Abstract
In this article, I move beyond prior efforts to explore the relationship between the risk of a coup and international conflict by considering alternatives that leaders can utilize to strengthen their regimes. I offer two theoretical expectations. First, I theorize that leaders lose the incentive and ability to use diversion when the structural coup-proofing apparatus is strengthened. Second, I expect military finances to lead to disparate behavior when considering regime type. Autocrats are expected to use military funds to provide private incentives to the armed forces, largely in the form of allowances. Democracies, in contrast, will be required to use expenditures to promote the public good of national security due to the transparency of their regimes. Autocrats are expected to lose the incentive to use diversion as the financial endowment of their militaries increase, while democracies will continue to show a diversionary trend due to their increased military capabilities. The theory is tested using global data from 1962 to 2000, with the findings strongly supporting the theory.
And by whose power I well might lodge a fear to be again displaced; which to avoid, I cut them off; and had a purpose now to lead out many to the Holy Land, Lest rest and lying still might make them look too near unto my state. Therefore, my Harry, be it in thy course to busy giddy minds with foreign quarrels; that action, hence borne out, may waste the memory of the former days. —Henry IV
Henry’s words to his son are illustrative of a common theme in conflict studies: the tendency of states to target an external scapegoat to distract from problems at home. The necessity of diversion, defined by Levy (1989) as the use of force for domestic political purposes, has largely been attributed to the potential to lose office through electoral cycles (e.g., Stoll 1984; Gaubatz 1991; James and Hristoulas 1994), usually in the context of the United States. Considerably less attention has been given to the theoretical processes that might promote diversionary activity in autocracies. A king such as Henry, for example, does not have to face the ballot. However, all leaders face removal to varying degrees, and those with a higher likelihood of being removed from office will be more likely to enter a dispute for diversionary reasons. The audience for diversion, however, differs in accordance with the likely means of removal.
While democratic leaders will attempt to divert the public by promoting a rally round the flag effect, those facing nonelectoral modes of regime change will need to consider those who can help to keep them in power or expedite their ouster. If we are to explain diversion in autocracies, we must more directly capture scenarios under which autocrats face a legitimate threat of being removed. Elections might be unlikely to unseat an authoritarian leader, but they often fall by other means of removal, such as the coup d’état. Decalo (1990, 547), for example, referred to coups as the “most visible and recurrent characteristic” of African politics, while Svolik (2009) shows that over two-thirds of authoritarian regimes end via coups. The coup is clearly an important piece of the diversionary puzzle, and recent scholarship has begun to put more emphasis on this dynamic (e.g., Miller and Elgün 2010).
Entering a dispute is only one method by which leaders can lessen the ability of would-be putschists to oust them, and a variety of commentaries have attested to “coup-proofing” strategies that states can implement. This study investigates two broad approaches that leaders can utilize to reduce the likelihood of a coup attempt: the use of structural coup-proofing efforts like armed counterweights to the armed forces and manipulation of military finances. Nielsen (2005) has suggested civil–military relations scholars have put far more emphasis on considering how to maintain civilian control of the armed forces than on considering how civil–military relations influence military capabilities. In the following pages, I suggest that civilian control strategies are critical determinants of military capabilities. Two trends are noted. First, the use of “structural” obstacles like counterbalancing will degrade a state’s military capabilities by redirecting important resources away from the regular armed forces and by creating coordination obstacles within the military. These trends both undermine the ability to attempt a coup (Powell 2012) and reduce a state’s effectiveness in combat (Pilster and Böhmelt 2011). Second, leaders can manipulate military finances in an effort to reduce their willingness to target the incumbent. However, an important distinction is worth noting. Autocrats will largely be immune from oversight and will be free to utilize military expenditures to provide a private good for a critical segment of the winning coalition: the military. Democracies, however, have a number of checks that will not allow such free use of expenditures. Instead, funds directed to democratic militaries will have to be dedicated to providing a verifiable public good: national security. Authoritarian leaders will lose the incentive to utilize diversion as military finances are strengthened, while democratic leaders will be equipped with an army that is particularly well suited for combat. These disparate trends have important implications for both diversionary theory and the study of the military effectiveness of democracies.
The remainder of this article attempts to theoretically explain and test the relationship between coup vulnerability and international conflict. I begin by briefly outlining the relevant diversionary literature, keeping in mind how prior studies relate to the theoretical offerings of this article. Second, I explain how leader survival strategies should impact the military capabilities of the state and, consequently, diversionary activity. Finally, I test the diversionary tendencies of states given their vulnerability to a coup d’état and the extent of coup-proofing in their regimes. The analyses suggest that regimes with heightened vulnerability to a coup do in fact initiate disputes for diversionary purposes. However, this lessens as alternative coup-proofing efforts are strengthened. Democracies are a notable exception, as well-funded militaries are particularly likely to be used in at-risk democracies.
The Utility of Diversionary Conflict
Considerable attention has been given to the characteristics of states involved in diversionary conflict. Recent scholarship has moved beyond elections by considering aspects such as executive–legislative relations (Brulé and Hwang 2010), but diversionary theory has still primarily attempted to explain the behavior of democracies. Indeed, even the rebuff of diversionary theory offered by Meernik and Waterman (1996) largely limits itself to democracies by pointing out states should refrain from diversionary use of force due to the “immoral and criminal” repercussions involved. Nondemocracies may be less concerned about violating international norms or laws, especially when confronted with threats to their rule.
