Are Military Regimes Really Belligerent?

Abstract

Does military rule make a state more belligerent internationally? Several studies have recently established that military autocracies are more likely than civilian autocracies to deploy and use military force in pursuit of foreign policy objectives. I argue that military regimes are more likely to resort to military force because they are located in more hostile security environments, and not because they are inherently aggressive. First, I show that rule by military institution is more likely to emerge and exist in states facing external territorial threats. Second, by examining the relationship between military autocracies and conflict initiation, I find that once I control for states’ territorial threats, the statistical association between military regimes and conflict initiation disappears. Additionally, more evidence suggests that civilian dictatorships are more conflict-prone than their military counterparts when I account for unobserved dyad heterogeneity. The results are consistent across different measures of international conflict and authoritarian regimes.

Keywords

interstate conflict militarized interstate disputes domestic politics political leadership

Since Geddes’s (2003) notable statement that authoritarian regimes differ from each other as much as they differ from democracies, a growing body of literature has paid attention to the institutional heterogeneity among autocracies to explain various outcomes.¹ An increasing number of studies on international conflict have examined how dictatorships differ from each other in their propensity to engage in belligerent international behavior. Particularly, three seminal studies (Lai and Slater 2006; Debs and Goemans 2010; Weeks 2012) have recently established that military autocracies are more likely than civilian autocracies to deploy and use military force in pursuit of foreign policy objectives. These studies attribute military regimes’ relative conflict proneness to various sources: their lack of institutional power leading to regime’s insecurity (Lai and Slater 2006), to the harsh, postexit punishments military rulers face (Debs and Goemans 2010), or to ruling elites’ military backgrounds (Weeks 2012, 2014). This research is indicative of a growing scholarly interest in examining the linkage between domestic politics and international affairs.

While recognizing the contributions of these studies, I argue that the “military belligerence hypothesis” must be subjected to further critical scrutiny. These previous studies pay little attention to the fact that political regimes are not randomly assigned across countries and over time. Drawing on the peace-to-democracy and territorial peace literatures (Gibler 2012; Hintze 1975; Rasler and Thompson 2004; Thompson 1996), I argue that military regimes are more likely to emerge and exist in states facing sustained territorial threats. Salient territorial threats produce high levels of militarization, which expands and politically empowers the military. Thus, the military’s capacity to intervene in politics increases when the country is exposed to salient external threats. If this is the case, territorial threats from neighboring countries may be causally responsible for generating both military regimes and militaristic behavior. That is, military regimes may be more likely to resort to military force or threat of military force because they are located in more hostile security environments, not because they are either institutionally fragile or predisposed toward using force. If as the previous research argues, military autocracies are indeed more prone to militarized conflict than civilian autocracies due to their inherent characteristics, a systematic relationship between military regimes and conflict initiation should be found even after controlling for countries’ external territorial threats.

My empirical analysis consists of two parts. First, I show that rule by military institutions is more likely to emerge and persist when countries face territorial threats from neighboring rival states.² The same is not true of other authoritarian regime types. Second, I test the relationship between political regime types and the initiation of militarized disputes. I find some evidence that military dictatorship, including both collegial and personalist military rule, regimes—are more likely than civilian party–based dictatorships to initiate militarized disputes. However, controlling for territorial rivalries removes the statistical association between military regimes and conflict propensity. Additionally, more evidence suggests that civilian dictatorships are more conflict-prone than their military counterparts when I account for unobserved dyad heterogeneity. The lack of a significant association between military autocracies and conflict initiation remains consistent when (1) using either dyadic or monadic specifications, (2) using militarized disputes from the militarized interstate disputes (MIDs) data, MIDs that feature the use of force, or international crises drawn from the international crisis behavior (ICB) data, (3) comparing military and civilian dictatorships with or without accounting for personalism, and (4) addressing unobserved dyad effects through random effects or fixed effects.

When I further distinguish between territorial and nonterritorial militarized disputes and do not control for territorial rivalries, military regimes’ aggressiveness is found only in territorial disputes. In the context of nonterritorial disputes, no evidence supports military regimes’ aggressiveness. The result provides additional evidence that territorial threats from neighboring countries are likely to produce both military autocracies and increased conflict propensity.

Overall, I find no compelling evidence that military rule increases a state’s propensity to initiate military conflict, compared to civilian rule. It appears premature to conclude that certain characteristics encourage military dictatorships to engage in foreign aggression. Instead, my empirical analysis consistently demonstrates that civilian personalist regimes are the most belligerent of all, and monarchy is the most peaceful. The results suggest that consistent with the previous studies (Weeks 2008, 2012), variations in domestic institutional constraints are important to explaining regimes’ conflict propensity.

Military Regimes and Conflict Initiation

Many studies on authoritarian regimes identify military dictatorships as a distinct subtype of authoritarianism. Military dictatorships behave differently than nonmilitary dictatorships and are systematically associated with a range of important political outcomes. The distinctive features of military dictatorships can be summarized as follows.³ First, military dictatorships are governed by those who specialize in the use of force. Military regimes’ greater capacity for violence makes military dictatorships better-equipped to use repression in response to popular dissent (Escriba-Folch and Wright 2010). Additionally, military rulers prefer to maintain the military’s internal unity to protect its corporate interests (Geddes 2003; Stepan 1971). However, military regimes’ advantages in coercive capacity and internal coherence do not lend them stability or durability. Instead, both military dictatorships and their leaders have the shortest life spans (Geddes 2003; Gandhi 2008). Military dictators frequently face violent ousting by other officers, followed by severe posttenure punishments of imprisonment and death (Debs and Goemans 2010). This fragility of military regimes could be either due to the nature of the military as an institution that emphasizes unity, as is argued by Geddes (2003), or to the absence of institutional infrastructure (Slater 2003). Military officers do not tend to retain their rule when faced with popular protests or economic crises because they value the unity of the military over political power (Geddes 2003). At the same time, military regimes are less institutionalized (Escriba-Folch and Wright 2010; Nordlinger 1977; Slater 2003). Military regimes lack the mechanisms of sociopolitical control because they tend to rely on terror and repression as a means of rule. Repression alone is not sufficient to hold a regime together. Military regimes may thus have difficulty surviving during hard times.

Existing Studies

Scholarly attention to the different attributes of military dictatorships has probed the relationship between military regimes and conflict propensity. These studies produce considerable disagreement about the mechanisms that cause military dictatorships’ belligerence. First, Lai and Slater (2006) focus on military dictatorships’ institutional deficiencies. According to Lai and Slater, the institutional power affecting a regime’s legitimacy and security relies on whether it is ultimately backed by the military or by a ruling party. Military dictatorships tend to lack institutional infrastructure for maintaining social control and elite cohesion. They rely on military apparatuses to maintain political control as they lack party infrastructure to enhance the regime’s stability and durability. Thus, military rulers are less secure in power and more likely to initiate militarized conflict to bolster its regime by mobilizing domestic support.

Meanwhile, Lai and Slater are skeptical that the other institutional dimension involving constraints on leaders’ decision-making power effectively explains conflict propensity. Instead, the only significant quality is whether the regime is led by civilians or the military. This implies that in their four-way classification of autocratic regimes, personalist (Strongman) and nonpersonalist military (Junt a) regimes tend to initiate more disputes than personalist (Boss) and nonpersonalist civilian regimes (Machine). Finally, Lai and Slater argue that military regimes’ belligerence is unrelated to the military background of the leaders because civilian rule controlled by military elites is just as conflict prone as direct military rule.

Challenging Lai and Slater’s argument, Weeks (2012, 2014) claims that their background and training make military officers more likely than civilians to view the status quo as threatening and to consider the use of force necessary and effective (see also Sechser 2004).⁴ These preexisting views result in military officers’ proclivity for using military force. Weeks also disagrees with Lai and Slater’s argument that diversionary incentives explain military regimes’ aggression. In Weeks’s view, domestic plights leading to diversionary conflicts do not arise often enough to drive up a regime’s conflict propensity in general. Thus, a regime’s institutional control over a society does not matter. The last point of disagreement between Weeks and Lai and Slater is that in explaining conflict propensity, Weeks emphasizes the extent to which members of the ruling group can impose limitations on rulers. When leaders face a domestic audience able to punish them for foreign policy mistakes, they are more cautious about using force. Conversely, unconstrained dictators are more willing to take risks and more emboldened to embark on aggression in interstate disputes. However, Weeks argues that a leader’s background is largely redundant in personalist regimes that tend to select for highly violent and ambitious leaders. Accordingly, she orders autocratic regimes from most to least belligerent: Strongman, Boss, Junta, and Machine, and she expects the difference between Strongman and Boss to be marginal.

