Generals in the Cabinet: Military Participation in Government and International Conflict Initiation

Transcription

1 Generals in the Cabinet: Military Participation in Government and International Conflict Initiation Peter B. White [Working paper, please do not circulate or cite without contacting author] Word Count: 12,317 How do military officers in a national cabinet affect a state's likelihood of international conflict? We know a great deal about how regime type affects international conflict, but much less about how intraregime variation in military involvement in government affects a state s decision to use force particularly in civilian-led regimes. I argue that higher levels of military involvement in civilian-led regimes increase the likelihood that a state will initiate international conflict, but this only occurs when military officers occupy positions outside of national defense. When the military intrudes into areas of civilian responsibility, it is indicative of the military becoming a political faction and not just an adviser on national defense. This increases a state s likelihood of using force in international relations through two mechanisms: The first is the diffusion of decision-making authority beyond a civilian leader. When military officers are deciders and not just advisers in a government, they have the ability to unilaterally escalate international tensions. The second mechanism is the politicization of the decision to use force. When the military is a political actor, civilian and military elites have added incentives to increase international tensions in order to strengthen their faction in intra-regime politics. I test this theory cross-nationally using a new dataset of the number and type of positions held by military officers in national cabinets. I find strong support for the theory. In civilian-led regimes, increases in the number of non-security government positions held by military officers greatly increases the probability that a state will initiate violent international conflict. 1

2 Introduction How does military participation in the government affect the likelihood that a country initiates international conflict? Recent research has substantially advanced our understanding of the conflict propensity of military regimes (e.g., Sechser 2004; Lai and Slater 2006; Weeks 2012, 2014). However the military is not absent from regimes that aren't military dictatorships. There is substantial variation in the level of military representation in the cabinets and state councils of personalist dictatorships, one-party regimes, monarchies, and even democracies. Outside of overall regime category, we don't understand the cross-national impact of military involvement in government on international conflict. In this article, I argue that military participation in civilian-led regimes increases the likelihood that they initiate international conflict; however, this only occurs when military officers occupy positions in the government outside of national defense. Specifically, when military officers step outside their traditional role as advisers on national defense, the military becomes another political faction in the government, and this introduces two institutional pathologies into a state s decision to use force. The first of these is the politicization of the decision to use force. When the military is not only the institution charged with carrying out national defense, but is also a political faction, civilmilitary relations becomes a conflictual bargain (e.g., Svolik 2012). Here, both civilian and military elites have incentives to use international conflict for political gain. International conflict can insulate civilian leaders from coups and professionalize politicized militaries. Factions within the military may see conflict as an opportunity to gain resources and autonomy from the government. The second pathology is the diffusion of decision-making authority beyond the civilian leader of the state. When military officers are not just advisers on the decision to use force, but also deciders, civilian control is undermined and a single, direct chain of command from the civilian leader to the military no longer exists (Brooks 2008). In this context, the number of actors who have the ability to escalate 2

3 international tensions or even initiate conflict increases, and there is the opportunity for the military or groups within the military to unilaterally take actions that begin conflict or make it more likely. I test this theory quantitatively using new yearly, cross-national data on the number of military officers in national cabinets and state councils and the type of role that they play the Military Participation in Government (MPG) data. This data allow for the relative militarization of national governments to be measured relatively continuously in all states, including all types of civilian-led regimes In examining the initiation of militarized disputes that rise to the level of actual violence (where there are fatalities), I find strong evidence in support of the theory. Increased military officers in government correspond to the increased likelihood that a state initiates violent international conflict, when the military is involved in the non-security aspects of government. This demonstrates that military officers exert a substantial impact on conflict propensity in civilian-led regimes. This article makes two main contributions. First, it moves the quantitative literature on civilmilitary relations beyond military regimes and coups to consider the spectrum of military involvement in government that exists in between military and civilian dominance. Second, it demonstrates that military participation in government in civilian regimes is a major driver of international conflict. The literature has hitherto focused on the conflict propensity of military regimes. This effectively disaggregates the internal politics of civilian regimes to allow us to judge their relative risk for international conflict based on an observable characteristic i.e., military participation in the government. This article proceeds in six sections. In the first, I outline the existing literature on regime type, the military, and international conflict. In the second, I discuss my theory of how the military influences international conflict in civilian-led regimes. In the third, I introduce the new Military Participation in Government (MPG) data, and discuss its application to international conflict and the 3

4 general research design. In the fourth section, I present the results of the empirical analysis. In the fifth and sixth sections, I conduct additional analyses to verify the robustness of the main results and conclude with some discussion of the scholarly and policy implications of the argument and results. Literature There is an expansive literature on the effect of regime type on international conflict. Much has been written about the rarity with which democracies fight each other (e.g., Doyle 1986; Maoz and Russet 1993; Schultz 1999; Huth and Allee 2002). 1 Beyond democracies, there has been extensive progress in disaggregating autocracy into distinct categories. Lai and Slater (2006) have found that military regimes are more conflict prone than civilian dictatorships. Weeks (2012, 2014) has developed a four-party typology of autocracies regimes headed by "juntas" (regimes where the military as an institution rules), "strongmen" (personalist dictators who are military officers), "bosses" (civilian personalist dictatorships), and "machines" (civilian party regimes). Weeks finds, importantly, that civilian party regimes are no more conflict-prone than democracies, while juntas, strongmen, and bosses are more conflict-prone. There has been another line of research on how the state leaders affect the conflict propensity of the state. Much of this links implicitly with the regime-type literature, focusing on how leaders' likelihood of punishment for defeat in international conflict varies across regime types and how this in turn affects leaders' conflict behavior. Leaders in mixed regimes and autocracies face a higher likelihood of punishment for defeat, and this punishment can also include substantial personal costs, including imprisonment and death (Goemans 2000, 2008). Though they may face removal from office in elections (Bueno de Mesquita et al 2004), particularly if they (or their party) are responsible for the 1 There is also work that attributes the Democratic Peace to other factors, such as trade (e.g., Gartzke 2007) or stable borders (e.g., Gibler 2007). 4

