

To this point, we have combined states and IOs as international actors in our discussion of how international reactions influence coup-born leadership. However, a large swath of literature focusing on liberal international institutions suggests that IOs may have a fundamentally different role from states in the international community (Doyle 1986; Moravcsik 1997). Realists contend that states pursue self-interest and power, while liberals argue that IOs serve as a testament to states’ ability to cooperate for collective interest (Mearsheimer 1991; Doyle 1986; Keohane 2005). If liberals are correct, states are apt to signal differently than IOs, and the effect of these signals should yield different postcoup outcomes. In contrast, if signals from IOs have little to no impact on the domestic affairs of other states or are no different from that of states, this bolsters the realist argument that IOs largely do not matter in international relations, at least in the postcoup context.

There is reason to suspect that responses have changed over time, both in terms of those reacting, the nature of the reactions, and the influence of the reactions. Following the end of the Second World War, political nationalism and state identity were of the utmost importance. Counterattacks to nationalist tendencies included the formation of the European Union, the UN, and the North Atlantic Treaty Organization. However, none of these were able to deter the arms race between the Soviet Union and the United States as the Cold War came to the political forefront. Up until the end of the Cold War, we expect to see state reactions dominate signals from IOs, if for no other reason than that nationalist ideology had benefited from a much longer period of path dependence than cooperative IOs.

The end of the Cold War bolstered the normative commitment to democracy through global institutions, and IOs expanded in number and prominence (Kadera, Crescenzi, and Shannon 2003; Pevehouse 2005; Milner 1997; Shannon, Morey, and Boehmke 2010). States could use membership in IOs to mount stronger, multilateral efforts for desired results. Such a contention is consistent with a recent study from Shannon and her colleagues (2015), which shows that IOs became the preferred mouthpiece for postcoup international signaling following the end of the Cold War. Due to the democratic nature of these organizations, we should also expect to see an increase in the likelihood of international reactions to push for democratization in coup states, consistent with the findings of Marinov and Goemans (2014) and Thyne and Powell (2016). Indeed, a variety of transnational frameworks were adopted with the intention of explicitly deterring coups. Although individual countries like the United States implemented policies that called for the suspension of both military and foreign aid against regimes that seize power through a coup, regional IOs took the most forceful steps following the end of the Cold War. ² OAS General Assembly Resolution 1080 (1991) provided the first comprehensive attempt by regional actors to guarantee constitutional power transfers by recommending a comprehensive set of punishments for coup-born governments. ³ The overwhelming response to the aforementioned Haitian case sent an example to others in the region. During the 1996 Paraguayan constitutional crisis, for example, some observers saw it as a foregone conclusion that General Lino Cesar Oviedo would seize power. Although Oviedo’s rivalry with President Juan Carlos Wasmosy reached a boiling point, he did not seize power, reportedly claiming to his colleagues that “with the OAS’s adoption of Resolution 1080 in Santiago, the era of Latin American coups had come to an end” (Valenzuela 1997, 54).

The Organization of African Unity (OAU) and, later, the AU took similar steps, though early OAU efforts were far less consistent than that of the OAS. ⁴ After being given the opportunity to pursue a Haitian-type precedent when a 1999 coup unseated Henrie Konan Bedie in Cote d’Ivoire, the organization allowed the coup leader to attend its subsequent meeting. The 2002 launch of the AU more clearly specified the required organizational response, and the AU has grown more consistent in punishing coups following the establishment of its Peace and Security Council (Omorogbe 2011; Souaré 2014).

The growing strength and consistency of reactions to coups from international actors is important because they provide important information to coup plotters, both prior to and in the aftermath of a coup. In the aftermath of coups, reactions from the international community can act to either stabilize or undermine the new regime. Strategic state interests dominated the landscape during the Cold War, creating responses that varied in both direction and strength. Following the conclusion of the Cold War, however, policies from IOs helped develop norms to condemn coups in the international community. This discussion leads to the following hypothesis:

Hypothesis 4: The influence of state reactions to coups on regime duration should have the strongest influence during the Cold War, and IO reactions should have the strongest influence afterward.

Precoup Relationships

Our final concern focuses on precoup relationships. The discussion thus far considers how reactions from domestic and international actors influence the postcoup tenure of coup leaders. These reactions can indeed be unpredictable and fleeting (Morey et al. 2012); however, it would be unwise to assume that coup leaders ignore precoup information when launching the putsch. Instead, we suspect positive precoup relationships to strengthen the relationship between international reactions to coups and the tenure of coup-born regimes.