On a different note, Debs and Goemans (2010) explain the war propensity of different regime types by focusing on both leaders’ sensitivity to war outcomes and their postexit fates. As leaders’ survival is more sensitive to war outcomes, and the cost of losing power is greater, they are less willing to make concessions to other states. Debs and Goemans find no significant difference in the sensitivity to war outcomes among dictatorships, although dictatorships are more sensitive to war outcomes than democracies. They instead find that military dictators and monarchs tend to face worse fates, such as death or imprisonment, after losing power than do civilian dictators. The fear of postouster fates looms in dictators’ minds even when the likelihood of losing office is low. Dictators, fearing severe posttenure punishments, cling more desperately to power and are less likely to make peaceful bargains with other states. Hence, military dictators and monarchs are more likely than civilian dictators and democratic leaders to be involved in war. However, they do not claim that military dictators are more likely to initiate war.

Overall, these studies claim that military autocracies are more conflict-prone than civilian autocracies. However, they produce conflicting theories and incongruous empirical results (see also Table 1). In addition, these studies use different regime-type data sets that are built on different definitions of military regimes. We thus know neither which mechanism is responsible for the observed pattern nor whether the previous findings are robust. However, since it is beyond the scope of this article to address these issues, I focus on one challenge to the military belligerence theories: a potentially spurious relationship between military autocracies and increased conflict propensity.

Table 1.

Summary of the Previous Studies on Military Dictatorship and Conflict Propensity.

Regime Classification	Regime Data	DV	Predicted Belligerence
Lai and Slater (2006):
Strongman, Boss, Junta, and Machine	Banks and Polity III	MID initiation	Strongman and Juna > Machine and Boss
Weeks (2012, 2014):
Strongman, Boss, Junta, and Machine	Raw regime data from Geddes (2003)	MID initiation	Strongman > Boss > Junta ≫ Machine
Debs and Goemans (2010): Civilian, Military, Monarchy	Cheibub et al. (2010)	War onset	Military and Monarchy > Civilian

Note: DV = dependent variable; MID = militarized interstate dispute.

The Argument

Previous studies attribute military regimes’ conflict proneness to military regimes’ institutional characteristics or to military leaders’ personal characteristics. However, they do not consider the possibility that other factors may be responsible for generating both military rule and heightened conflict propensities. Building on the literature that emphasizes a state’s security environment in explaining democratization, I argue that military rules, particularly characterized by collegial forms, are more likely to emerge in states with external territorial threats. When faced with salient external threats, states tend to engage in more aggressive policies. Accordingly, military regimes will be more likely to initiate international conflicts because they are located in more hostile security environments, not because they are either institutionally fragile or predisposed toward using force.

External Threat Environment → Political Regime Types

The prominent so-called war-making and state-making literature emphasizes the role wars and external threats play in state centralization and development, analyzing the interrelationship between wars, the military, and state building (Hintze 1975; Tilly 1985). States facing wars and external threats mobilize resources to build and maintain large standing armies, which in turn require a highly centralized state to raise and administer revenues and expenditures. Building on this literature, several scholars argue that a country’s hostile security environment fosters authoritarianism and undermines democratic rule (Gibler 2012; Hintze 1975; Rasler and Thompson 2004; Thompson 1996). A state’s centralization and militarization in response to external threats interact to undermine constraints on executive control and to suppress domestic opposition. Thus, external threats retard the development of democratic rule while the absence of these threats improves the prospects for democratization.

Drawing on the peace-to-democracy literature, Gibler (2012) develops the “territorial peace” theory that when states have peacefully settled borders, they no longer rely on military force to resolve disputes. Gibler posits that contested borders generate more salient and lasting external threats than any other factors. Contested borders encourage government centralization and militarization, which generates more aggressive foreign policies and worsen security environments (see also Vasquez 2009). Meanwhile, the presence of large standing armies, necessitated by territorial disputes, reduces the costs of domestic repression and empowers the military and elites. Accordingly, unsettled borders not only increase security threats to the state but also hinder democratization. Contrarily, when a state peacefully resolves border disputes with its neighbors, a hospitable environment emerges, reducing the need for large standing armies and decentralizing political power. In sum, settled borders between two countries improve both interstate relations and the prospects for joint democracy within the dyad.

Territorial Threats and Military Regimes

The military, created to protect against foreign and domestic enemies, is at the heart of the existing theories on external threats and domestic politics. High levels of external threat pressure states to anticipate violent challenges and, in response, develop sufficient defenses. “Even if a state could avoid the temptation to expand, being in a neighborhood harboring some expansive aspirants meant that one had to develop adequate defenses against the possibility of attack” (Rasler and Thompson 2004, 882). Hence, rulers develop large, standing land-based armies in anticipation of such external threats (Huth 1996; Rasler and Thompson 2004). External threats also allow rulers to better extract the resources necessary for militarization at the expense of other sectors (Gibler 2012; Thies 2005; Tilly 1985), leading to the expansion of a coercive military organization.

Existing research on interstate conflict demonstrates that a state perceives greater military threats particularly when external threats emanate from its immediate neighborhood and are mainly concerned with the possession of territory (e.g., Gibler 2012; Rider 2013; Vasquez 2009). Rivals close to home pose more substantial threats due to simple proximity. Moreover, because people tend to have strong attachments to their homeland for material and/or symbolic reasons and, thus, are willing to fight to defend it, states are likely to engage in provocative and violent behavior in order to protect or acquire that territory. States are highly attentive to the possibility of violent transfers of territory and actively develop war plans based on acquiring or retaining territorial control (Gibler and Tir 2010, 954). Therefore, rivalries with neighboring states over territories are more intense and have stronger repercussions on domestic politics and political institutions than threats from other actors.

High levels of military preparations in turn place the military in a politically pivotal role (Hintze 1975; Gibler 2012; Lasswell 1941). When a country is confronted with external threats to its security, its military is better positioned to demand and obtain greater institutional autonomy in personnel, education, budgetary, organization, and procurement decisions. Rulers delegate extensive power to the military in order to defend against external threats. Furthermore, the need for military effectiveness tends to increase unity and cohesion within the military (Desch 1998), making it better able to intervene in politics (Belkin and Schofer 2003). Hence, a military equipped with greater resources, autonomy, and cohesion is better able to expand its political influence. As Svolik (2013) puts it, “Only once such preeminence translates into the military’s ability to garner greater autonomy and resources is the military in a position to intervene in politics should its political preferences or institutional interests be undermined” (769).

Rulers face a fundamental dilemma in that any military strong enough to defend a regime against external threats is also strong enough to subvert that regime. Additionally, salient threat environments discourage political leaders from weakening the military’s political power since the tactics employed to prevent the military from seizing power simultaneously erode the state’s military effectiveness while decreasing the risk of coups (e.g., Pilster and Böhmelt 2011). For example, promoting and appointing officers based on loyalty and ethnic affiliation diminish leadership qualities and discourage the exercise of initiative (Pilster and Böhmelt 2011, 335; Brooks 2003, 162). Similarly, counterbalancing impedes coordination among different military units, which is critical to the implementation of modern system tactics and operations (Pilster and Böhmelt 2011, 335). Once the military obtains its privileged position under sustained external threats, therefore, the military’s capacity to intervene in politics is hard to curb.

Observable Implications

The discussion above suggests that military rule is more likely to emerge in states facing sustained territorial threats to their homelands. Such sustained threats expand and empower the military as an institution, paving the way for military rule. Particularly given that sustained threats expand and empower the military as an institution, rule by the military as an institution is more likely to emerge in hostile security environments. At the same time, states tend to rely on coercive tactics (such as arming, military mobilization, and seeking alliances) to address territorial disputes rather than disputes over other issues (Vasquez 2009). Numerous studies show that territorial disputes are more prone to violent conflict than disputes over other issues (Hensel 1996; Huth 1996), produce higher fatalities (Senese 1996), are more escalatory (Hensel 1996), and are more likely to persist (Hensel 2001).