5 conflict (Croco 2011, 2015), elected leaders are unlikely to face high personal costs. Accordingly, they are better able to seek peace and make concessions (Debs and Goemans 2010). More recent literature has examined the effect of leader attributes on international conflict initiation. Horowitz and Stam (2014) and Horowitz, Stam, and Ellis (2015) have found that the military experience of state leaders matters a great deal, even when holding constant regime-type specifically, they find that former rebels and military officers without combat experience are particularly likely to initiate international conflict. Relatedly, Colgan (2013) finds that revolutions select more conflictprone leaders, which lead revolutionary states to become more conflict-prone; Weeks and Colgan (2015) find that this is particularly true when the revolution leads to a personalist dictatorship. We know a great deal about how leaders and political systems affect the conflict propensity of states, yet there are significant gaps in our understanding. Notably, while we know that militarydominated regimes are more likely to initiate international conflict (Weeks 2012, 2014; Lai and Slater 2006) and that certain types of military experience make leaders more likely to initiate conflict (Horowitz and Stam 2014; Horowitz, Stam, and Ellis 2015), we don't understand the impact of military involvement in government outside of the leader and the overall regime category particularly in civilian-led regimes. Existing literature does not take into account the degree of military or civilian power in single-party states, personalist dictatorships, and democracies. Yet it is clear that the military is not absent from government in regimes that aren't military dictatorships. In a personalist context, in Stalin's Soviet Union, the foremost active-duty general in the Red Army, Marshall Georgi Zhukov, served as member of the Central Committee of the Communist Party, and later with the army's backing played a king-maker role in supporting Kruschev's ascension and the downfall of NKVD 2 chief Lavrenty Beria in 1953 following Stalin's death (Roberts 2012, Ch. 11). He was made a member 2 The NKVD was the precursor to the KGB. 5

6 of the ruling Presidium of the Politiburo in 1957 the first active-duty Soviet military officer to do so (Colton 1981). In a single-party context, the People's Liberation Army (PLA) has played an extensive, and varied, role in government and politics throughout Communist rule in China. At the height of its political power during the latter part of the Cultural Revolution, the PLA took control of many government ministries as well as most provincial governments (MacFarquhar and Schoenals 2006). Since then, while taking a reduced role and focusing on national security affairs, the PLA has maintained multiple seats in both the Central Committee and Politburo of the Communist Party (Miller 2015). Democracies also are not immune to military involvement in government, with the records of Bangladesh, Pakistan, Turkey, and Sri Lanka showing periods where military cabinet officials or parallel military ruling councils co-existed (at times uneasily) with elected civilian governments. The recent appointment of a substantial number of recently-retired and active duty generals to high-level positions in the Trump Administration has demonstrated that even in consolidated democracies, high levels of military involvement in government are possible. How does this type of military involvement in government where the military is involved in the highest levels of government, but does not necessarily control the government affect the likelihood that a state seeks international conflict? There are clear cases of civil-military dysfunction contributing to international conflict that cannot easily be explained by overall regime type: for example, the Pakistani military's possible role in autonomously initiating the 1999 Kargil War with India 3 and the role that political struggles between civilian and military factions played in miscalculations by the Egyptian government in the lead-up to the 1967 war with Israel (Brooks 2008). In 1999, Pakistan was headed by an elected civilian government, and Egypt in 1967 by a dominant 3 Nawaz blames Musharraf for Kargil. 5/28/2006. The Times of India. Accessed 4/7/

7 party regime split between civilian technocrats and military officers. In each case, the military was a political actor, but not the dominant one in the regime. What is needed is a theory to explain how the military participation in civilian-led regimes affects the likelihood that those regimes use force. Theory As the Egyptian and Pakistani cases suggest, when the military has a role in civilian-led government, but is not the dominant actor, it can significantly affect the state's conflict propensity. Within civilianled regimes, how does variation in military involvement in government affect conflict propensity? In this section, I outline my theory. The argument I put forward focuses on institutional pathologies that emerge when the military takes a political role outside that of adviser in a government. I see institutional, not individual pathologies as being the key mechanism. I do not argue, for example, that military officers, individually, are inherently more aggressive with regards to the decision to use force. Indeed, there is strong evidence that military officers, at the individual level, are no more likely to favor military solutions to foreign policy problems than are civilian elites (e.g,. Betts 1977, 1991; Feaver and Gelpi 2011). 4 However, when militaries take on a political role beyond advising on national defense, the decision-making processes surrounding the use of force in international relations becomes distorted, and international conflict becomes more likely. The key insight of my theory is that military officers do increase the likelihood that a country will initiate international conflict, but only when the military occupies leadership positions in the government outside of the confines of defense and security. When the military occupies positions in the government beyond security e.g., as ministers with non-defense portfolios or as members of a ruling council it is indicative of the military being a part of the political leadership of the country and 4 Though this does not extend to the use of force in conflicts that have already begun, when military officers tend to favor escalation (e.g,. Betts 1977, 1991; Feaver and Gelpi 2011). 7

8 taking a role in the decision to use force, rather than just implementing that decision. In many states without undue military intrusion into politics, including in democracies, an active-duty or retired military minister of defense or national security adviser is common or even de riguer and does not necessarily reflect military intrusion into traditionally civilian politics. 5 Rather it reflects the military's primary role in advising non-military decision-makers on defense-related decisions. For example, in the present-day Chinese government, the PLA is heavily represented in national security-bodies, such as the Central Military Commission, but is absent from the highest-level general decision-making bodies, such as the Presidium of the Politburo of the Chinese Communist Party (CCP), and has ceded ultimate decision-making authority to the civilian cadres of the CCP (Miller 2015). It is military intrusion into traditionally civilian areas of the decision-making process when military officers become deciders rather than advisers that leads to higher conflict propensity. There are two mechanisms behind this: The first is the politicization of national security decision-making due to competition over political power between civilian and military elites. When the military is a political actor, this can make international conflict more attractive to civilian and military leaders. Civilian leaders facing politically-active militaries may see military conflict as a way to politically weaken and professionalize militaries, while factions within militaries may see conflict as an opportunity to gain resources, power, and autonomy. The second mechanism is the diffusion of decision-making authority to both military and civilian elites. When military officers intrude into traditionally civilian decision-maker roles, this undermines civilian control and the chain of command, making it possible for military units to mobilize and even use force without explicit orders from civilian leaders, 5 However, a military minister of defense is not considered the ideal in democratic civil-military relations. In democracies, where civil-military relations can be conceived of as a principal-agent interaction (Feaver 2005), a military minister of defense likely weakens civilian oversight and may make shirking by the military easier. Indeed, Desch (1999, 65) terms a civilian minister of defense the sine qua non of civilian control. However, a military minister of defense in and of itself is not necessarily indicative of the military having an outsized role in the politics of a country or the decision to use force rather than the implementation of that decision. 8