The most direct way that external actors can shorten the life span of coup-born regimes is by imposing a new regime by force, as we saw the United States do in Panama. Undertaking such maneuvers requires both ability and will, however, which is not constant across states that may react harshly to coups. US condemnation of the 2000 Ecuadorian coup likely influenced the junta more than a similar reaction from the UK, for example, due to US strength and its long history of influence in Latin America. Seeking to avoid the fate of leaders like Hudson Austin in Grenada or Johnny Paul Koroma in Sierra Leone, we expect coup-born regimes to be short lived when they overthrow a regime that had friendly relations with major powers.

A secondary way that precoup relations matter is less direct, though similarly important for the survival of a coup-born regime. States form security arrangements to protect their long-term interests. The Egypt–Israel (1979) peace treaty, for example, not only normalized relations between the two states but also started the flow of economic and military aid from the United States to Egypt. Over time, relations between Egypt and Israel warmed, and the two states now cooperate extensively in counterinsurgency operations (Schenker 2015). Given the importance of this relationship, it is no surprise that military leaders immediately promised to uphold international agreements after ousting Presidents Mubarak and Morsi, and it is unsurprising that Israel reacted positively to the coups following such overtures (Fahim 2011; Times of Israel 2014). Had either the United States or Israeli responses to the Egyptian coups been hostile, we suspect the military’s run in office would have been appreciably shorter. This discussion yields our final hypothesis:

Hypothesis 5: As precoup ties between coup states and powerful international actors increase, the influence of negative postcoup reactions from international actors on the tenure in office for leaders who came to power via a coup d’état should increase.

Independent Variables

We first expect leadership duration to decrease if the domestic population responds negatively to the coup. We followed four steps to capture domestic protests. First, we began with data from the Social, Political, and Economic Event Database (SPEED) Project. ⁸ Coded for 165 countries in the post–World War II era, these data focus on “human-initiated destabilizing events,” which are defined as “happenings that unsettle the routines and expectations of citizens, cause them to be fearful, and raise their anxiety about the future” (Nardulli, Hayes, and Bajjalieh 2013, 1). SPEED data are collected using a hybrid computer–human approach with autoextraction of news reports that are categorized into event data by human coders. ⁹ Coded events contain information to analyze a variety of research questions, including event type (e.g., political violence, terrorism), location, government response, and fatalities.

The unaltered SPEED data set begins with 62,074 events coded by precise date range, location, initiator, and target (inter alia) from 1946 to 2005. For our domestic protests variable, we are interested in how nonstate, domestic actors responded to the government. Thus, our second step was to purge the data to capture only events where domestic actors demonstrated a negative response (either “political expression” or “political attacks”) toward the government, which reduced our observations to 18,509 events. Having defined the event type of interest, our third step was to capture the number of protest events that took place in each month during the postcoup period, which we define as the first six months following the coup. ¹⁰ Finally, we filled this number down beyond the six-month postcoup period, assuming that each response lasted until the end of the regime’s tenure (or the end of the data set for censored cases). ¹¹ This resulted in 4,892 protest months, which represents 29.4 percent of the sample. We expect this measure to produce a negative and significant coefficient to support our first hypothesis.

Our second hypothesis predicts that the duration of postcoup leadership tenure will decrease if external actors react to the coup in a hostile manner. We test this expectation using data from Shannon et al. (2015), who code international responses to coups from IOs and states during the six-month postcoup period. As with the measure for domestic reactions, these authors began with the Powell and Thyne (2011) data set and then coded all international responses to coup states using the Goldstein (1992) scale, which ranges from −10 (most hostile) to +8.3 (most supportive). These scholars searched sources “by hand” to look for instances where international actors responded specifically to coups. Their final measure captures 1,259 reactions to 98 postcoup periods. To remain consistent with the measure for domestic protests, we multiplied this measure by −1, so that positive values capture increasing levels of international protests and then captured the mean signal sent from all international actors during the six-month, postcoup period. ¹² The final measure has a marginally supportive mean of −.133 (s = 1.87). As with the measure for domestic protests, we expect to see a negative and significant coefficient to support our second hypothesis.