Taken together, this suggests that military regimes, particularly collegial military regimes characterized by “rule by the military as an institution,” may be more likely to initiate international conflicts because they are often located in hostile security environments with a high likelihood of militarized conflict initiation. If collegial military regimes are indeed more prone than collegial civilian autocracies to militarized conflict due to their own characteristics rather than to territorial threats from neighboring countries, I should be able to identify a systematic relationship between military regimes and conflict initiation even after controlling for external threats to a countries’ homeland. This should hold true, because the sequential relationship, territorial threats → military rule → conflict, is possible.

Below, I first establish that sustained territorial threats to states increase the likelihood that collegial forms of military rule emerge and persist. Next, I test whether military autocracies are more likely to initiate militarized conflict than civilian autocracies even when controlling for states’ sustained territorial threats.

Testing Relationship between External Territorial Threats and Military Regimes

The dependent variable is the emergence and incidence of collegial military regimes (denoted as Junta). The incidence of Junta is an indicator equaling one in years of ongoing Junta, and the onset of Junta is an indicator equaling one in the year a new Junta emerges. To measure Junta, I use the regime-type data constructed by Geddes, Wright, and Frantz (hereafter GWF, 2014) since GWF emphasize rule by military institutions in defining military regimes. They define military regimes as those in which “the dictator consults with other high-ranking officers and can be constrained by them” (p. 152). Military dictatorships in Argentina 1955–1982, Brazil 1964–1984, and Uruguay 1973–1983 belong to this category.

The GWF data set classifies autocracies as military regimes, dominant-party dictatorships, personalist regimes, hybrids of these three pure types, and monarchies. To distinguish among dictatorships, GWF focus on “whether control over policy, leadership selection, and the security apparatus is in the hands of a ruling party (dominant-party dictatorships), a royal family (monarchies), the military (rule by the military institution), or a narrower group centered around an individual dictator (personalist dictatorships)” (p. 318). For example, they code a regime as a military regime if the proportion of questions regarding military rule answered by “yes” is high and the proportion of questions regarding personalist and party rule answered by yes is low. A regime with high scores in multiple categories is coded as a hybrid regime.

To fully utilize the information on military regimes from GWF regime data, I construct a measure of collegial military regimes aggregating all military hybrids, including “party-military” and “party-personal-military” hybrids. This coding rule is slightly different from what GWF suggest. For their analysis, GWF include all party hybrids and oligarchies in the category of dominant-party dictatorships by prioritizing a party dimension, grouping only “personal-military” with military regimes and classifying pure “personal” as personalist autocracies. Grouping party-military and party-personal-military hybrids with party-based regimes is not appropriate for my research since military belligerence theories indicate that these hybrids should not be less aggressive than pure party-based dictatorships. For instance, Honduras 1964–1971, El Salvador 1950–1982, and Congo 1969–1991 are coded as party-military hybrids, and Paraguay 1955–1993, Egypt 1953–2008, and Indonesia 1967–1999 are coded as party-personal-military hybrids. In all these countries, the military exerts effective control on important policies and key positions of power. These countries should behave differently than countries coded as purely party-based dictatorship, such as Cambodia 1975–2010, Hungary 1947–1990, and Zambia 1968–1991.

A key independent variable is salient and prolonged threats to a state’s territories. To measure this variable, I focus on interstate rivalries. Interstate rivalry involves a pair of states that regard each other as competitive, threatening enemies in protracted conflict (Colaresi, Rasler, and Thompson 2008; Klein, Goertz, and Diehl 2006). Rivalries, often characterized by mutual threat perception and intense hostility, are the context in which the vast majority of interstate conflicts occur. Militarized foreign policies are prevalent in a rivalry context. Several studies use interstate rivalries to capture a country’s external threats (e.g., Gibler 2012; Rasler and Thompson 2005; Thies 2005). For the measure of rivalries, I rely on two widely used data sets: Klein, Goertz, and Diehl (2006) and Colaresi, Rasler, and Thompson (2008). Klein et al. emphasize the occurrence of militarized disputes in conceptualizing rivalry relationships and define a rivalry as a dyadic relationship in which two states engage in militarized disputes at least three times over the same set of issues. Conversely, the Colaresi et al.’s data employ a perception-based approach to identify strategic rivalry. They focus on leaders’ perception, by evaluating leader statements and historical narratives, rather than on actual dispute participation. I employ both measures because I expect that both participation in repeated militarized disputes and perceived military threats affect the need for security and military buildups. For the measure of strategic rivalry, I use the Thompson and Dreyer (2011) data set that updates the Colaresi et al. data set and covers the time period from 1816 to 2010.

To explore the effect of territorial threats, I distinguish rivalries competing over territorial issues or sharing land borders (what I call territorial rivalries) from rivalries competing over other issues or not sharing land borders. To this end, when I employ strategic rivalry, I utilize spatial rivalries as coded in Colaresi, Rasler, and Thompson (2008). Colaresi et al. distinguish between spatial rivalries primarily concerned with territorial issues and positional rivalries concerned with power position. I measure Territorial rivalry (strategic) as a dichotomous variable that takes the value one when a country is involved in at least one spatial rivalry in the prior year, and zero otherwise. Nonterritorial rivalry (strategic) is an indicator for countries that experience strategic rivalry but no spatial rivalry. For the Klein et al. rivalry measure, I create a binary variable Territorial rivalry (KGD) that is coded one when a rivalry’s most frequent reason for militarized disputes is territory.⁵ Nonterritorial rivalry (KGD) includes the remaining rivalries.

I also include control variables. First, I include an indicator of civil war taken from the Correlates of War (COW) data (Sarkees and Schafer 2000) to capture the possibility that internal conflicts encourage the military to take on a more active political role (Desch 1998). A binary indicator of internal armed conflicts is coded as one for country-years with at least one corresponding internal conflict occurring in the previous year, and zero otherwise. Second, I control for regime type by including dichotomous indicators for democracies and anocracies. Regimes whose Polity score is higher than 5 in the previous year are classified as democracies while those whose Polity score is between −5 and 5 in the previous year are classified as anocracies (Marshall, Gurr, and Jaggers 2013). Next, I include a natural log of real GDP per capita and the annual percentage change of real gross domestic product (GDP) per capita.⁶ Fourth, global and regional environments may influence the establishment of military regimes. This is captured by a dummy variable for the post–Cold War period and the proportion of democratic neighbors.⁷ Last, I include a natural log of the amount of the time elapsed between the last regime change and the military regime’s emergence to control for potential negative duration dependence.

Figure 1 displays the estimated coefficients of territorial and nonterritorial rivalries along with their standard errors from logit regressions. The full regression tables are reported in Table A1 of the Online Appendix. Models 1 and 2 examine regime onset,⁸ and models 3 and 4 analyze regime incidence.⁹

Figure 1.

Association between territorial rivalries and different types of autocracies. The graph shows the logit regression coefficients from separately estimated models. Circles show the point estimates, and horizontal line segments associated with circles show the 95 percent confidence intervals. All models include full control variables.

Regardless of whether I examine regime incidence or onset, the left-most panel of Figure 1 shows that the coefficients on Territorial rivalry (KGD), built on the dispute-density approach, and Territorial rivalry (Strategic), based on the perception-based approach, are positive and statistically significant at the 5 percent level, which comports with my argument.¹⁰ This suggests that Junta is more likely to emerge when a country has engaged in a series of militarized disputes fought over territory or has long-standing competition with rivals over territorial controls. Substantively, collegial military regimes are about three (when using Territorial rivalry [KGD]) or twice (when using Territorial rivalry [Strategic]) more likely to emerge when a country has territorial rivalries than when it has no rivalries. On the other hand, Nonterritorial rivalry (KGD) and Nonterritorial rivalry (Strategic) are not significantly associated with Junta, which confirms the importance of territorial threats.

For comparison, I examine the impact of territorial and nonterritorial rivalries on other authoritarian regime types (see the next section for how to measure them). Figure 1 suggests no systematic relationship between territorial rivalries and other authoritarian regimes.

Testing the Relationship between Military Regimes and Conflict Initiation

Next, I test whether military dictatorships are more likely to initiate militarized conflict than civilian dictatorships when I control for a country’s territorial threats increasing the probability of dispute behavior. Previous studies characterize military belligerence as a monadic effect of military autocracies, operating independently of both domestic political conditions in other states and interactions with other states. However, they use different empirical strategies: Lai and Slater (2006) and Debs and Goemans (2010) use monadic tests in which country-years are the unit of analysis while Weeks (2012, 2014) employs dyadic tests in which directed dyad-years are the unit of analysis. To ensure robustness, I use both monadic and dyadic specifications.