9 potentially leading to the unintended escalation of conflict. The two mechanisms relate to each other in that the politicization of the military creates added incentives for civilian and military elites to start international conflict and escalate tensions, while the diffusion of decision-making authority to both military and civilian elites makes it easier for international conflict to be started or escalated, even unintentionally. These two mechanisms mean that states with higher levels of military participation in the traditionally civilian, non-security functions of government will be more likely to initiate international conflict. Turning first to the politicization of the decision to use force: When the military occupies positions in the government outside of those related to national defense, it is stepping into traditionally civilian politics and becoming a political faction. This may be a way to keep tabs on civilian politics and protect the military s traditional institutional interests, such as budget, autonomy, and defense policy preferences. However, where the military participates in non-defense politics, it often has institutional interests that do not pertain directly to defense, such as commercial holdings (e.g., Egypt, Sri Lanka, Indonesia), criminal activity or smuggling (e.g., Syria), or other perquisites and privileges of the officer class. Where the military takes an active role in non-defense politics, civil-military relations can be regarded as a conflictual bargain where both civilian and military elites are seeking a greater share of the state s power and resources at the expense of the other (Svolik 2012). Civilian and military leaders have incentives to use international conflict to shape this bargain. Since bargaining failure in highly conflictual civil-military relations is often a coup attempt (Svolik 2012), at the most basic level, civilian leaders may see international conflict as a way to make the launching of a coup more difficult, as it occupies and distracts the military and makes it functionally more difficult for a coup to be plotted and launched (Piplani and Talmadge 2016). Civilian leaders may pursue international conflict as a way to coup-proof (Belkin and Schoffer 2005; Goemans and 9

10 Chiozza 2011; Powell 2014). This was, for example, a major motivation for Idi Amin, in ordering Ugandan military incursions into Tanzania in the late 1970s an action which could demonstrate strength and occupy and distract a rebellious officer corps (Mambo and Schofield 2007). While initiating an international conflict is not something we would expect a state leader to undertake lightly, in the modern era the odds of state death or some other catastrophic outcome from international conflict are relatively rare (e.g., Fazal 2007) when compared to the relative frequency with which coups depose leaders (e.g., Svolik 2012). Accordingly, it is likely that in states with highly conflictual civilmilitary relations, leaders view the threat from their militaries as more salient that foreign-imposed regime change. Even absent the immediate threat of an outright coup, there are other reasons for both civilian and military leaders to view international conflict more favorably when the military involves itself in non-defense politics. For civilian leaders: High levels of international threat tend to encourage attitudes of professionalism in the military as well as an acceptance of civilian control (Desch 1999). International crises can even lead the military to willingly surrender governmental power to civilian elites in order to focus increasingly on defense as professionally-oriented officers increasingly assert themselves in intra-military debates over the military s appropriate role in the state (White 2017). 6 This is what occurred in Egypt in between the 1967 and 1973 wars (e.g., Cooper 1982; Gamasy 1993; Brooks 2008) and in China in the 1990s as the Chinese People s Liberation Army (PLA) observed the startling military successes of the United States military a likely antagonist in the 1991 Persian Gulf War (Saunders and Kiselyznyck 2010; Finkelstein 2007, 102). International conflict often jumpstarts the professionalization and de-politicization of the military. 6 This may reflect the struggle inherent in military rule where a politically-active military bifurcates into what Stepan terms the military as institution and the military as government (Stepan 1988). 10

11 In addition to empowering professional officers in the military, the prospect of international conflict can benefit junior officers at the expense of the more senior military leadership. The incentives of officers at the beginning of their careers are different from those at the end. Junior officers need opportunities for advancement, while the same is not true for senior officers, who have already reached the top of the military hierarchy (Sechser 2004, 750). Calcification of the military hierarchy and the persistence of a gerontocracy of generals is often a factor in politically active militaries where the lines between politician and military officer are blurred and therefore the same vocational pressures to retire may not exist as in professional militaries. For example, in the 1980s in China, there was an extreme bottleneck in the promotion pipeline in the PLA due to the "political longevity of Long March [i.e., Chinese Civil War] generation" (Paltiel 1995, 784), with one observer noting that "The Politburo looks like a reunion of Army veterans from the historic long march of " 7 In the Chinese case, the prospect of international conflict against technologically sophisticated adversaries compelled the retirement of many of the politically active Long Marchers and contributed to the rise of a younger, much less politically-active generation of officers something welcomed by civilian reformers, such as Deng Xiaoping (e.g., Dreyer 1996; Joffe 1987; Kanwal 2007). 8 Generally, international conflict tends to elevate younger, more professionally-oriented officers at the expense of the older, more politicized faction. Accordingly, civilian leaders are likely to see international conflict as politically attractive in that it should empower younger, less politically-inclined officers in the military at expense of the older, politically-active military leadership. 7 Washington Post. 10/18/1982. "China's Army, Party Keep Uneasy Truce." The author adds that No less than 12 of its 28 members actually participated in the guerilla-war [in the civil war against the Nationalists]...including three marshals in their 80s. 8 International conflict can sometimes offer a pretext for a purge of political rivals in the military, which is unlikely to be opposed by junior officers who stand to benefit from the vacant senior billets. In Syria, Hafez al-assad was able to use the failure of the 1970 invasion of Jordan to justify the purging of a large number of politically-active senior officers, a move welcomed by the middle and junior ranks of the Syrian officer corps (Pollack 2002, Ch. 6). 11