The conditions that are expected to influence diversionary conflict are those that will impact a leader’s prospects for maintaining power and that different states will have different turmoil thresholds that will allow them to tolerate crises to varying degrees (Morgan and Bickers 1992). On the surface, it would seem that autocrats might have a high threshold given their insulation from democratic processes, but recent work has begun to make note of the theoretical processes behind authoritarian diversion. Mitchell and Prins (2004), for example, theorize that autocrats are more likely to use diversionary conflict because the transparent nature of democracies will allow potential targets to strategically avoid an altercation. Identifying the motivation to accompany this opportunity has remained more elusive, though there are exceptions. Miller (1995) surmised that authoritarian regimes must utilize diversionary war to deal with domestic dissent since they have few other means to deal with crises. This was later extended to show that states ruled by military strongmen or juntas are more likely to use force than even other authoritarian governments (Lai and Slater 2006). They explain this trend by theorizing that military governments lack institutional capability and must resort to “desperate measures” when attempting to quell dissent. The source of dissent, of course, will vary by regime. Morgan and Bickers (1992) have suggested leaders should act differently when faced with challenges from different groups and show US foreign policy takes a diversionary turn during low levels of partisan support. Other efforts have similarly pointed to the importance of elites in the US context. Gowa (1998), for example, demonstrates the importance of party control of the Congress, while Brulé (2006, 2008) points to periods of low approval and high legislative constraints. 1
The literature regarding democracies is intriguing in that it points to the importance of cohesion within the political elite even though it is the masses that will ultimately determine political survival. Elite cohesion should be of particular importance for authoritarian regimes because the primary survival threat is within the state apparatus or ruling coalition.
The ruling coalition has been dealt with most explicitly in the context of authoritarian diversion by Pickering and Kisangani (2010) and Nicholls, Huth, and Appel (2010). The political incentive theory of the former predicts that autocrats will be more belligerent when the distance between available resources to provide patronage and the amount spent on patronage becomes small. They conclude that authoritarian regimes with the largest winning coalitions, single-party states should be the most likely to utilize international conflict in times of economic crisis due to the strains that their coalition size puts on patronage resources. Though theoretically compelling, their analysis fails to find robust support for their theory of “despotic diversions.” One potential explanation is the lack of vulnerability in one-party regimes, which has been found to be among the most stable in existence (Geddes 1999; Hadenius and Teorell 2007). As Pickering and Kisangani (2010, 479) note, “the military’s subservient state within the hierarchy is almost always well established” in single-party regimes. The chances for a coup, then, are minimal, calling into question the claim that single-party states should have the incentive to attempt diversionary actions, even during a time of crisis. Nicholls, Huth, and Appel (2010) have more directly argued that opposition from within the ruling coalition was a major determinant of Japanese foreign policy following the establishment of the Imperial Diet. Diversion, they argued, would be more likely when leadership was actively rejecting demands from other elites. I similarly contend that we must look specifically at the likelihood of a credible movement to unset them, specifically through a coup.
The suggestion of diversion in the face of coup risk has been most directly tested in the quantitative literature by Miller and Elgün (2010), who find strong evidence that the risk of a coup does in fact promote the initiation of militarized interstate disputes (MIDs). However, background factors that might increase the likelihood of a coup are only one part of the leader’s decision-making calculus. Also important are the other options that leaders can implement in order to survive. The following section considers two broad coup-proofing strategies that can be seen as influencing the decision to initiate a dispute.
Diminishing Conflict Utility through Coup Proofing
A number of factors will influence a leader’s vulnerability to a coup and his or her perceived need for diversion. Belkin and Schofer (2003) have acknowledged that “coup risk” is a potentially loaded term and that their measure for the concept (discussed below) does not capture opportunity. It merely captures background factors that might predispose a state to have a coup. Background factors (e.g., prior coup history) can potentially be exacerbated by or offset by the characteristics of the security apparatus. This section looks at background characteristics of the state and civil-control strategies that attempt to guarantee the subjugation of the armed forces.
A number of works have investigated the processes of civil–military relations, especially regarding sustained civilian authority. Perhaps, the most influential is that of Huntington (1957), who believed that objective civilian control would maximize both military effectiveness and continued authority over the military. Subsequent theoretical efforts have almost exclusively focused on the latter of these outcomes, giving little attention to how civil–military relations influence military capabilities (Nielsen 2005). This trend is surprising, giving the tendency of leaders to willfully alter military capabilities in order to guarantee their own political survival. This willingness can be seen as an extension of the most basic goal of politicians. In the context of democracy, Downs (1957, 11) assumes that “every government seeks to maximize political support” and that “its primary goal is reelection . . .”. To attain and maintain power is the first order of business for those in office, being more important than even strengthening the state. Once in power, leaders will seek out “the most efficient means” by which they can maintain their rule. Just as leaders operating under the rules of democracy will promote public goods in order to maximize their support among voters, those in regimes facing a high likelihood of a coup will seek out a strategy that will directly account for that particular means.