To code conflict initiation, I use the COW MID data set, and for the purposes of comparison, focus on “side A,” the state that initiated the first militarized move, because Lai and Slater (2006) and Weeks (2012, 2014) use the same dependent variable. As Ghosn and Bremer (2004, 38-39) note, however, the state on side A is not necessarily responsible for the conflict. “If a country perceives a potential threat, it may choose to attack first, and it is not clear that data focusing on the direction of attack are always able to account for such preemptive strikes” (Caselli, Morelli, and Rohner 2015, 287). Thus, I also examine MID initiation in terms of “revisionist” that sought to revise the status quo by force. In monadic models, the dependent variable is a count of a country’s total number of MID initiations in year t + 1. In dyadic tests, the dependent variable is a dummy variable coded one if state A initiated a new MID against state B in the directed dyad in year t + 1, and zero otherwise. The time period for the empirical analysis is 1946 to 2000.

A potential concern is that the MID data set includes many minor disputes, not explicitly authorized by state leaders, and noninterstate conflict cases (Downes and Sechser 2012, 463-64). The inclusion of these cases may be problematic since the theories under examination explicitly focus on political leaders’ choices to engage in militarized disputes. To address this concern, I also limit MIDs only to those in which force is used.¹¹ Additionally, I employ the initiation of international crises as coded in the ICB project. The ICB project specifies two defining conditions for an international crisis: “(1) a change in type and/or an increase in intensity of disruptive, that is, hostile verbal or physical, interactions between two or more states, with a heightened probability of military hostilities; that, in turn, (2) destabilizes their relationship and challenges the structure of an international system—global, dominant, or subsystem” (Brecher and Wilkenfeld 1997, 4-5). The ICB data set is attractive in that it excludes conflicts resulting from unauthorized or “accidental” uses of force.¹²

Following previous studies (Lai and Slater 2006; Weeks 2012, 2014), I differentiate between authoritarian regimes using personalism dimension as well as military–civilian dimension: personalist military (Strongman), personalist civilian (Boss), nonpersonalist military (Junta), and nonpersonalist civilian regimes (Machine). To measure autocratic regimes, I again use the GWF regime-type data. As explained above, GWF military regimes measure the rule by the military, corresponding to Junta. To measure Strongman, I follow Geddes, Frantz, and Wright’s (2014) recommendation: a country-year qualifies as military strongman rule when it is coded both as personalist by GWF and as military by Cheibub, Gandhi, and Vreeland (2010). The Cheibub et al. classification of dictatorships depends solely on the identity of the regime leader without considering institutional configurations and the composition of the political leadership. It codes a dictatorship as a military dictatorship if the effective leader is or was a current or former military officer prior to seizing power. Thus, military dictatorships coded by Cheibub, Gandhi, and Vreeland identify military-led autocracies, and the strategy of combining the GWF data set with the Cheibub et al. data set can measure military strongman rule, a subset of military-led autocracies. Idi Amin in Uganda, Mobutu Sese Seko in Democratic Republic of the Congo, Rafael Trujillo in the Dominican Republic, and Muammar Gaddafi in Libya are the notable examples of Strongman.

Additionally, I include two civilian counterparts, Boss and Machine. I code the remaining civilian personalist regimes in GWF data as Boss. Dominant-party dictatorships, including “party-based,” “party-personal,” and “oligarchy,” are coded Machine. In addition to these four autocratic regimes, I include Monarchy as a separate regime category. Lai and Slater (2006) and Weeks (2012, 2014) do not consider the conflict propensity of monarchies, theorizing military dictatorships’ belligerence only compared to civilian dictatorships. In contrast, Debs and Goemans (2010) argue that monarchs would be more war-prone than civilian autocrats because of their adverse postouster fates. Last, about 2 percent of all country-years are coded as “not independent,” “occupied by foreign troops,” “ruled by a provisional government charged with overseeing a transition to democracy,” or “lacking a central government.” I code these country-years as Others and include it in the model instead of removing these observations.

I include a set of control variables that might be correlated with political regime type and international conflict. First, I include a state’s national military capabilities, as measured by a state’s composite index of national capabilities score and a binary indicator of major power status in the international system, both as coded in the COW project. Dyadic models control for each state’s military capabilities and major power status, and additionally include an initiator’s proportion of dyadic capabilities. Second, I include a measure of geographic conditions. In the monadic analysis, I include the number of contiguous territorial borders with other states (separated by a land or river border). In the dyadic analysis, I include a dummy variable for contiguity. Last, I control for a state’s alliances. Monadic tests include the total number of a state’s allies, as measured in the COW alliance data (Gibler 2009). Dyadic models include the similarity of the two states’ alliance portfolios. To control for duration dependence, I include a cubic polynomial of the number of years since the last MID initiation.

To control for potential unobserved unit-specific factors, I follow King’s (2001) suggestion to use random effects models. I thus include country-level random effects for the monadic analysis and dyad-level random effects for the dyadic analysis. A pooled model maintains a very strong assumption that the average rate of conflict initiation is the same for all countries (or dyads) and that control variables fully account for the unobservable determinants of a country’s belligerence that may be spuriously correlated with regime type. Countries that have more frequently been under military rule may be fundamentally different from countries that have not. A pooled regression will be heavily confounded with other factors likely to be simultaneously correlated with military regimes and conflict propensity. I also use a conditional fixed effect logit model as an alternative estimator but report the results in the Online Appendix due to space considerations.

Dyadic Tests

Tables 2 and 3 present the results of the twelve different models in which the dependent variable is Side A initiation (Table 2) or Revisionist initiation (Table 3). For comparison, I present both pooled and random effects logit estimates. However, likelihood-ratio tests indicate that there is significant unobserved heterogeneity at directed-dyad level, confirming the need to account for these differences. All the models include six regime-type dummy variables for state A and set Democracy as the baseline category. To test whether military autocracies are more likely to engage in militarized foreign policies than civilian autocracies, I compare the coefficients of military autocracies with those of civilian autocracies. I report the p-values of two-tailed Wald tests assessing the statistical significance of the differences at the bottom of the tables. If hostile security environments of Junta contribute to its conflict propensity, the belligerence of Junta will disappear when I control for territorial threats. Otherwise, I should be able to identify the belligerence of Junta compared to other civilian autocracies even when I include territorial threats.

Table 2.

Militarized Interstate Dispute Initiation (Side A) in Directed-dyad Years.

	Pooled Logit			Random-Effects Logit
	(1)	(2)	(3)	(4)	(5)	(6)
Junta	0.589*** (0.152)	0.528*** (0.159)	0.270* (0.152)	0.307** (0.136)	0.244* (0.137)	0.153 (0.137)
Machine	0.210 (0.134)	0.296** (0.142)	0.313*** (0.115)	0.416*** (0.130)	0.462*** (0.129)	0.404*** (0.117)
Strongman	0.612*** (0.162)	0.614*** (0.168)	0.110 (0.144)	0.564*** (0.177)	0.496*** (0.169)	0.189 (0.173)
Boss	0.934*** (0.156)	1.009*** (0.154)	0.870*** (0.152)	1.197*** (0.167)	1.209*** (0.167)	1.159*** (0.155)
Monarchy	−0.129 (0.222)	−0.156 (0.205)	−0.417** (0.191)	−0.727*** (0.226)	−0.762*** (0.225)	−0.533** (0.217)
Others	0.247 (0.398)	0.327 (0.415)	−0.122 (0.359)	0.210 (0.418)	0.328 (0.417)	−0.171 (0.369)
Power parity	0.429*** (0.155)	0.488*** (0.154)	0.516*** (0.162)	0.718*** (0.166)	0.751*** (0.167)	0.739*** (0.158)
Alliance similarity	−0.774*** (0.153)	−0.757*** (0.160)	−0.230 (0.156)	−0.677*** (0.198)	−0.659*** (0.201)	−0.367** (0.178)
Contiguous dyad	3 741*** (0.128)	3.378*** (0.148)	1.910*** (0.228)	5.193*** (0.205)	4.603*** (0.217)	2.998*** (0.206)
Trade dependence	−2.145 (6.761)	2.120 (5.178)	8.195 (6.792)	6.256 (7.755)	6.332 (7.992)	8.114 (6.932)
Territorial rivalry (strategic)		0.688*** (0.208)			1.446*** (0.219)
Territorial rivalry (KGD)		1 447*** (0.233)			2.335*** (0.205)
Strategic rivalry			0.398*** (0.140)		1	1.080*** (0.161)
KGD rivalry			3.900*** (0.222)		1	3.584*** (0.164)
Constant	−5.821*** (0.194)	−6.115*** (0.226)	−7.470*** (0.230)	−9.753*** (0.343)	−9.838*** (0.327)	−9.451*** (0.286)
Variance (dyad RE)				4.707*** (0.448)	4.241*** (0.415)	1.714*** 0.244)
Test of equality (p values)
Junta = Machine	0.02	0.19	0.80	0.48	0.16	0.10
Junta = Boss	0.06	0.01	0.00	0.00	0.00	0.00
Strongman = Machine	0.01	0.05	0.17	0.39	0.84	0.21
Strongman = Boss	0.06	0.02	0.00	0.00	0.00	0.00
N	849,862	849,862	849,862	849,862	849,862	849,862
Log likelihood	−6,085.6	−5,943.8	−5,224.4	−5,585.6	−5,463.2	−5,062.2