12 In addition to civilian leaders, the military may also find that it has institutional incentives to favor international conflict. Heightened international tensions and conflict increases military resources (Secsher 2004) i.e., the military can anticipate a bigger budget and other resources if the likelihood that it will be needed to defend the state increases. Indeed, some scholars argue that military conflict is a way for civilian leaders to credibly commit to the military a certain level of resources (Arbatli and Arbatli 2014) and in a non-democratic context, interstate war also tends to reduce the onerous, coup-proofing controls sometimes placed on militaries (Talmadge 2013), increasing their autonomy and potential leverage over the civilian side of the government. In the context of civil wars, Acemoglu et al. (2010) suggest that, during conflict, civilian governments may be forced to create excessively large militaries, giving the military a great deal of leverage over the government, in order to assure the military that the government will not seek to challenge military prerogatives after the war. Indeed, largescale military expansion may allow a calcified, gerontocratic military to address promotion bottlenecks thereby relieving tensions with junior officers without having to retire more senior, politically-active officers. This literature suggests that international conflict may benefit the military in a political contest with civilian elites, incentivizing its leaders to favor international conflict. However, if the military stands to benefit from international conflict, wouldn t civilian leaders seek to avoid it, and vice-versa? The answer to this is that in the context of civil-military competition over political power, often both civilian leaders and factions within the military will find reasons to support international conflict, perceiving its likely impact on the civil-military conflict differently. This makes international conflict more likely as it creates a situation where both sides of the civil-military conflict see potential political benefits. A brief discussion of the Sino-Soviet border conflict of early 1969 illustrates this dynamic. 12

13 There is strong evidence that the clashes between Soviet and Chinese troops at the Ussuri River in 1969 were instigated by the Chinese side and that the cause of this was domestic politics. Some scholars suggest that the attack on Soviet border troops at the Ussuri River by the People s Liberation Army (PLA) was ordered by Marshall Lin Biao as a way to demonstrate the PLA s essential role in China in the lead-up to the Ninth Communist Party Congress and consolidate the political power it had accrued during the Cultural Revolution (MacFarquhar 1997, 263). The PLA stood to gain a great deal from heightened tensions with the Soviet Union, which would increase its prestige and power in the factional politics that characterized the Cultural Revolution. Indeed, there is evidence that Mao s overtures to the Nixon Administration in 1971 were motivated in part by a desire to minimize the Soviet threat in order reduce much of the rationale for the PLA s heightened power (MacFarquhar and Schoenals 2006, Ch. 18). However, others suggest that Mao instigated the clashes as a way to regain control of the Cultural Revolution, generating unity among its disparate civilian and military factions and loyalty to him (e.g., Goldstein 2000). Mao was losing control of the Cultural Revolution in 1969, and a major concern of Mao s was the enormous power that the PLA had accrued during the Cultural Revolution as it took over many functions of national and provincial administration from various Red Guard factions (MacFarquhar and Schoenals 2006, Ch. 18). He may have perceived that this conflict, rather than generating power for the PLA, would unify it behind him, or at least, any leverage the PLA gained would be offset by the unity of the population and Red Guards behind him. While historians have not settled on exactly which faction inside China was responsible for instigating the Ussuri River clashes, there is compelling evidence to suggest that both factions within the military and Mao perceived a political benefit to conflict with the Soviet Union. Accordingly, in analysis of the case, there are plausible narratives whereby either Mao or the military was responsible for the Ussuri incident (e.g., Macfarquhar 1997; Goldstein 2000). Another possibility is that Mao may 13

14 have sought a limited conflict or merely the threat of conflict in order to galvanize the public, Red Guards, and military units behind him, but as the conflict worsened and the prospect of actual war became more immediate, it began to benefit the military. It is possible also that at the outset both the military and Mao agreed to instigate the Ussuri river clashes ostensibly for mutual benefit or for the national interest, but with each side perceiving potential political gains at the expense of the other. When the military is a political faction, the decision to use force is inherently politicized. Military and civilian leaders may both perceive political benefit to instigating international conflict. If both the military and civilian leaders had perfect information on the future course of a potential international conflict as well as its political consequences, then there would likely not be increased conflict propensity in response to the military s political participation. This is because one side either the military or civilian leadership seeing how the other side might benefit, would seek to block the initiation of conflict. However, both sides may perceive the likely course of a conflict and its political ramifications differently. Indeed, in conflictual civil-military relations where there is shared political power between civilian and military elites as was the case in China in 1969 information-sharing between the civilian and military branches of government is acutely poor, as both sides have incentives to conceal information from each other (Brooks 2008). The military likely has superior information on its strength and that of an international foe, and the civilians may have superior information on domestic political or economic ramifications of conflict, but neither side has an incentive to share information with the other that might benefit them in civil-military competition. This creates a situation where both the state s civilian and military leaders can anticipate that an international conflict will benefit them and weaken the other side. The especially conflictual civil-military relations that arise with a politically-active military create added incentives for state leaders and their militaries to initiate international conflict. The 14