Not vulnerable to the citizenry, autocrats will instead have an incentive to provide private goods to those who can credibly threaten their survival. Policy will reflect this goal, not the overall well-being of the state. For example, the “omnibalancing” theory of alliance formation argues that theorists erroneously assume a leader asks, “How does this policy affect the power of the state?” He or she should instead be asking, “How does this policy affect the probability of my remaining in power,” as many leaders “will sometimes protect themselves at the expense of the interests of the state” (David 1991, 236). Such an approach reflects Downs’s belief that individual survival interests prevail and David believes that third world leaders form alliances with states that are most likely to keep them in power. A similar trade-off is seen in domestic coup-proofing.
The available options for deterring coups are limited. Legal provisions placing restraints on the military, for example, are useful only to the extent that the military is willing to conform to the law (Feaver 1999). Attempts to professionalize the military are similarly flawed, as such efforts can take a considerable amount of time (Huntington 1957). Further, professionalization could actually increase the military’s ability to attempt a coup. For example, Mobutu Sese Seko was heavily committed to promoting a professional fighting force following his assumption of power. He dramatically increased military spending in a modernization effort that saw the armed forces’ prestige increased. Also implemented were efforts to “ensure that the military’s role remained restricted to external defense,” and promote education in the officer corps by having hundreds of its members given tutelage at the “best military schools in the west” (Kisangani 2000, 211). Finally, the patronage politics for which he would become famous remained separated from the military, a fact that would soon be resented by those in uniform. In short, Mobutu began down a path that would have proved exceptional in region: promoting a modernized, professional armed forces. However, he ultimately shifted policy after facing strong resistance from the army high command after promoting Leonard Mulamba to prime minister. Mulamba’s appointment was retracted and Mobutu further agreed not to appoint him to the defense ministry (Williame 1970, 149). The Mulamba affair led to a drastic shift in which coup fears were met not with efforts to promote professionalism, but rather alternative strategies that would serve to both strengthen his grip on power while reducing his military’s capabilities.
The coup-proofing efforts that Mobutu and others have utilized can be categorized as either “those that affect the disposition of the military to subvert” or those that reduce “the ability of the military to subvert control” (Feaver 1999, 225). Consideration of both aspects of coup-proofing is important for two reasons. First, they are both theoretically expected to reduce the utility of diversionary conflict, as a vulnerable leader will grow more secure with an increase in either strategy. Second, the respective categories have important implications for fighting capacity. A growing body of literature suggests that efforts to reduce the ability to undertake a coup will similarly reduce the ability of militaries to function in combat (Welch 1976; Biddle and Zirkle 1996; Feaver 1999; Quinlivan 1999; Pilster and Böhmelt 2011). In addition to lowering the utility of diversion, these strategies can lower the capacity for undertaking it. Meanwhile, efforts to influence coup disposition are traditionally seen as having no ill effect on the fighting capacity of the regular armed forces (Feaver 1999). I offer a partial critique of this assumption by considering theoretical differences between democracies and authoritarian regimes.
The following two sections describe the impact of coup proofing on military effectiveness. I begin with a discussion of structural coup-proofing efforts that see critical resources redirected to paramilitary units, increased coordination obstacles, and a reduction of professionalism in the regular armed forces. I then shift to a discussion of coup proofing through “spoils.” Though prior scholarship has suggested that efforts to target coup disposition have no ill effect on capabilities, I amend this claim by showing the impact of military spoils is dependent on regime type. Autocracies that provide increased resources to their militaries are expected to see a negligible increase in capability, while democracies will see a substantial boost in military effectiveness. These have important consequences for diversion and, more generally, conflict.
Structural Coup Proofing
Coordination of efforts is a necessity in a struggle between groups. In large groups it can be achieved only on the basis of subordination and unitary command. The proposition that unitary command is advantageous in war is so evident that no elaborate proofs are required. —Stanislav Andreski (1968, 92)
Structural coup-proofing efforts include “setting various branches against each other,” “using parallel chains of command,” or using numerous institutions such as “border guards, secret police, paramilitary forces, militaries, presidential guards, and so on” to act as an armed counterweight to the military (Feaver 1999, 225). These efforts require even the most willing putschists to carefully evaluate the prospects of successfully executing a coup. Geddes (1999) has noted that the worst possible outcome for a military is to have to fight other factions of the armed forces, and over half of successful coups reported by Marshall and Marshall (2009) are indeed bloodless. This reality has led many leaders to reduce the ability of their armed forces to attempt a coup by providing a number of structural obstacles to organizing and executing a conspiracy. These structural divisions are expected to reduce both coup likelihood and military effectiveness in four ways.
First, structural coup proofing creates obstacles in the form of armed counterweights. These require conspirators to approach and gain the cooperation of other armed bodies and in many cases overtake them by force during the coup. Belkin and Schofer (2003, 596) point to these “counterbalancing” organizations as being “mutually suspicious rival forces that check and balance on another.” For example, Wrong (2002, 258) pointed to Mobutu’s “multiplication of special units and security organizations, often vying for identical duties.” These organizations included “the DSP, the Garde Civile, SARM, the Kamanyola division, the para-commandos, the 21st brigade, the 31st brigade, SNIP, and, bringing up the rear, the gendarmerie . . .” (Wrong 2002, 258). These rivalries will see multiple security organizations refuse to cooperate with their counterparts.