Note: KGD = Klein, Goertz, and Diehl; RE = random effects. Logit estimates with standard errors clustered by dyad (reported in parentheses). All models include each state’s military capabilities, the major power status of each state in the dyad, and a cubic polynomial of peace years (not reported). All regime variables are lagged by one year.

*p < .1.

**p < .05.

***p < .01.

Table 3.

Militarized Interstate Dispute Initiation (Revisionist) in Directed-dyad Years.

	Pooled Logit			Random-Effects Logit
	(1)	(2)	(3)	(4)	(5)	(6)
Junta	0.589*** (0.185)	0.526*** (0.189)	0.245 (0.180)	0.255 (0.167)	0.225 (0.167)	0.136 (0.169)
Machine	0.374*** (0.144)	0.445*** (0.145)	0.465*** (0.122)	0.613*** (0.143)	0.652*** (0.142)	0.590*** (0.129)
Strongman	0.750*** (0.176)	0.729*** (0.172)	0.243 (0.155)	0.729*** (0.195)	0.671*** (0.184)	0.358* (0.184)
Boss	1.095*** (0.173)	1.147*** (0.166)	0.988*** (0.165)	1.387*** (0.180)	1.393*** (0.181)	1.297*** (0.171)
Monarchy	−0.196 (0.298)	−0.244 (0.264)	−0.517** (0.263)	−0.823*** (0.296)	−0.838*** (0.289)	−0.611** (0.287)
Others	−0.401 (0.512)	−0.331 (0.518)	−0.734 (0.507)	−0.370 (0.554)	−0.170 (0.522)	−0.562 (0.502)
Power parity	0.525*** (0.179)	0.576*** (0.174)	0.576*** (0.184)	0.905*** (0.185)	0.924*** (0.185)	0.860*** (0.179)
Alliance similarity	−0.662*** (0.173)	−0.610*** (0.180)	−0.075 (0.169)	−0.494** (0.219)	−0.474** (0.222)	−0.184 (0.195)
Contiguous dyad	3.710*** (0.144)	3.345*** (0.164)	1.741*** (0.254)	5.198*** (0.241)	4.637*** (0.252)	2.843*** (0.233)
Trade dependence	0.821 (6.432)	3.276 (5.159)	11.849** (5.835)	9.123 (7.449)	8.260 (7.834)	10.535 (6.683)
Territorial rivalry (strategic)		0.881*** (0.245)			1.298*** (0.259)
Territorial rivalry (KGD)		1.227*** (0.269)			2.107*** (0.251)
Strategic rivalry			0.392** (0.152)			0.969*** (0.186)
KGD rivalry			3.949*** (0.244)			3.561*** (0.189)
Constant	−6.189*** (0.234)	−6.427*** (0.254)	−7.652*** (0.245)	−10.558** (0.420)	−10.503*** (0.401)	−9.794*** (0.336)
Variance (dyad RE)				5.386*** (0.578)	4.844*** (0.537)	1.893*** (0.287)
Test of equality (p values)
Junta = Machine	0.25	0.68	0.26	0.04	0.02	0.01
Junta = Boss	0.01	0.00	0.00	0.00	0.00	0.00
Strongman = Machine	0.02	0.08	0.15	0.54	0.92	0.19
Strongman = Boss	0.06	0.02	0.00	0.00	0.00	0.00
N	849,862	849,862	849,862	849,862	849,862	49,862
Log-likelihood	−4,991.5	−4,886.1	−4,279.0	−4,554.5	−4,480.7	−4,146.0

Note: Logit estimates with standard errors clustered by dyad (reported in parentheses). All models include each state’s military capabilities, the major power status of each state in the dyad, and a cubic polynomial of peace years (not reported). All regime variables are lagged by one year.

*p < .1.

**p < .05.

***p < .01.

I begin with a pooled logit model that does not include a country’s threat environments. Column 1 of Table 2 shows that Machine is the least aggressive and not statistically different from democracies, but another civilian autocracy, Boss, is the most aggressive of all. The differences not only between Machine and military autocracies but also between Boss and military autocracies are statistically significant. No previous research on the military belligerence predicts that Boss is more aggressive internationally than Strongman. Only Junta > Machine is consistent with the previous studies.

Column 2 adds two variables measuring territorial rivalries, Territorial rivalry (strategic) and Territorial rivalry (KGD), to column 1. As expected, all territorial rivalry indicators have positive and statistically significant coefficients, which is consistent with the existing research (Colaresi, Rasler, and Thompson 2008; Diehl and Goertz 2001). A country is more likely to initiate militarized disputes against its territorial rivals than against other countries. Additionally, both Strategic rivalry and KGD rivalry are positive and statistically significant. This confirms that they capture different aspects of interstate rival relationships. Once I control for a potential initiator’s external security environment, Junta decreases in magnitude, but Machine increases. Consequently, the statistical difference between Junta and Machine disappears. Interestingly, Strongman also decreases in magnitude, making the relative aggressiveness of Boss greater.

A similar pattern emerges when I include interstate rivalries including both territorial and nonterritorial rivalries in column 3. The coefficients on Junta and Strongman decline more in magnitude, and the coefficient on Machine increases, making Machine more belligerent than military autocracies. Thus, the ordering Boss > Machine > Junta > Strongman is found. This is the opposite of what the military belligerence hypothesis predicts.

Columns 4 through 6, my preferred specifications, add dyad random effects to the specifications of columns 1 through 3. These models provide no evidence that military autocracies initiate MIDs at a higher rate than their civilian counterparts, irrespective of whether I include territorial rivalries or not. The results indicate the opposite: not only Boss but also Machine is more conflict-prone than their military counterparts although the difference between Machine and Junta is not statistically different.

The examination of Revisionist initiation also fails to provide a strong support for the relationship between military regimes and conflict initiation. In column 1 of Table 3, the pooled logit model, Junta and Strongman are more belligerent than Machine, but only the difference between Strongman and Machine is statistically significant. However, the inclusion of territorial rivalries wipes out the difference between Junta and Machine and substantially reduces the difference between Strongman and Machine (column 2). When column 3 includes interstate rivalries, I find Boss > Machine > Strongman ≫ Junta. Finally, random-effects models, reported in columns 4 through 6, consistently show that Machine initiates MIDs at a higher annual rate than Junta, a difference significant at the 5 percent level. As in other models, Boss remains more belligerent than Strongman.

Overall, no evidence indicates that military rule makes states more aggressive internationally than civilian rule. All models demonstrate that civilian personalist autocracies are more belligerent than both types of military autocracies, a difference statistically significant. The only evidence for Junta > Machine comes from column 1 of Table 2, the pooled model of Side A initiation. However, the association between military rule and conflict propensity is not robust to including dyad random effects or rivalry relationships. Combined with the finding displayed in Figure 1, this suggests that the conflict-proneness of collegial (and personalist) military dictatorships, compared to elite-constrained party dictatorships, seems driven by countries’ external threat environments rather than by regimes’ inherent characteristics.