15 positive externalities of international conflict that leaders may enjoy in their contests with the military may lead them to discount the costs and risks of an international adventure, and overvalue its merits (e.g., Goemans and Chiozza 2011). And in some cases, militaries may find that international conflict is more appealing because of increased resource flows, autonomy, and potential leverage over civilian leaders that may stem from the conflict. Again, this does not mean that either military or civilian leaders conjure up an international conflict solely to increase their leverage over the other side in conflictual civil-military relations; however, it does make conflict more appealing, and makes the government less likely to pursue peaceful means to address extant international disputes such as the long-standing border disagreements from which the 1969 Ussuri River clashes stemmed. International disputes that might have otherwise been resolved peacefully may escalate to violent conflict, because the potential benefits to civilian and military leaders of violent conflict makes conflict in toto less costly. The second mechanism increasing the likelihood of international conflict is the diffusion of decision-making authority, which compounds the politicization of the decision to use force. Brooks (2008) argues that when politicians and military officers share political power, foreign policy decisionmaking authority is contested and unclear i.e., it is not concentrated in the hands of a civilian leader. This can have severe consequences in the lead-up to a potential international conflict. At key moments, military leaders, endowed with decision-making authority, need not wait for civilian orders, they can unilaterally escalate tensions, or even start an international conflict by themselves. Immediately prior to Egypt s 1967 war with Israel, both civilian and military elites shared power in the Egyptian government, and disagreements between President Gamal Abd al-nasser and army chief Field Marshall Hakm Amer led initially to indecision and then contradictory orders. However, the diffusion of ultimate decision-making authority between civilian and military leaders meant that Amer was able to unilaterally escalate tensions with Israel by pre-positioning troops close to the vital Straits of Tiran without authorization from Nasser (Brooks 2008, 89-91). A similar dynamic occurred in China in

16 at the height of Sino-Soviet tensions, when PLA chief Marshall Lin Biao unilaterally ordered the mobilization of almost a million soldiers and thousands of aircraft in preparation for possible war with the Soviet Union. When Mao Zedong heard about the massive mobilization, he quickly rescinded the order, likely fearing unintended escalation with the Soviet Union (MacFarquhar and Schoenals 2006, Ch. 18). The actions of Pakistan s military in the 1999 Kargil conflict provide another example wherein a government with contested civilian-military authority, the military pursued entrepreneurial military action without authorization from or coordination with the civilian arm of the government. Here, in the planning stages, Pakistan s military decision-makers focused myopically on the tactical dimension of their offensive in Kashmir rather than the broader international and regional context in which the conflict would take place. The lack of consultation with civilian leaders meant that the broader diplomatic and political ramifications of an offensive against India were discounted (Brooks 2008). 9 In the implementation stage, at the very least, Pakistan s elected civilian leadership was not fully informed of the scale and conduct of the offensive, with some sources suggesting that the offensive was begun unilaterally without the civilian leadership s knowledge. 10 When the military intrudes into areas of government outside of national defense in civilian regimes it is indicative of the military becoming part of intra-regime politics and is often indicative of high levels of civil-military competition over political power. Both military and civilian leaders are incentivized to use international conflict for political gain. The politicization of the decision to use force is compounded by the shared decision-making authority between the two groups. The decision to use force is no longer solely vested in the civilian leader, and military leaders have the ability to 9 Desch (1999, 6) notes that civilian politicians, diplomats, and technocrats have a more nuanced view of how to approach international affairs their toolbox includes diplomacy, economic measures, and a range of other actions in addition to military action. 10 Nawaz blames Musharraf for Kargil. 5/28/2006. The Times of India. Accessed 4/7/

17 escalate international tensions. For example, the civilian leader may seek a limited demonstration or issuance of rhetorical threats to increase international tensions, but the weak hierarchy engendered by sharing decision-making authority with the military means that the conflict can be escalated without clear authorization from the leader. This leads to a clear hypothesis: H1: As military participation in areas of government outside of defense increases, so too does the likelihood of a state initiating international conflict. Research Design I argue that in civilian-led regimes, military participation in government will increase the likelihood that a state initiates international conflict, but only when the military expands its political role beyond advising on national defense. What is needed is data that captures the number and type of high-level government positions held by military officers in a range of states over time and disaggregates according to whether the military officers have security or non-security responsibilities. The new Military Participation in Government data (MPG) that I have collected fulfills this need. The MPG data is a human-coded dataset where coders consulted the Europa World Yearbook from for yearly rosters of national cabinets, state councils, and other decision-making bodies in the executive. 11 From these rosters, coders counted the number of individuals in the government and the number that were military officers as evidenced by their military rank, whether active-duty or retired. 12 To minimize the amount of subjective decision-making by coders, coders 11 Europa records information once per year for each country and in most cases does not account for changes in government that may occur after the date of publishing. Accordingly, individuals who both take and leave office in between two volumes may not appear in the data. This is an issue common in other country-year data on cabinet composition that uses almanacs or other sources that are published once per year (see, e.g., Arriola and Johnson 2013, Appendix A). 12 In rare country-years where military ranks were omitted, I examined additional secondary sources to address any gaps in the Europa rosters. Retired military rank is indicated in Europa with ret. next to the individual s military rank. This is less objective and consistent than active-duty rank, given that retired in the military context does not merely mean that the individual served in the armed forces. In most countries retired status is granted after a certain number of years in the military (20 in 17

18 counted individuals not portfolios. 13 The state-leader was not considered in the counts so as to properly measure the military's institutional role in national politics and avoid conflating an institutional role for the military with the leader having a military title. 14 Other than the state leader, all listed individuals in the Europa rosters were counted with the exclusion of political party leadership, 15 the legislature, and agency or ministry-heads explicitly noted to not be a part of the government. The coding incorporated a range of executive-level bodies, including cabinets, state councils, "revolutionary command councils," presidiums, and privy councils. Ministers, deputy ministers, state secretaries, and council-members were counted. Vice presidents and deputy prime ministers were also included, as were leadership positions, such as monarch, prime minister, and president if they were not held by state leader. The counts were divided into security- and non-security components. Government-members were counted in either the security or non-security count based on whether any of their listed portfolios pertained to security or the armed forces. For example, if an individual were Deputy Prime Minister and Minister of Defense, s/he would be counted in the security count, because of his/her Defense portfolio. In this way, the coding of the non-security count is relatively conservative in that it only counts individuals whose portfolios have no direct relevance to defense or the armed forces. the United States military) or a certain rank; the specifics of the regulations governing retired status may vary by country, but usually involve high ranks (e.g., generals or admirals). In robustness checks, the analysis is limited to only active-duty officers, who comprise the large majority of military officers represented in the MPG data, with no substantive change to the results. 13 So if an individual were both Minister of Defense and Minister of the Interior, s/he was only counted once. In many cases, the number of portfolios in a particular minister's brief was ambiguous. For example, while "Minister of Defense and Rural Development" clearly entails two separate portfolios; "Minister of Rural Development and Agriculture" could pertain to one or two separate portfolios. Counting individuals, accordingly, presented a much more objective approach that would be consistently applied across coders. 14 The state leader was determined using the Archigos data (Goemans, Gleditsch, and Chiozza 2009). 15 In the coding it was observed that there is substantial overlap between the leadership of the dominant party in one-party regimes and cabinet-level positions in the government, so an accounting of cabinet-level positions should provide a good overall picture of the level of military involvement in government and politics in these regimes. 18