Second, the effectiveness of specific organizations is undermined by an inability to undertake coordinated action. Unity of command has been noted by Durell-Young (1997, 23) as a “sine qua non for the successful prosecution of military operations” in regard to coalition efforts, and the work of Pilster and Böhmelt (2011) note that effective interunit cooperation can only be accomplished through joint peacetime training. Coup-fearing leaders ensure that different organizations in the security apparatus are not experienced in coordinating. Returning to the Zaire example, challenges to interbranch communication were so well established that the Zairean air force accidentally bombed its own ground soldiers just prior to Shaba I (Young and Turner 1985, 252). In terms of international conflict, Muammar Qaddafi similarly undermined the fighting capacity of his own forces while at war with Chad by “frequently and unexpectedly” rotating command positions and division-level commands had to be created “on an ad hoc basis in the field” (Pollack 2002, 386). The Libyan case indicates Qaddafi was willing to take these coup-proofing measures while the country was already at war. Personal survival was clearly of more importance than was state military capabilities, and Libyan forces found “concerted” action to be “nearly impossible,” and failed to undertake a single tactical counterattack against the Chadian army (Pollack 2002, 417).
Third, structural coup proofing is also accompanied by the ancillary factor of reduced quality of individual soldiers. The literature is rife with examples of structural obstacles being accompanied by selective purges of the armed forces (e.g., Horowitz 2002; Roessler 2011). Mobutu, for example, purged the “old oligarchy,” with many being executed after the Mulamba affair (Kisangani 2000, 214). These included what Young and Turner (1985, 266) describe as Mobutu’s “ablest soldiers.” Restocking the military would put “political considerations” over “military ones,” with promotions isolated to those with “enthusiasm for the existing regime” (Kisangani 2000, 215–16). Such efforts undermine the quality of new recruits and reduce individual initiative (Pilster and Böhmelt 2011).
Fourth, crucial military resources are redirected from the regular armed forces to the coup-proofing portion of the security apparatus. Though these units will often be well trained and well equipped, leaders are unlikely to deploy these soldiers away from the capital except for the gravest of circumstances. Quinlivan (1999, 145), for example, noted Hussein’s reluctance to deploy Iraq’s Republican Guard despite the Popular Army having “frequently collapsed” in combat. Mobutu even refused to deploy his own coup-proofing apparatus beyond the capital during the Shaba conflicts and against the westward march of Democratic Alliance for the Liberation of the Congo.
While coup risk is expected to be an important determinant of conflict initiation, structural cohesion obstacles in the military will reduce the necessity of diversion as well as undermine the ability to utilize it.
Military Financing
In all countries officers are deeply concerned with their material status . . . Give them toys. That is, provide them with fancy tanks, planes, armored cars, artillery, and sophisticated electronic equipment. New equipment will make them happy and keep them busy learning how to operate it. —Samuel Huntington (1991, 250–52)
Huntington (1991) has made a number of relevant suggestions for democratizing states that wish to avoid praetorianism. First, he warns that soldiers will think “they are badly paid, badly housed, and badly provided for—and they are probably right” (Huntington 1991, 252). Regimes should address these deficiencies by reducing the size of their militaries and increasing individual salaries, pensions, and benefits. Second, leaders should give their militaries “toys” such as “fancy tanks, planes, armored cars, [and] artillery” (Huntington 1991, 252). Desch (1999) has similarly claimed that states would be more immune to military interventionism when “supplying sufficient resources” for external missions, and Powell (2012) finds that increased expenditures per soldier is an effective method for reducing coups.
However, the influence of expenditures on military capabilities is expected to be contingent on regime type. Democratic leaders, to whom Huntington was making suggestions, will be more likely to dedicate military expenditures to provide the public good of national security. The primary driver of this distinction is transparency, as third parties like the media and think tanks will keep a watchful eye on military policy (Pilster and Böhmelt, forthcoming). Democratic leaders may have the incentive to use military expenditures as a patronage resource, but the inherent restraints that accompany transparency will keep these efforts to a minimum. A lack of such constraints gives authoritarian leaders considerable freedom in defense spending and two major trends have been noted in the squandering of military financial resources.
First, Henk and Rupiya (2001, 18) point to “an overwhelming proportion” of spending that goes to salaries and personnel allowances in Africa. For example, Zimbabwe has seen nearly 70 percent of expenditures go toward officer allowances and was a typical case on a continent whose militaries are “chronically short of ammunition, fuel, other basic supplies, and spare parts” (Henk and Rupiya 2001, 18). This trend is seen beyond Africa as well. Saudi Arabia, for example, illustrates a coup-fearing leadership that ultimately became the highest per-capita defense spender in the world; yet, as little as 5 percent of these expenditures were dedicated to military hardware (Pollack 2002, 427). Instead, finances were mainly directed and keeping soldiers happy, not skilled.
Second, a considerable amount of money is dedicated to symbolic spending. These expenditures are put toward bona fide military hardware, but these resources often remain useless due to a lack of adequate training or a lack of upkeep. The point of sophisticated weapons systems was at best a symbolic effort, a trend that is more difficult in democracies. Democratic South Africa, for example, saw the government become the subject of intense scrutiny following the purchase of submarines and Swedish fighter jets, both of which acted to increase the prestige of the South African Defense Force but had little connection to extant national security concerns.