It is worth noting that controlling external threat environments does not wipe away the differences among authoritarian regimes. The results consistently show that Boss initiates MIDs at the highest annual rate. Similarly, the differences between democracies and monarchies on the one hand and other types of autocracies remain significant. Particularly, monarchies initiate MIDs at the lowest rate. The probability of MID initiation for Boss is about three times greater than for Democracy and about six times greater than for Monarchy.

Monadic Tests

Tables 4 and A7 report the result of monadic tests using a negative binomial model in which the dependent variable is the number of MID initiations in a given year. Table 4 examines Side A initiation, and Table A7 (reported in the Online Appendix) examines Revisionist initiation. I again report both the results of pooled and random effects models.¹³ Most models offer common results. First, Machine is likely to initiate more MIDs than Junta and Strongman, although the differences are not statistically distinguishable from zero.¹⁴ Second, the inclusion of security environments decreases the coefficients on both military regimes but does not much affect the coefficients on civilian autocracies. All rivalry indicators have positive and statistically significant coefficients, suggesting that a country is more likely to engage in foreign aggression when it resides in hostile neighborhoods. Last, Boss is the most aggressive internationally, and Monarchy is the least aggressive of all. In sum, monadic tests show little empirical relationship between military regimes and conflict propensity. These results provide no support for the idea that military autocracies are more belligerent than civilian autocracies due to their inherent characteristics or leaders’ personal backgrounds.

Table 4.

Militarized Interstate Dispute Initiation (Side A) in Country-Years.

	Pooled NB			Random-Effects NB
	(1)	(2)	(3)	(4)	(5)	(6)
Junta	0.144 (0.133)	0.049 (0.117)	0.035 (0.130)	0.203 (0.126)	0.110 (0.127)	0.081 (0.128)
Machine	0.215 (0.195)	0.287 (0.177)	0.226 (0.168)	0.140 (0.189)	0.060 (0.185)	0.105 (0.162)
Strongman	0.136 (0.167)	0.165 (0.163)	−0.083 (0.161)	0.259 (0.202)	0.075 (0.167)	−0.050 (0.199)
Boss	0.485** (0.222)	0.529*** (0.182)	0.494*** (0.185)	0.555*** (0.202)	0.491** (0.192)	0.422** (0.185)
Monarchy	−0.170 (0.183)	−0.364* (0.204)	−0.470** (0.184)	−0.864** (0.424)	−0.975*** (0.363)	−0.937*** (0.345)
Others	−0.188 (0.397)	0.012 (0.475)	−0.361 (0.361)	−0.034 (0.522)	−0.033 (0.542)	−0.325 (0.402)
Total borders	0.070*** (0.018)	0.050*** (0.019)	0.083*** (0.016)	0.092*** (0.031)	0.097*** (0.029)	0.103*** (0.022)
Military capabilities	5.310** (2.159)	3.387 (2.091)	2.491 (2.059)	−2.108 (1.757)	−4.747* (2.441)	−4.210 (2.713)
Number of allies	0.001 (0.005)	−0.001 (0.005)	0.003 (0.005)	0.012* (0.006)	0.007 (0.007)	0.009 (0.006)
Major power	0.171 (0.278)	0.309 (0.256)	0.195 (0.265)	1.321*** (0.379)	1.359*** (0.389)	0.970** (0.395)
Trade openness	0.022 (0.052)	0.037 (0.044)	0.029 (0.044)	−0.033 (0.096)	−0.023 (0.083)	−0.007 (0.072)
Territorial rivalry (strategic)		0.344** (0.154)			0.715*** (0.216)
Territorial rivalry (KGD)		0.578*** (0.172)			0.440*** (0.146)
Strategic rivalry			0.450*** (0.116)			0.589*** (0.141)
KGD rivalry			1.295*** (0.146)			1.264*** (0.158)
Constant	−1.271*** (0.204)	−1 791*** (0.196)	−2.814*** (0.259)	−2.065*** (0.378)	−2.519*** (0.302)	−3.272*** (0.288)
Variance (country RE)				0.715*** (0.156)	0.602*** (0.126)	0.338*** (0.089)
Test of equality (p values)
Junta = Machine	0.73	0.25	0.36	0.73	0.79	0.90
Junta = Boss	0.11	0.01	0.03	0.06	0.03	0.07
Strongman = Machine	0.72	0.59	0.16	0.55	0.93	0.47
Strongman = Boss	0.11	0.08	0.01	0.20	0.02	0.04
N	5,986	5,986	5,986	5,986	5,986	5,986
Log likelihood	−2,984.9	−2,930.9	2,848.0	−2,887.5	2,851.3	−2,791.5

Note: NB = Negative binomial. Negative binomial estimates with standard errors clustered by country (reported in parentheses). All models include a cubic polynomial of peace years. All regime variables, trade openness, and the number of alliances are lagged by one year.

*p < .1.

**p < .05.

***p < .01.

Testing Additional Implication: Differentiating between Territorial and Nonterritorial MIDs

I test an additional observable implication that flows from my argument. If territorial threats from neighboring countries tend to produce both Junta and increased conflict propensity, territorial MIDs rather than nonterritorial MIDs would be responsible for the association, found in column 1 of Table 2 and previous studies, between Junta and conflict initiation. Contrarily, if the military belligerence idea holds true, Junta should be more prone than its civilian counterpart to initiate militarized disputes in both territorial or nonterritorial MIDs. To test this observable implication, I do not include territorial rivalries and differentiate between territorial and nonterritorial MIDs.

Figure 2 displays the coefficients on authoritarian regimes from pooled logit models (panel a) and from random effects logit models (panel [b]). The baseline category is democracies. Models 1 and 3 probe the initiation of territorial MIDs, while models 2 and 4 investigate the initiation of nonterritorial MIDs. Congruent with expectations, Junta (and Strongman) initiates territorial MIDs at a higher rate than Machine, and the difference is statistically significant. However, this does not hold true in the context of nonterritorial MIDs. No evidence suggests that military autocracies are more likely to initiate nonterritorial MIDs than civilian autocracies. As the bottom panel of Figure 2 illustrates, a similar pattern emerges when I estimate a random effects model.

Figure 2.

Differences in coefficients between territorial MIDs and nonteritorial MIDs from logit regressions. The graph shows the logit regression coefficients from separately estimated models. Circles show the point estimates, and horizontal line segments associated with circles show the 95 percent confidence intervals. All models include full control variables for dyadic models.

It is also worth noting that the inclusion of territorial rivalries substantially reduces or removes the difference between Junta and Machine in the models of territorial MIDs (see Tables A5 and A6). However, the same cannot be said for the models of nonterritorial MIDs: the inclusion of territorial rivalries does not significantly affect the regime coefficients. These results provide additional evidence that military regimes’ territorial threats are important to explaining to the relationship between military regimes and international conflict.

These results have a significant implication for the possibility of selection effects: when a country has weak constraints on the use of violence, the military is more likely to acquire political power, and it is less likely to settle territorial disputes peacefully.¹⁵ This selection effect may pose a challenge to my argument that external territorial threats promote the emergence of military regimes as well as increase the likelihood of conflict initiation. However, the results illustrated by Figure 2 and reported in Tables A5 and A6 lessen this concern. If circumstances favoring the use of violence cause both the emergence of military regimes and their greater likelihood of conflict initiation, we should observe that military regimes are more likely than civilian regimes to initiate not only territorial MIDs but also nonterritorial MIDs. Additionally, the inclusion of territorial rivalries should not significantly affect the relationship between military regimes and territorial MIDs initiations. However, Figure 2 and Tables A5 and A6 demonstrate that this is not true.

Finally, another interesting finding from Figure 2 is that the peacefulness of Monarchy and Democracy and the belligerence of Boss stand out only in the models of nonterritorial MIDs. Conversely, Junta, Machine, and Strongman differ little from each other in the models of nonterritorial MIDs. This may suggest the need to differentiate between territorial and nonterritorial disputes in analyzing the effect of political regime types on conflict initiation, confirming the importance of the issue-based approach (Hensel et al. 2008).

Robustness Checks

To ensure the robustness of my results, I perform several additional analyses. Due to space considerations, the results of these robustness checks are discussed briefly but are available in the Online Appendix unless indicated otherwise.

First, to ensure that the main results are driven by minor or accidental disputes, I use alternative measures of international conflict: Use of Force MIDs and ICB crises. The use of alternative MID measures does not alter the key findings (see Tables A8 through A11 for Use of Force MIDs, and Tables A12 and A13 for ICB crises).