19 Despite this conservative coding rule, incidences of military officers holding non-security positions occur with relatively high frequency in 21.69% of country-years coded (see Table 1). Table 1: Military officers in security and non-security roles (all countries , cell percentages are proportion of total) Military officer in non-security role Military officer in security role No Yes Total No 3, , % 6.15% 60.38% Yes 1,420 1,174 2, % 17.93% 39.62% Total 4,971 1,577 6, % 24.08% In more detail, the security/non-security distinction between government members portfolios was made based on the responsibilities implied by the position's title. The Ministries of Defense, Foreign Affairs, Justice, and Interior were considered to be security-related in addition, so too were positions with security in the title excluding Social Security and Food Security and also positions related to borders, counter-narcotics, military branches, the policy-gendarmerie, or military industries. Cabinet-level positions related to the administration of specific international or domestic territories e.g., Minister of Chittagong Hill Affairs (Bangladesh) or Minister of Mainland Affairs (Taiwan) were also considered security-related positions. 16 Other positions counted in the coding of security/armed forces positions included positions related to nuclear energy, secondary foreign affairs ministries that did not explicitly pertain to economic matters, development, international aid, and positions pertaining to civil defense or public order. All positions that did not meet the criteria for security-related were 16 This does not include provincial governorships rather it captures positions in the national government related to a specific, usually strategic or conflict-prone territory. 19

20 considered to be non-security-related. Examples in the data of non-security positions held by military officers include various economic ministries, such as the "Minister of Civil Aviation" (e.g., Egypt, ) and ministries related to social services such as the Minister of Health (e.g., Sierra Leone, 1967), as well as, in many cases, positions that relate to the general political leadership of the country, such as the royal Privy Council members in Thailand ( ). Figure 1: Proportion of states with military officers in government, security and non-security roles (all countries, ) Figure 1 shows the proportion of states that have military officers in cabinets and state council equivalents for each year from , differentiating by whether there were officers in a security or non-security role. Two things are immediately clear. Military involvement in the security sector for military officers is always more common than in non-security areas. The gap between the percentage of states with military officers in cabinet-level security roles and those with military officers in non- 20

21 security roles is always at least seven percentage points, and is often more than ten percentage points. Also, it is clear that a greater share of states in the international system see military involvement in government during the Cold War. Military involvement in the security sector of government reaches its highest point in 1989 with more than 50% of states having a military officer in a cabinet-equivalent security role in the national government. The peak in military involvement in non-security areas of government was reached in 1978, when more than 37% of states had a military officer in a cabinetequivalent non-security role. This likely stems from the relative frequency of military regimes during that period. For both types of military involvement in government, there was a steep drop off with the end of the Cold War. Some of this drop-off can likely be attributed to the proliferation of new regimes with the break-up of the Soviet Union many of which began with civilianized governments. It is notable, however, that decline in military involvement in government at the end of the Cold War has not continued in the post-cold War period, with non-security involvement never dropping below 12% of states and security involvement never below 29% of states. To examine the effect of military involvement in government on conflict initiation, I use as independent variables counts of the number of military officers in government, aggregating all military officers in government (MPG) and differentiating by role i.e., security (Security MPG) vs. non-security (Non-security MPG). For the dependent variable, I generated an indicator for whether or not a state initiated a militarized interstate dispute (MID) in a given year using the MID 4.0 data from the Correlates of War Project (Palmer et al 2015). I also generated an indicator for whether or not a state initiated a MID that resulted in fatalities in a given year to examine whether any effect for Security MPG or Non-security MPG extends to MIDs that involve at least some interstate violence (as opposed to just threats). One of the mechanisms discussed in the theory is the diffusion of decision-making authority to military leaders. Given the unique ability of military leaders to mobilize armed forces and initiate armed clashes, the positive effect for Non-security MPG should be strong for those disputes that 21

22 involve at least some violence. I used both of these indicators to examine international conflict initiation at the country-year level of analysis (with directed-dyads examined in the online appendix). Critically, the theory argues that MPG has an effect on conflict initiation outside of regime type. To ensure that the analyses do not conflate any effect for military involvement in government with overall regime characteristics, I include Weeks' (2012) autocratic regime-type indicators for "boss" (civilian personalist dictators) and "machine" (single-party systems). "Strongman" (military personalist dictators) and "junta" (institutional military systems) regimes are excluded from the main analysis, given that the scope of the theory is civilian-led regimes. I include also Weeks' indicator for other non-democratic regime-types, so that the base category is democracies. Effectively, this approach gives each civilian-led regime type its own intercept and captures the effect of within-regime variation in military participation in government on conflict initiation. Figure 2 shows the averages of the different MPG counts across Weeks regime typology including also military regime types for comparison. Here, it is evident that juntas and strongmen regimes are very similar in their (high) levels of military involvement in government, as are boss regimes, machines, and other non-democracies, while democracies have substantially lower levels of military involvement than all non-democratic regime-types. 22

23 Figure 2: Mean MPG ( ) by regime-type MPG averages by regime-type MPG Security MPG Non-security MPG It is important to also separate the effect of MPG from the characteristics of the state leader, which may make particular leaders more or less likely to bring military officers into the government. Horowitz, Ellis, and Stam (2015) have examined extensively the role that leader military experience plays in the decision to use force in international relations, including in civilian-led regimes. They find that former soldiers without combat experience and former rebels are particularly likely to initiate conflict. Accordingly, I include indicators for whether the leader was a former soldier or officer with combat experience, without combat experience, or a former rebel with the base category being no rebel or military experience. Importantly, while HSE conduct leader-year analysis, I am using the country-year, so I aggregate HSE's leader-year data to this level of analysis. Essentially, this means that the three HSE indicators for leader military experience capture whether or not a country had a leader with those attributes in a given year. In years where there were one or more leadership changes, these indicators will account for the background of multiple leaders. 23