On the other hand, Henk and Rupiya (2001) have pointed to the lack of a legislative mandate and expertise to monitor defense spending in the average African state as a major impediment to the quality of the continent’s armies. Furthering this problem is the danger of attempting to institute stronger oversight, as efforts of authoritarian leaders to reduce patronage could act as a catalyst for military intervention. Milton Obote, for example, offered the British Broadcasting Corporation (BBC) a rather direct explanation after being ousted in Idi Amin’s 1972 coup (Lofchie 1972, 24):
a person very close to Major General Amin . . . ordered materials worth Shs 40,000,000 [approximately $6 million]. There are no documents, no copies of the invoices, no copies of any delivery note . . . On the same day I left for Singapore, I asked General Amin, on my return, to give me a written report on how the Shs 40,000,000 was to be spent. I have no doubts at all that what is now developing in Uganda is another attempt to hide the loss of Shs 40,000,000 and an attempt to prevent me from getting back to the country and punish the culprits.
In summary, the preceding theory has offered three predictions regarding the diversionary behavior of leaders. First, leaders whose regimes lack structural coup-proofing strategies will have an increased incentive to enter disputes. Conflict should be less likely as these strategies are strengthened. Second, authoritarian leaders with high coup risk are expected to utilize diversion, but the incentive and ability to do so will decrease as they provide their militaries with more financial resources. Third, democratic leaders who dedicate more financial resources to their militaries are expected to be more capable militarily, leading to a consistently positive use of military diversion.
Research Design
The analysis is conducted within a pool of politically relevant directed dyads in order to capture the dyad member that is the initiator of the dispute and to account for relevant control variables (e.g., joint democracy, power ratio). Directed dyads are not only attractive but also essential. 2 Availability of the independent variables limits the scope of the study to 1962–2000, though different specifications will see fluctuations in temporal range.
Multiple specifications are used to account for conflict. For the diversionary literature alone, conflict has been treated as “diplomatic rebuffs” (Collins 1973), dispute initiation (Chiozza and Goemans 2003), “low-level uses of force” (Morgan and Anderson 1999), and war (James 1987). Diehl (2006, 204) has hinted that inconsistent findings in the diversionary literature could be due to the failure to distinguish between dispute initiation and escalation, as some leaders wishing to “wag the dog” are not willing to pay the higher costs associated with war. Morgan and Bickers (1992) have also suggested diversionary tactics should come short of war, a sensible assumption given the negative consequences of casualties on public opinion (Gartner and Segura 1998, 2000; Gartner 2008) and the potential to be ousted due to military fiasco (Goemans 2000).
Given these considerations, I primarily rely on the MIDs data set (Ghosn and Palmer 2003). For the purposes of this study, MID initiation is coded as 1 if state A at a minimum issued a threat of force toward state B in a given directed dyad-year and is coded 0 if no dispute is recorded. More belligerent classifications of MIDs are also included and only the first year of the MID is used in the analysis. Alternative specifications of the dependent variable are included to test the robustness of the findings. First, I consider a higher level of conflict intensity by including war, or the initiation of a MID that resulted in over 999 battle deaths. Next, I include two dependent variables from the International Conflict Behavior (ICB) data set (Brecher and Wilkenfeld 2010). An international crisis (IC) refers to events in which state A acted as the trigger for a foreign policy crisis, defined as an “act by a state as creating a basic threat to values, time pressure, and heightened probably of military hostilities” (Brecher and Wilkenfeld 2010, 14). A violent international crisis (VIC) considers only events that included a violent military act, such as a border clash, invasion of air space, bombing of a large target, or war. Due to the dichotomous and rare nature of the dependent variable, a rare events logistic regression is employed (King and Zeng 2001). 3
Coup risk is adopted from Belkin and Schofer’s (2003) index accounting for country’s structural vulnerability to a coup d’état by considering its suffering of a coup attempt in the previous five years, the strength of civil society, and regime legitimacy. The existence of a recent coup has been widely accepted as a predictor of future coups (Londregan and Poole 1990). Civil society is treated as the number of international nongovernmental organizations and legitimacy as leadership competitiveness as defined by Polity. The index omits some factors that have been found to have an impact on coups (e.g., wealth and economic growth), but these omissions are not problematic. Belkin and Schofer were more interested in structural characteristics that would allow crises to matter, not the specific economic or political crises themselves. Given the need to control for economic performance in diversionary conflict models, the exclusion makes the Belkin and Schofer measure particularly attractive.
Feaver’s (1999) summary of civil–military relations has simplified coup-proofing efforts to two categories: efforts that attempt to reduce the ability of a military to intervene and those that reduce the disposition of the military to intervene. Two measures are included to account for structural obstacles that will decrease the ability to execute a coup. I first use the Pilster and Böhmelt (2011) index for the number of effective organizations, which can be thought of as ground-capable military bodies that could credibly combat a coup. I also include Belkin and Schofer’s (2003) counterbalancing variable. This is a continuous variable that considers the ratio between regular and paramilitary forces, as well as the number of personnel in those forces. The first hypothesis predicts the use of diversion should subside as structural coup proofing increases.