Second, I compare military-led autocracies with civilian-led autocracies, monarchies, and democracies using Cheibub, Gandhi, and Vreeland’s (2010) regime-type data without accounting for the personalism dimension (see Tables A14–A17). Recall that the Cheibub et al. regime classification depends solely on the identity of the regime leader. Next, I further combine the measure of personalism constructed by Weeks (2012, 2014) with the Cheibub et al.’s regime classification to create the four-way dictatorship classification (see Tables A18–A21).¹⁶ Last, I use the original GWF measures of party-based regimes and military regimes for Machine and Junta (see Tables A22–A25). In all cases, monadic and dyadic tests fail to find that military-led autocracies initiate MIDs at a higher annual rate than civilian-led autocracies. These results indicate that the main results presented above are not an artifact of my decision to use the GWF data or aggregate all military hybrids.

Third, I use alternative samples to test the relationship between military dictatorships and conflict propensity. I restrict the dyad sample to politically relevant dyads that include at least one major power or two states separated by no more than twenty-four miles of water (Tables A28 and A29). Similarly, I include only autocracies for the monadic analysis or autocratic initiators for the dyadic analysis (Tables A30 and A31). The central findings hold in both contexts.

Last, I estimate a conditional fixed effects logit model to control for time-invariant unobserved heterogeneity between dyads. This estimator allows me to explore the within-unit (country or directed dyad) effects of military autocracies. Therefore, if either leaders’ personal backgrounds or regime attributes are the predictors of a military dictatorship’s conflict propensity, we should be able to find that Argentina becomes more aggressive when transitioning from a civil to a military dictatorship and not only that Argentina is more aggressive than Mexico. I find that the fixed effects logit estimates are similar to the random effects logit estimates reported here (see Tables A32 and A33).

Conclusion

The contribution of this article is twofold. First, it establishes that military regimes are more likely to emerge and exist when countries are faced with sustained territorial threats from neighboring countries. This relationship does not hold for other types of autocracies. Second, building on this finding, I show that the empirical evidence for the military belligerence hypothesis is significantly weakened once I control for territorial rivalries. In fact, by capturing territorial threats, military regimes’ relative aggressiveness disappears. Further analysis demonstrates that territorial threats from neighboring countries likely drive the relationship between military autocracies and increased conflict propensity reported in the previous studies.

These findings indicate that military regimes initiate militarized conflicts because they are located in more hostile security environments, not because they are either institutionally fragile or predisposed toward using force. These results imply that the military belligerence hypothesis should be subjected to further examination. As briefly discussed, conflicting theories and incongruous empirical results mark the literature proposing this hypothesis. In addition, we know neither which mechanism is responsible for the observed pattern, nor whether the previous findings are robust, since these studies use different regime-type data sets that draw on different definitions of military regimes. Future studies should also examine whether the key assumptions of these previous studies are empirically supported. For example, does either civilian or military regime leadership predict autocrats’ posttenure fates or governing parties’ institutionalization? Questions like these are central to the previous arguments regarding military regimes’ conflict-proneness.

The argument and findings presented in this article have further implications for future study. First, future study should explore the relationship between military regimes and rivalry (particularly territorial rivalry) initiation. In this article, I treat territorial rivalries as exogenous. However, territorial rivalries may reflect leaders’ purposeful choices in the sense that military autocrats may initiate rivalries with neighboring countries as a means of strengthening their political power. Owsiak and Rider (2013) and Rider and Owsiak (2015) recently examine the onset and termination of contiguous rivalries, but they do not differentiate among different types of autocracies. Accordingly, it is important to determine whether military autocracies are more likely to initiate and sustain territorial rivalries than civilian autocracies.

Second, this article highlights the need for continued research into the relationship between political regimes and conflict propensity. As previous studies contend, the conflict propensity of different types of autocracies varies significantly. However, my research departs from the existing studies in that I fail to support military regimes’ belligerence relative to civilian regimes. Instead, civilian personalist regimes are the most belligerent of all, which is consistent with the previous studies (Weeks 2008, 2012). Meanwhile, monarchies are found to be the most peaceful regime type, along with democracies. This may be because monarchies successfully construct stable ruling coalitions by sharing power via consultative councils (Gandhi 2008) or by utilizing political culture to enhance cohesion among ruling members (Menaldo 2012). The relationship between monarchies and international conflict, which (to the best of my knowledge) has yet to be subjected to a systematic investigation, warrants further research.

Last, future study should further probe the impact of sustained territorial threats on military regimes. For instance, the effects of territorial rivalries are likely to proceed and accumulate over time. Thies (2005) argues that interstate rivalries may represent a slow-moving, causal process that has more of an incremental impact on domestic political bargaining and political institutions. It is thus possible that the longer a country engages in territorial rivalries, the greater the military’s capacity to intervene in politics. If so, I should be able to examine both the short- and long-term effects of territorial rivalries on domestic politics. However, the binary indicator of whether a country is experiencing territorial rivalries, adopted in this article, is not well suited to capturing the long-run effect of territorial rivalries. I may borrow the empirical strategy used in Gerring, Thacker, and Alfaro (2012). They calculate a “stock” measure of democracy, extending back to 1900 with several annual depreciation rates to examine the impact of a country’s democratic history on its level of human development. This strategy would be helpful for investigating the long-term effect of a country’s history of territorial rivalries on military regimes and domestic politics.

Supplemental Material

Supplemental Material, Online_Appendix - Are Military Regimes Really Belligerent?

Supplemental Material, Online_Appendix for Are Military Regimes Really Belligerent? by Nam Kyu Kim in Journal of Conflict Resolution

Supplemental Material

Supplemental Material, Replication_files_(1) - Are Military Regimes Really Belligerent?

Supplemental Material, Replication_files_(1) for Are Military Regimes Really Belligerent? by Nam Kyu Kim in Journal of Conflict Resolution

Footnotes

Author’s Note

Nam Kyu Kim is now affiliated to Sungkyunkwan University, Department of Political Science & Diplomacy, Jongno-gu, Seoul, South Korea.

Acknowledgments

I am grateful to Alex Kroeger, Ross Miller, participants in workshops at the University of Nebraska–Lincoln, and the anonymous reviewers for comments on earlier versions of this article.

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.

Supplemental Material

Supplementary material is available for this article online.

Notes

References

Belkin

Aaron

Schofer

Evan

. 2003. “Toward a Structural Understanding of Coup Risk.” The Journal of Conflict Resolution 47 (5): 594–620.

Brecher

Michael

Wilkenfeld

Jonathan

. 1997. A Study of Crisis. Ann Arbor: University of Michigan Press.

Brooks

Risa A.

2003. “Making Military Might: Why Do States Fail and Succeed? A Review Essay.” International Security 28 (2): 149–91.

Caselli

Francesco

Morelli

Massimo

Rohner

Dominic

. 2015. “The Geography of Interstate Resource Wars.” The Quarterly Journal of Economics 130 (1): 267–315.

Cheibub

José Antonio

Gandhi

Jennifer

Vreeland

James Raymond

. 2010. “Democracy and Dictatorship Revisited.” Public Choice 143 (1-2): 67–101.

Colaresi

Michael P.

Rasler

Karen

Thompson

William R

. 2008. Strategic Rivalries in World Politics: Position, Space and Conflict Escalation. Cambridge, MA: Cambridge University Press.

Debs

Alexandre

. 2016. “Living by the Sword and Dying by the Sword? Leadership Transitions In and Out of Dictatorships.” International Studies Quarterly 60 (1): 73–84.

Debs

Alexandre

Goemans

H. E.

. 2010. “Regime Type, the Fate of Leaders, and War.” American Political Science Review 104 (03): 430–45.

Desch

Michael C.

1998. “Soldiers, States, and Structures: The End of the Cold War and Weakening U.S. Civilian Control.” Armed Forces & Society 24 (3): 389–405.

10.

Diehl

Paul F.

Goertz

Gary

. 2001. War and Peace in International Rivalry. Ann Arbor: University of Michigan Press.

11.

Downes

Alexander B.

Sechser

Todd S.

. 2012. “The Illusion of Democratic Credibility.” International Organization 66 (03): 457–89.

12.