24 Other aspects of the model specification build on Weeks' analyses of international conflict initiation. In addition to state capability controls, such as the state's value in the Composite Index of National Capabilities (CINC score, Singer, Bremer, and Stuckey 1972) and major power status, I include Weeks' indicator for whether a state has had "substantial domestic institutional change within the past three years" (Weeks 2012, 338), which controls for the conflict-proneness of new and transitioning regimes (e.g., Colgan 2010; Mansfield and Snyder 2007). Given the demonstrated connection between intra- and interstate armed conflict (e.g., Gleditsch, Salehyan, and Schultz 2009), I included also an indicator coded as one if the state was experiencing an internal armed conflict within its own borders that resulted in at least 25 battle-deaths in that year using the UCDP/PRIO Armed Conflict Dataset (Themner and Wallensteen 2014). Another critical issue is to ensure that any relationship found between military involvement in government and international conflict propensity is not driven by a dynamic whereby states with histories of conflict anticipate future conflict and accordingly see militarizing of their governments in advance of future conflict. Accordingly, I control for state's history of conflict in two ways. The first is the inclusion of a count of the total number of MIDs that a state has experienced since The second, capturing how recent the last initiated conflict was, is the inclusion of a cubic polynomial of a count of the years since the last MID initiation by the state (or 1946, whichever is more recent). This has the added benefit of addressing temporal dependence with a binary dependent variable (Carter and Signorino 2010). 24

25 Table 2: Descriptive Statistics Variable Mean Std. Dev Min Max Obs. MID initiation ,793 Fatal MID initiation ,793 MPG ,200 Security MPG ,199 Non-security MPG ,200 Machine regime ,793 Boss regime ,793 Other non-democracy ,018 New/unstable regime ,220 Leader combat experience ,566 Leader non-combat experience ,564 Leader rebel experience ,570 Major power ,793 Civil war ,497 CINC < ,659 Cumulative MIDs ,793 In all models, the right-hand side variables were lagged by one year to address potential reverse causality i.e., to address concerns that the militarization of government may occur in response to conflict. Robust standard errors were calculated by clustering on country. Given the binary nature of the dependent variable, I used logistic regression. In the following analyses of all civilian-led regimes from , I examine separately the impact of the independent variables on all MIDs as well as those that resulted in at least some fatalities. I first examine the effect of overall MPG, and then disaggregate into its security and non-security component. Results Table 3 shows the results from country-year analysis of MID initiation, where the independent variable is the aggregated count of military participation in government (MPG), collapsing both security- and 25

26 non-security military involvement. The analysis examines separately MPG's impact on the initiation of MIDs and also MID's that result in fatalities. MPG is not significant in the MIDs model, but is highly significant and positive throughout the models where the dependent variable is whether or not the state initiated a MID that resulted in fatalities. This suggests that where military involvement in government influences conflict initiation, it is generally in those conflicts that ultimately result in actual violence. This is in line with the logic of the theory, where the diffusion of command authority should have a particularly strong effect on those disputes that actually involve the use of military force as opposed to those that escalate to only threats, maneuvers, or other "sabre-rattling." Table 4 shows results for the same analysis, but where MPG is disaggregated into its security and non-security components. The results here are striking and demonstrate that the positive effect for MPG on the initiation of MIDs with fatalities shown in Table 3 is driven primarily by military involvement in the non-security aspects of government, as anticipated by the theory. In the fatal MIDs analysis, Non-security MPG is positive and significant at at least the.01 level in all models, while Security MPG is negative and not significant. When MIDs that result in fatalities are examined, the results from the country-year analysis demonstrate that MPG is positively associated with conflict initiation, but this effect is driven by non-security participation in government. In line with the theory, this demonstrates that in civilian-led regimes, where military officers hold only positions related to national security, there is not a significant increase in conflict propensity. Only when the military s role in government expands beyond defense, is there a significant effect. 26

27 Table 3: Military participation in government and international conflict initiation (1) (2) (3) (4) (5) (6) Basic Regime & Leader Full Model Fatal - Basic Fatal - Regime & Leader Fatal - Full Model MPG * 0.083* 0.108* (0.028) (0.031) (0.029) (0.040) (0.041) (0.045) Machine regime * * (0.215) (0.170) (0.246) (0.341) Boss regime (0.200) (0.177) (0.252) (0.265) Other nondemocracy New unstable regime (0.190) (0.162) (0.208) (0.254) (0.149) (0.172) (0.232) (0.279) Leader combat (0.164) (0.155) (0.205) (0.232) Leader non-combat (0.204) (0.138) (0.249) (0.267) Leader rebel * (0.166) (0.129) (0.190) (0.204) Major power (0.150) (0.191) Civil war 0.350* 0.467* (0.143) (0.206) Cumulative MIDs 0.051** 0.040** (0.010) (0.011) CINC 8.059** (2.254) (1.966) Peace-years ** ** ** (0.035) (0.036) (0.029) (0.045) (0.049) (0.056) Constant ** ** ** ** ** ** (0.144) (0.162) (0.134) (0.165) (0.201) (0.336) Observations 4,200 3,894 3,894 4,200 3,894 3,894 Robust standard errors in parentheses ** p<0.01, * p<0.05, + p<0.1 (two-tailed) (Polynomial of peace-years not shown) 27