The financial aspect of coup proofing considers military expenditures per soldier as a measure for spoils. This measure has been used by a variety of conflict scholars as a measure for “troop quality” 4 and has been found to reduce the likelihood of a coup (Powell 2012). To be clear, the expenditures per soldier measure is not meant to be a proxy for military professionalism in the spirit of Huntington (1957), as this is assuredly not the case in authoritarian regimes. It can be more generally thought of as a measure of contentment with the regime, though it is expected to be accompanied by an increase in fighting capacity in the context of democracies. The additive impact of coup risk on conflict initiation is expected to subside as spoils increase in authoritarian regimes (Hypothesis 2a), while spoils are not expected to significantly reduce diversion in democracies (Hypothesis 2b).
A number of necessary control variables are also included. Wealth refers to the logged value of gross domestic product per capita (GDP), held constant in 1996 dollars (Gleditsch 2002). Growth measures the year-to-year change in real per capita GDP. A benefit of dyadic analysis is the ability to control for important factors such as alliances, joint democracy, power ratio, peace years, and distance. Allies are coded with the Alliance and Treaty Obligations (ATOP) data set and is coded 1 if the dyad members are considered to be in a “formal agreement . . . to cooperate militarily in the face of a potential or realized military conflict” (Leeds 2005, 4). Joint democracy is a dichotomous variable coded 1 if both members of the dyad score above +5 on the Polity IV scale. I also incorporate dummy variables for whether the initiator (Dem. Initiator) or target (Dem. Target) is a democracy. Power ratio relies on Correlates of War's (COW's) Composite Index of National Capabilities (CINC). A state’s CINC score is constructed by aggregating the number of military personnel, military expenditures, iron and steel production, energy consumption, and total and urban population. For power ratio, I divide the initiator’s capabilities by the total capabilities of the dyad. A value of 1 would indicate the initiator has a complete preponderance of power over the other dyad member. A state is believed to be more likely to initiate a dispute as the power ratio increases. Peace years records the years since a conflict was last initiated in the dyad. I also incorporate squared and cubic polynomials for peace years at the suggestion of Carter and Signorino (2010). Finally, I use the natural log of the distance between capitals to address geographical constraints on conflict. As two states are farther from one another, they should be less likely to wage conflict. 5
Results
In this section, I report the results of the models suggested by the hypotheses. I begin by offering a base model that assesses the impact of coup risk on international conflict while omitting coup proofing. I look at all versions of the dependent variable: MIDs (model 1), war-level MIDs (model 2), international crises (IC; model 3), and violent international crises (VIC; model 4). These alternative measures will allow me to check the robustness of the findings across different data collection efforts, different conflict levels, and compare the fit of the naïve model to the later specifications that account for coup proofing. Results are presented in Table 1.
The Impact of Coup Risk on International Conflict, 1962–2000.
Note. ICD = International Conflict Behavior; LR = log-likelihood ratio; MIDs = militarized interstate disputes; VIC = violent international crisis. Models reflect rare events logistic regression. Robust standard errors (clustered by dyad) are in parentheses. The log-likelihood ratio test compares each model to a naïve model that excludes coup risk (i.e., the analogous specification from models 1–4).
***p < .01. **p < .05. *p < .1 (two-tailed).
The coup risk measure is significant at the .01 level with the expected positive sign for all four specifications, indicating that the impact of coup risk on interstate conflict is consistently additive across conflict levels. Further, a likelihood ratio test reveals that each specification represents a significant improvement to the fit of the model when compared to a naïve model that omits the coup risk measure. Moving from the 20th to 80th percentile of coup risk increases MID likelihood by 72 percent, the likelihood of a war-level MID by 150 percent, the likelihood of international crisis initiation by 106 percent, and the likelihood of a VIC by 252 percent. In addition to providing strong support for the diversionary theory, the findings suggest that the diversionary incentive is not constrained by the level of force and is actually substantively stronger at higher conflict levels. The suggestion of Morgan and Bickers (1992) that unpopular leaders should avoid higher levels of conflict is not born out, as diversion is seen across multiple levels of conflict and multiple data sources.
A consideration of the coup proofing apparatus suggested a number of important expectations. The first hypothesis predicted that the diversionary impact of coup will subside as structural coup proofing is strengthened. This is due to two mechanisms. First, leaders are expected to lose the incentive to utilize conflict when alternate means of survival are already employed. Second, structural coup proofing will hamper military capabilities and consequently reduce the feasibility of diversion. The marginal influences of coup risk and structural coup proofing on conflict are tested in models 5 through 10 in Table 1. For presentation purposes, I present models utilizing MIDs and VICs as the dependent variable for each coup-proofing measure. 6
Ai and Norton (2003) warn that interpreting continuous interaction coefficients from a logit or probit is misleading, as an interaction might be significant even if not indicated by a t-test. Following this warning and the more recent suggestion of Brambor, Clark, and Golder (2006), multiplicative effects are not interpreted from the models in Table 1. I instead rely on the Grinter utility developed by Boehmke (2006) to graphically illustrate the marginal influences of coup risk and coup-proofing on conflict initiation. A graphic representation of the models is available in Figure 1, with each model noted in the upper-right corner of the Grinter output (e.g., M5 refers to model 5). The left column of Figure 1 presents the results for MIDs while the right column presents the results for VIC initiation.
The marginal influence of coup risk and coup proofing on conflict initiation militarized interstate dispute violent international crisis.