Escriba-Folch

Abel

Wright

Joseph

. 2010. “Dealing with Tyranny: International Sanctions and the Survival of Authoritarian Rulers.” International Studies Quarterly, International Studies Quarterly 54 (2): 335–59.

13.

Gandhi

Jennifer

. 2008. Political Institutions under Dictatorship. Cambridge, MA: Cambridge University Press.

14.

Geddes

Barbara

. 2003. Paradigms and Sand Castles: Theory Building and Research Design in Comparative Politics. Ann Arbor: University of Michigan Press.

15.

Geddes

Barbara

Frantz

Erica

Wright

Joseph G.

. 2014. “Military Rule.” Annual Review of Political Science 17:147–62.

16.

Geddes

Barbara

Wright

Joseph

Frantz

Erica

. 2014. “Autocratic Breakdown and Regime Transitions: A New Data Set.” Perspectives on Politics 12 (02): 313–31.

17.

Gerring

John

Thacker

Strom

Alfaro

Rodrigo

. 2012. “Democracy and Human Development.” The Journal of Politics 74 (01): 1–17.

18.

Ghosn

Faten

Bremer

Stuart

. 2004. “The MID3 Data Set, 1993–2001: Procedures, Coding Rules, and Description.” Conflict Management and Peace Science 21 (2): 133–54.

19.

Gibler

Douglas M.

2009. International Military Alliances, 1648–2008. Washington, DC: CQ Press.

20.

Gibler

Douglas M.

2012. The Territorial Peace: Borders, State Development, and International Conflict. Cambridge, MA: Cambridge University Press.

21.

Gibler

Douglas M.

Tir

Jaroslav

. 2010. “Settled Borders and Regime Type: Democratic Transitions as Consequences of Peaceful Territorial Transfers.” American Journal of Political Science 54 (4): 951–68.

22.

Hensel

Paul R.

1996. “Charting a Course to Conflict: Territorial Issues and Interstate Conflict, 1816–1992.” Conflict Management and Peace Science 15 (1): 43–73.

23.

Hensel

Paul R.

2001. “Contentious Issues and World Politics: The Management of Territorial Claims in the Americas, 1816–1992.” International Studies Quarterly 45:81–109.

24.

Hensel

Paul R.

Mitchell

Sara McLaughlin

Sowers

Thomas E.

Thyne

Clayton L.

. 2008. “Bones of Contention Comparing Territorial, Maritime, and River Issues.” Journal of Conflict Resolution 52 (1): 117–43.

25.

Heston

Alan

Summers

Robert

Aten

Bettina

. 2011. “Penn World Table Version 7.0.” Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania. Accessed September 2, 2013. https://pwt.sas.upenn.edu/php_site/pwt_index.php

26.

Hintze

Otto

. 1975. “The Formation of States and Constitutional Development: A Study in History and Politics.” In The Historical Essays of Otto Hintze, edited by Felix

Gilbert

, 178–15. New York: Oxford University Press.

27.

Horowitz

Michael C.

Stam

Allan C.

. 2014. “How Prior Military Experience Influences the Future Militarized Behavior of Leaders.” International Organization 68:527–59.

28.

Huth

Paul K.

1996. Standing Your Ground: Territorial Disputes and International Conflict. Ann Arbor: University of Michigan Press.

29.

King

Gary

. 2001. “Proper Nouns and Methodological Propriety: Pooling Dyads in International Relations Data.” International Organization 55 (2): 497–507.

30.

Klein

James P.

Goertz

Gary

Diehl

Paul F.

. 2006. “The New Rivalry Dataset: Procedures and Patterns.” Journal of Peace Research 43 (3): 331–48.

31.

Lai

Brian

Slater

Dan

. 2006. “Institutions of the Offensive: Domestic Sources of Dispute Initiation in Authoritarian Regimes, 1950–1992.” American Journal of Political Science 50 (1): 113–26.

32.

Lasswell

Harold D.

1941. “The Garrison State.” American Journal of Sociology 46 (4): 455–68.

33.

Marshall

Monty G.

Gurr

Tedd R.

Jaggers

Keith

. 2013. “Polity IV Project: Political Regime Characteristics and Transitions, 1800–2012.” Center for Systemic Peace. Accessed December 5, 2013. http://www.systemicpeace.org/polity/polity4.htm

34.

Menaldo

Victor

. 2012. “The Middle East and North Africa’s Resilient Monarchs.” The Journal of Politics 74 (03): 707–22.

35.

Mettler

Simon A.

Reiter

Dan

. 2013. “Ballistic Missiles and International Conflict.” Journal of Conflict Resolution 57 (5): 854–80.

36.

Nordlinger

Eric A.

1977. Soldiers in Politics: Military Coups and Governments. Englewood Cliffs, NJ: Prentice Hall College Div.

37.

Owsiak

Andrew P.

Rider

Toby J.

. 2013. “Clearing the Hurdle: Border Settlement and Rivalry Termination.” The Journal of Politics 75 (03): 757–72.

38.

Pilster

Ulrich

Böhmelt

Tobias

. 2011. “Coup-proofing and Military Effectiveness in Interstate Wars, 1967–99.” Conflict Management and Peace Science 28 (4): 331–50.

39.

Rasler

Karen

Thompson

William R.

. 2004. “The Democratic Peace and a Sequential, Reciprocal, Causal Arrow Hypothesis.” Comparative Political Studies 37 (8): 879–908.

40.

Rasler

Karen

Thompson

William R.

. 2005. Puzzles of the Democratic Peace: Theory, Geopolitics and the Transformation of World Politics. London, UK: Palgrave Macmillan.

41.

Rider

Toby J.

2013. “Uncertainty, Salient Stakes, and the Causes of Conventional Arms Races.” International Studies Quarterly 57 (3): 580–91.

42.

Rider

Toby J.

Owsiak

Andrew P.

. 2015. “Border Settlement, Commitment Problems, and the Causes of Contiguous Rivalry.” Journal of Peace Research 52 (4): 508–21.

43.

Sarkees

Meredith Reid

Schafer

Phil

. 2000. “The Correlates of War Data on War: An Update to 1997.” Conflict Management and Peace Science 18 (1): 123–44.

44.

Sechser

Todd S.

2004. “Are Soldiers Less War-prone than Statesmen?” Journal of Conflict Resolution 48 (5): 746–74.

45.

Senese

Paul D.

1996. “Geographical Proximity and Issue Salience: Their Effects on the Escalation of Militarized Interstate Conflict.” Conflict Management and Peace Science 15 (2): 133–61.

46.

Slater

Dan

. 2003. “Iron Cage in an Iron Fist: Authoritarian Institutions and the Personalization of Power in Malaysia.” Comparative Politics 36 (1): 81–101.

47.

Stepan

A. C.

1971. The Military in Politics: Changing Patterns in Brazil. Princeton, NJ: Princeton paperbacks Princeton University Press.

48.

Svolik

Milan W.

2013. “Contracting on Violence the Moral Hazard in Authoritarian Repression and Military Intervention in Politics.” Journal of Conflict Resolution 57 (5): 765–94.

49.

Thies

Cameron G.

2005. “War, Rivalry, and State Building in Latin America.” American Journal of Political Science 49 (3): 451–65.

50.

Thompson

William R.

1996. “Democracy and Peace: Putting the Cart before the Horse?” International Organization 50 (01): 141–74.

51.

Thompson

William R.

Dreyer

David R.

. 2011. Handbook of International Rivalries, 1494–2010. Washington, DC: CQ Press.

52.

Tilly

Charles

. 1985. “War Making and State Making as Organized Crime.” In Bringing the State Back In, edited by Evans

Peter

Rueschemeyer

Dietrich

Skocpol

Theda

, 169–91. Cambridge: Cambridge University Press.

53.

Vasquez

John A.

2009. The War Puzzle Revisited. Cambridge, MA: Cambridge University Press.

54.

Weeks

Jessica L.

2008. “Autocratic Audience Costs: Regime Type and Signaling Resolve.” International Organization 62 (1): 35–64.

55.

Weeks

Jessica L.

2012. “Strongmen and Straw Men: Authoritarian Regimes and the Initiation of International Conflict.” American Political Science Review 106 (02): 326–47.

56.

Weeks

Jessica L. P.

2014. Dictators at War and Peace. Ithaca, NY: Cornell University Press.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

131.30 MB

0.00 MB

0.24 MB