28 Table 4: Military security- and non-security participation in government and international conflict initiation. (1) (2) (3) (4) (5) (6) Basic Regime & Leader Full Model Fatal - Basic Fatal - Regime & Leader Fatal - Full Model Security MPG (0.067) (0.067) (0.070) (0.095) (0.097) (0.113) Non-security MPG ** 0.150** 0.184** (0.043) (0.044) (0.038) (0.053) (0.046) (0.043) Machine regime * (0.204) (0.168) (0.253) (0.360) Boss regime (0.195) (0.178) (0.255) (0.279) Other nondemocracy New unstable regime (0.186) (0.159) (0.211) (0.267) (0.149) (0.171) (0.228) (0.272) Leader combat (0.165) (0.154) (0.206) (0.234) Leader non-combat (0.204) (0.133) (0.255) (0.272) Leader rebel * (0.166) (0.129) (0.193) (0.202) Major power (0.156) (0.183) Civil war 0.359* 0.506* (0.143) (0.199) Cumulative MIDs 0.051** 0.041** (0.010) (0.012) CINC 8.206** (2.224) (1.893) Peace-years ** ** ** (0.035) (0.036) (0.029) (0.045) (0.049) (0.055) Constant ** ** ** ** ** ** (0.146) (0.163) (0.134) (0.169) (0.195) (0.335) Observations 4,199 3,893 3,431 4,199 3,893 3,431 Robust standard errors in parentheses ** p<0.01, * p<0.05, + p<0.1 (two-tailed) (Polynomial of peace-years not shown) While the logit coefficients indicate the significance and direction of the effect, they do not give an easily interpreted sense for the substantive impact of MPG on the likelihood that a state initiates international conflict in a given year. Figures 3 plots the predicted probabilities of a civilian-led state initiating a MID that results in fatalities in a given year at different levels of Non-security MPG 28

29 generated from the full model specifications shown in Table 4. This demonstrates large increases in the probability of fatal MID initiation as the level of non-security military involvement in government increases. When there are no military officers present in a non-security function in the government of a civilian-led regime, the predicted probability of fatal MID initiation is 3.79%. The addition of one military officer increases it to 4.46% (a 17.68% increase); with two, it increases to 5.25% (a 38.52% increase). When there are five military officers, the probability of fatal MID initiation more than doubles to 8.37%. Figure 3: Substantive effect of Non-security MPG on fatal MID initiation in civilian regimes Taken together, these analyses suggest strong support for the theory. Military participation in government has a strong, positive effect on the initiation of fatal MIDs in civilian-led regimes, but this effect is driven primarily by military officers taking non-security roles in government, indicating that 29

30 the military has some political power outside of an advisory role in defense and security matters. In contrast, military participation in government in a security capacity has no independent effect at conventional levels of statistical significance. This does not mean that military ministers of defense or national security advisers never exert a systematic positive effect on conflict initiation, but rather suggests that when they do, it is in regimes where the military has expanded its role in the government beyond advising on defense matters. Additional Analyses To provide added confidence in the main findings, I explored a range of additional control variables and model specifications. In all robustness checks, I focused on fatal MID initiation as the dependent variable. First, given the relative rarity of fatal MID initiation in the data, I reran the analysis using rare events logit (King and Zeng 2001). In this specification, none of the findings for the MPG counts change substantively, and indeed, in most cases achieve greater statistical significance. Second, as Figure 1 indicates, there are clear system-wide time-trends in military participation in government. In order to ensure that these temporal trends were not an omitted factor that biased the results, I added yearly ( ) fixed effects to the analyses, which did not substantively change the main findings. Replacing the Weeks autocratic regime typology with similar indicators derived from the Geddes, Wright, and Frantz (2014) data also do not substantively change the result. 17 Since coup risk may also motivate state leaders to pursue aggressive foreign policies (e.g., Goemans and Chiozza 2011), and coup risk is likely associated with higher levels of military involvement in government, I include a counter for the number of years since the state last experienced a coup attempt. This was 17 Here, the Party and Personal (civilian) GWF indicators replaced Weeks Machine and Boss indicators, respectively. Weeks Other Nondemocracies indicator was replaced with one coded as 1 for monarchies and all other non-democratic regimes in the GWF data. In all specifications, GWF s coding of democracy was the base category. Weeks indicator for New/Unstable Regime was replaced with an indicator coded as 1 if the regime was no more than five years-old in the GWF data. 30

31 derived from Powell and Thyne s (2011) data. This also did not substantively alter the main findings. Limiting the MPG counts to only active-duty military officers also did not substantively change the results either. This, and the relative rarity of retired officers in the data, 18 suggests that the trends observed in the main analyses are driven primarily by the participation of active-duty military officers in the government. In order to ensure that the results were not driven by extreme values on the MPG counts, I also generated the natural log of all MPG counts and reran the analyses with these measures; this also did not change the main findings. To ensure that the effect of the MPG counts was not merely picking up the effect of overall cabinet/state council size and to better capture the military s share of political power, I reran the analyses with proportions that captured the share of high-level government positions held by military officers rather than the raw counts. Figure 4 show the impact of the proportion of non-security government positions held by military officers on conflict initiation in civilian-led regimes which shows a similar trend to that shown when the independent variable is the count of military officers (e.g., Figure 3). 18 As discussed in the research design, retired does not merely capture former military experience, it captures an individual having spent enough time in the military to have achieved sufficient rank and or enough years of service to qualify to officially retire rather than merely be discharged after a term of service. 31

32 Figure 4: Substantive effect of military s share of non-security positions on fatal MID initiation in civilian regimes And while the scope of the theory in this article is civilian-led regimes, I also reran the analyses, but no longer restricted the sample to civilian-led regime i.e., I added to the sample military regimes, both juntas and military-personalist dictatorships. The results of these analyses do not present major differences from those in the civilian regimes analyses though there is a slight reduction the magnitude of the effect for MPG and Non-security MPG. Another potential concern in the analyses is endogeneity. Specifically, the anticipation of international conflict by the government may both contribute to a state s propensity to actually initiate conflict and also to bring more military officers into the government. This should not be a major concern in the analyses here for several reasons: 1) Existing literature on coups (e.g., Talmadge and Piplani 2015) and on civilian control of the military more generally (e.g., Desch 1999; White 2017) has demonstrated that the effect of international conflict on military involvement in politics is negative the opposite of that presented here. Given this, any potential bias should set a higher bar for finding 32