Although the mean reaction from international actors captures the influence of international protests on regime tenure, this measure does not necessarily capture agreement, as suggested with our second hypothesis. We more precisely capture agreement among international actors by interacting the mean international response with the standard deviation (SD) in responses (Intl prot. × Int prot. SD). We expect to see a positive coefficient for this interaction term, which would indicate that the regime-shortening effect of hostile international reactions is weakened if there is disagreement among international actors. Likewise, we test our third hypothesis by interacting domestic protests with international protests (Domestic prot × Intl prot). We expect to see a negative and significant coefficient for this interaction term, suggesting that the negative influence of domestic/international reactions on coup leader duration strengthens when the reactions are consistent with international/domestic reactions.

Our final hypotheses focus exclusively on reactions from external actors, though we retain domestic protests as a control variable when testing these hypotheses. Our fourth hypothesis predicts state reactions to matter the most during the Cold War, and for IO reactions to have the strongest influence afterward. We test this expectation by splitting international reactions into two categories, states and IOs, and then splitting the sample between years prior to 1989 and those afterward.

Finally, we expect international signals to have the strongest influence on postcoup tenure if they come from states that have precoup ties with powerful international actors. We control for IO reactions and then focus on states to capture this concept in three ways. First, we split the measure for international signals into two groups: major powers and nonmajor powers, as defined by the Correlates of War project (2011). This approach assumes that major powers have meaningful ties with all postcoup states, which may not necessarily be true. Thus, we more precisely capture precoup ties by first coding whether or not the coup state had a meaningful trading relationship with the reacting state (Barbieri and Keshk 2012; Barbieri, Keshk, and Pollins 2009). We define a relationship as meaningful if the coup state’s trade with the reacting state comprised at least 10 percent of its total trade. We then split international reactions into two categories: (1) reactions from trading partners and (2) reactions from nontrading partners. Like the trade measure, we also captured whether or not the coup state had a formal alliance with the reacting state (Gibler 2009). We then split international reactions into two categories: (1) reactions from allied states and (2) reactions from nonallied states. We expect reactions from major powers, trading partners and allied states to have the strongest tenure shortening effect to support our fifth hypothesis. ¹³

Control Variables

We include two sets of control variables that are meant to capture other causal processes that might explain coup leader duration. ¹⁴ The first set captures the postcoup climate by drawing on previous studies of leader duration (e.g., Escribá-Folch and Wright 2015). Our first two measures, gross domestic product (GDP)/capita and yearly Change in GDP/capita, come from K. S. Gleditsch (2002). Following previous studies, we predict that leaders who enjoy a strong economy, measured in either static or dynamic terms, are more likely to retain their leadership positions (Cheibub and Przeworski 1999; Chiozza and Goemans 2004). Next, previous work has discovered that large populations are more difficult to govern (Fearon and Laitin 2003); thus, we expect leaders in states with high populations to have shorter tenures (K. S. Gleditsch 2002). Finally, even failed coups have been found to destabilize leaders (Belkin and Schofer 2003; Powell 2012). We thus include a measure counting the number of failed coups that took place in the previous two years.

The second set of control variables captures the precoup status of the state. These measures are important because they capture the strategic, forward-thinking nature of coup plotters. ¹⁵ When coups are launched during protests, coup leaders might rightly expect domestic support for the putsch because they are likely to be seen as guardians of the people. Similarly, precoup sanctions indicate that international actors are likely to support regime upheaval, so must hold the precoup status of domestic and international viewpoints constant in order to gauge the exogenous effect of postcoup reactions on regime tenure. We capture the precoup domestic climate using the same data as our domestic protest measure. Precoup protests is a dummy that captures whether a protest event took place in the month preceding the coup. For international viewpoints, we turn to the threat and imposition of economic sanctions data to capture precoup sanctions (Morgan, Bapat, and Kobayashi 2014). Capturing events like embargoes, travel bans, and cutting foreign aid, this data set captures the precise dates that an international actor threatened or applied sanctions against another state. We include a dummy variable that indicates whether sanctions were applied or threatened in the month preceding a coup. We might also expect negative domestic and international reactions to be more likely if the coup overthrew a democratic regime. Previous experience with democracy might likewise make the coup-born government less likely to solidify their power under authoritarianism. We include a measure called precoup democracy, which equals one if the state was a democracy (+6 or above on the Polity IV’s polity2 measure) prior to the coup (Marshall, Jaggers, and Gurr 2011). Similarly, plotters may be able to anticipate responses based on how the economy is doing, expecting greater support for the putsch in more dire periods. We consequently include measures for precoup GDP/capita and the yearly change in GDP/capita (K. S. Gleditsch 2002).