The vertical axis of each output represents the coefficient for the effect of coup risk on conflict initiation. Moving along the horizontal axis shows how the influence of coup risk changes in accordance with each respective value of coup-proofing. Each specification (M5–M8) displays a positive association between risk and conflict in states with low levels of structural coup-proofing, such as ranking less than two for effective ground-capable organizations or less than four (50th percentile) on the counterbalancing scale. The influence of coup risk declines as each measure increases, until the relationship is insignificant at the upper extremes of coup proofing. Models 5 and 8 indicate that coup risk actually becomes negative and significant at the upper extremes, lending support to the idea of conflict aversion. In cases of extreme coup-proofing, leaders appear to have crippled the fighting capacity of the state to the point that diversion is not only unnecessary but perhaps even dangerous. Substantively, when looking at high risk states (90th percentile of coup risk), moving from the 20th to the 80th percentile in effective organizations reduces the likelihood of MID initiation by over 80 percent (.011 to .0032). A similar change in counterbalancing reduces MID initiation by 49 percent (.022 to .011).
Table 1 also reports the results of two likelihood ratio tests. The first compares the fit of the presented models to a base model that omits measures for coup risk and coup proofing. This test shows that the inclusion of the coup risk, coup proofing, and interactive terms significantly improves model fit in models 5 through 10. The second likelihood ratio test compares the fit between the interactive models (5–10) and the analogous models (1–4) that exclude the coup proofing measure. The test indicates that with one exception (model 6), including coup proofing and the interaction significantly improves model fit even over the coup risk models. These trends provide strong support for the first hypothesis.
The results for the spoils measure are included in Figure 1 for comparison. Here, we see results that are inconsistent when distinguishing MID and VIC models. For MIDs, spoils appears to have a similar impact as structural coup proofing (model 9), while higher levels of spoils make coup risk additive to the initiation of VICs (model 10). A more careful evaluation of the data clarifies the discrepancy. The second hypothesis predicted that the impact of military spoils will be dependent on regime type. Specifically, the transparent nature of democracies will lead military expenditures to be a more accurate reflection of military power. Democracies with high military expenditures may lose the incentive to utilize diversion, but should be particularly capable of doing so. In the absence of strong accountability measures, authoritarian regimes will have expenditures more likely to be dedicated to officer allowances or symbolic spending, investments that while coup inhibiting have little impact on improving fighting capacity. Highly funded authoritarian armies, then, are not expected to be as well equipped, trained, or effective as their democratic counterparts and are expected to be less likely to be used for diversion. Models 11 through 18 distinguish between regime type and can be found in Table 2. Figure 2 graphically illustrates the relationship.
Marginal influence of coup risk and coup proofing on conflict initiation, by regime democracies and nondemocracies.
Marginal Influence of Coup Risk and Coup-Proofing on Conflict Initiation, by Regime Type.
Note. LR = log-likelihood ratio; MIDs = militarized interstate disputes; VIC = violent international crisis. Models reflect rare events logistic regression. Robust standard errors (clustered by dyad) are in parentheses. The log-likelihood ratio test compares each model to a naïve model that excludes coup risk and the coup-proofing measures.
***p < .01. **p < .05. *p < .1 (two-tailed).
The left column of Figure 2 displays the marginal influence of coup risk on MID initiation for democracies, while the right column considers nondemocracies. 7 I begin the figure by splitting the population for the structural coup proofing models. Military funding is expected to be contingent on regime type due to differences in its use, but the nature of structural coup proofing is expected to influence different regime types in the same manner. The top two rows illustrate the influence of effective organizations and counterbalancing, respectively. A similar slope for the coefficient is quickly apparent for all four models. The influence of coup risk on MID initiation subsides as structural coup proofing increases, though the split samples indicate a modest difference in behavior when considering the impact of effective organizations. Democracies (model 11) reveal a similar trend as the full model, with coup risk acting as a positive influence on MID initiation at low ends of the measure, while at the higher ends we see conflict avoidance. Nondemocracies (model 12) only display the conflict avoidance trend. Counterbalancing, meanwhile, displays similar slopes for both democracies and autocracies (models 13, 14) and continues to provide strong support for the first hypothesis.
The bottom two rows of the figure illustrate the effect of spoils on MID and VIC initiation. The split samples reveal important differences in state behavior. Coup risk has no influence on conflict initiation (MID or VIC) at lower levels of spoils in democracies (M15, M17), but gains significance with an increasingly positive effect on conflict behavior at higher levels. Increasing spoils from the 20th to 80th percentile increases MID initiation by 706 percent in democracies (.0014 to .015).
Autocracies, however, display a stark contrast. High-risk states are more likely to use poorly funded militaries to initiate international conflict. This trend, as expected, subsides as spoils increase (model 16). A 20th to 80th percentile change reduces MID initiation by 63 percent in high-risk autocracies (.021 to .008). The substantive influence of coup proofing and high coup risk can be found in Figure 3, which distinguishes between regime type for each mode of coup proofing. These results provide substantial support for Hypotheses 2a and 2b. Democratic militaries are more likely to be used for diversion when they are endowed with significant resources, while authoritarian armies will only be used when they are being given modest endowments. These findings indicate the fungible nature of military expenditures and point to a more general implication for the study of conflict. Specifically, the results of this analysis provide an important addition to the story of democratic military effectiveness. A number of previous studies have pointed to the military superiority of democracies (e.g., Reiter and Stam 1998), and others have pointed to the negative implications of coup-proofing for military effectiveness during hostilities (Pilster and Böhmelt 2011). This analysis builds on these seminal offerings by illustrating the importance of coup proofing for conflict initiation.
Marginal influence of high coup risk on MID initiation. Note. Substantive effects in the figure are calculated with CLARIFY (Tomz, Wittenberg, and King 2003). Effective organizations data are derived from models 5, 11, and 12. Counterbalancing data are from models 7, 13, and 14. Spoils data are from models 15 and 16.
The control variables behaved as expected, though the economic measures (wealth and growth) performed poorly. This is unsurprising, given the inclusion of the coup risk measure. Economic indicators would seem to be important for diversionary studies only to the point that they will serve as proxies for the likelihood of losing office. The inclusion of the coup risk variable is a more direct measure for survival prospects and performs considerably better. Joint democracy was negatively associated (79 percent reduction in model 1) with conflict initiation in each specification except for model 16, which looks at violent international crises and is temporally limited to the 1967–1986 time frame. The power variable shows that conflict initiation was more likely when the initiator was the stronger member of the dyad. Moving from the 20th to 80th percentile in power to ratio increased conflict initiation by 281 percent in model 1. Allies showed a more modest 18 percent reduction in conflict likelihood than nonallied dyads. The most powerful indicators of conflict, unsurprisingly, are distance and peace years. Increasing from the 20th to 80th percentile in distance (contiguous to 6,000 miles) reduced conflict likelihood by 98 percent. Increasing years since the last MID from four to twenty-seven years brought a 99.9 percent reduction.
Conclusion
Though there is a rich tradition of diversionary literature, few efforts have attempted to identify the conditions that would see an unelected leader have the need to implement diversionary tactics. Even fewer have considered the impact that coup-proofing can have on the necessity and ability for diversion. This article sought to move beyond an election-focused approach to leadership turnover by providing a proper theoretical treatment to the threat of a coup—the most common form of authoritarian removal (Svolik 2009). The analyses presented here indicate that interstate belligerency can be attributed to the conditional influences of coup risk and coup proofing. Specifically, leaders in coup-proofed regimes will be less likely to use diversion than leaders who lack a strong coup-proofing apparatus.
In addition to the general findings regarding coup risk and coup proofing, the analyses point to disparate findings regarding the diversionary activities of democracies and autocracies. These results suggest that authoritarian leaders lose both the will and the ability to utilize diversion when providing more resources to the armed forces. In short, such funds are unlikely to go toward improving the fighting capacity of the rank and file. Already treated as inferior to well-funded and well-armed coup-proofing units, members of coup-proofed regular armies lack the ability to fight in the international arena but could still pose a threat to domestic security if alienated. Sarr (2007, 40–41), for example, has noted that the motivation for the Gambia National Army’s 1994 ouster of President Dawda Jawara included grievances surrounding an intervention in Liberia. In addition to lacking both training and equipment, the government refused to bring home the bodies of fallen Gambian soldiers. These soldiers were then buried at the battlefield, with their families receiving a conciliatory sum of less than $50. Kisangani and Pickering (2007) have noted the tendency to utilize such interventions for diversionary purposes, though the Gambian case illustrates the potential fallout for regimes that do so with limited capabilities. Sarr (2007) suggests that the government never gave consideration to the consequences of soldiers being killed, but this analysis suggests other coup-fearing leaders are more cautious with their decisions.
Meanwhile, democratic rulers have heightened ability under the same conditions and readily use these capabilities. Despite the incentive to use such tactics, democratic leaders do not seek diversion in the absence of a well-funded military. Such a trend supports prior literature suggesting democracies will be risk averse when choosing to enter a conflict (e.g., Bueno de Mesquita and Siverson 1995; Reiter and Stam 1998).
A few words of caution should be made in regard to this analysis. First and foremost, this work does not suggest diversionary conflict will be sought in order to accomplish a rally, round the flag effect among the general population. Previous research on the rallying strength of external aggression is at best mixed (Gartner and Segura 1998, 2000), and this analysis does not offer direct support for the “rally” theory. Nor does this article suggests that diversion is attempted in order generate divisions within military, as described by Belkin and Schofer (2005). Instead, I argue that diversion is utilized in an effort to distract elites that can credibly undermine a regime by providing them with a foreign focus. Civil–military relations scholarship has long noted that civilian control can be strengthened when the armed forces are provided with an external mission (Huntington 1991; Desch 1999). For those that fear a coup and lack alternative means to preserve their rule, such a strategy seems to be particularly attractive.
Footnotes
Acknowledgments
I thank Dan Morey, Clayton Thyne, Sara McLaughlin Mitchell, Phil Arena, Paul Huth, and the anonymous reviewers at the Journal of Conflict Resolution for their comments on this project. Earlier versions of the article were presented at the 2009 American Political Science Association Annual Meeting and the 2009 Annual Conference of the International Studies Association—Southern Region. Replication materials and the appendix can be found at 
.
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) received no financial support for the research, authorship, and/or publication of this article.