VETO PLAYERS AND MILITARIZED INTERSTATE CONFLICT

Transcription

1 The Pennsylvania State University The Graduate School College of the Liberal Arts VETO PLAYERS AND MILITARIZED INTERSTATE CONFLICT A Dissertation in Political Science by Jeremy E. Lloyd c 2014 Jeremy E. Lloyd Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy August 2014

2 This dissertation of Jeremy E. Lloyd was reviewed and approved by the following: Glenn Palmer Professor of Political Science Dissertation Adviser Chair of Committee D. Scott Bennett Distinguished Professor of Political Science Department Head Scott Gartner Professor of International Affairs Christopher Zorn Liberal Arts Research Professor of Political Science and Sociology James Piazza Associate Professor of Political Science Director of Graduate Studies Signatures are on file in the Graduate School. ii

3 Abstract My dissertation seeks to clarify the relationship between domestic institutions and interstate conflict by examining the link between veto players and interstate conflict initiation, duration, and outcome. All else equal, states with higher levels of policy stability should, arguably, find it harder to initiate/militarize interstate conflicts as the decision to militarize often reflects a definable change in the status quo policy position (see Baumgartner & Jones 2010, Allison & Zelikow 1999, Halperin & Clapp 2006). Once a decision has been made to militarize, the interjection of veto players policy should make it more difficult to reach negotiated settlements, which subsequently should increase the duration of wars fought by states with higher levels of veto player constraint. Finally, the presence of veto empowered independent actors should increase the likelihood of victory by improving a state s ability to accurately select vulnerable targets. Consistent with the underlying theory, the findings indicate that while veto player based constraint is negatively correlated with the likelihood of states deciding to take militarized action, veto player based constraint is positively correlated with the likelihood of a state achieving a favorable outcome. However, contrary to theoretical expectations, veto player constraint is negatively correlated with war duration. iii

4 Table of Contents List of Tables vii 1 Introduction Interstate Conflict Veto Player Framework Veto Power as Legislative Constraint Veto Players and Conflict Initiation Veto Players and Conflict Duration Veto Players and Conflict Outcomes Initiation Introduction Veto Players and Conflict Initiation The Predictors of Conflict Regime Type and Conflict Initiation Veto Players and Conflict Initiation Veto Power and Legislative Constraint Research Design Directed dyads vs. Non-directional dyads Temporal Dependence Dyad Heterogeneity Militarized Action Veto Players Modeled as Legislative Constraints Control Variables Results and Discussion Conclusion iv

5 2.6 Appendix: Tables for Chapter Duration Introduction Theoretical Framework Interstate War Termination: Process and Inefficiencies Regime Type and War Duration War Termination Bargaining as a Two-Level Games Veto Players and War Termination Bargaining Veto Power and Legislative Constraint Population of Cases War Duration Legislative Constraint Control Variables Results and Discussion Conclusion Appendix: Tables for Chapter Outcome Introduction Theoretical Framework Conflict Outcomes: Initiation and Victory Victory through Selection: Accuracy and Risk Veto Players and Conflict Selection Legislative Constraint and Conflict Selection Research Design Bivariate Probit Model Selection Model Conflict Outcome Veto Players Control Variables Results and Discussion Conclusion Appendix: Tables for Chapter v

6 5 Conclusion Summary Initiation Duration Outcome Significant Implications Electoral Accountability and Conflict War by Committee Democratization and Peace Democratic Victory Non-state actors Extensions Bibliography 172 vi

7 List of Tables 2.1 Summary statistics Multicollinearity Diagnostics Total Number of Militarized Interstate Disputes by Type Bivariate Analysis of Legislative Constraints and Militarized Action, Bivariate Analysis Legislative Constraints and Militarized Action (expansion cases excluded), Marginal Effects of Legislative Constraints on Militarized Action, Bivariate Analysis Multivariate Analysis Legislative Constraints and Militarized Action (All MIDs), Multivariate Analysis Legislative Constraints and Militarized Action (Original Initiators, Expansions Excluded), Multivariate Analysis Legislative Constraints and Militarized Action (All), Multivariate Analysis Legislative Constraints and Militarized Action (All MIDs), (cont.) Multivariate Analysis of Militarized Action, Original Initiators (Expansion cases excluded) Multivariate Analysis of Militarized Action, Original Initiators (Expansion cases excluded) (cont.) Multivariate Analysis of Militarized Action, Democratic States Multivariate Analysis of Militarized Action, Mixed Regimes Multivariate Analysis of Militarized Action, Autocratic States Summary statistics Bivariate Analysis of Legislative Constraints and War Duration, Hazard Model Coefficient Estimates, War Duration Hazard Model Coefficient Estimates, War Duration (cont.) Reduced Hazard Model Coefficient Estimates, War Duration vii

8 4.1 Summary statistics Bivariate Analysis of Legislative Constraints and Conflict, Bivariate Analysis of Legislative Constraints and Conflict (Joining & Reciprocating States included), Multivariate Analysis of Legislative Constraints and Conflict Outcome, Multivariate Analysis of Legislative Constraints and Conflict Outcome (Joining and Reciprocating States Included), Multivariate Analysis of Legislative Constraints and Conflict Outcome by Regime Type, Multivariate Analysis of Legislative Constraints and Conflict Outcome by Regime Type ( Includes Reciprocating States and Joining States), viii

9 Chapter 1 Introduction This dissertation is about militarized interstate conflict. Violent conflict has always been and, perhaps, will always be a integral component of interstate relations. In the absence of hierarchy, conflicting states have limited options for settling their more intractable disputes (see Waltz 1959, Keohane 1986). Violent conflict has arguably served as something akin to a "court of last resort" for conflicting states that were unable to settle their disagreements through the normal course of bargaining (see Clausewitz 1832/1984). Though often costly and far from permanent, post-war settlements tend to be more durable than those reached through arbitration and mediation (see Beardsley 2011). Like many others that preceded it, this dissertation discusses the effects of domestic level factors on the conflict behavior of states. Though an innumerable host of studies, both theoretical and empirical, have been devoted to understanding the factors that govern state conflict behavior, social science and the field of international relations has yet to reach a functional consensus on the factors that determine a state s conflict behavior (see Bennett & Stam 2004). Traditionally, scholars have focused on the role played by systemic factors under the assumption that domestic political factors were largely irrelevant (see Morgenthau 1962, Waltz 1959, Waltz 1979, Mearsheimer 2001). While 1

10 systemic factors undeniably influence state behavior, only domestic-level and dyadic-level factors can adequately account for both the variation in conflict proclivity amongst states within a particular systemic environment, and the variation in the overall level of conflict across functionally similar, yet temporally distinct, international systems. Within the field of international relations, there is a burgeoning consensus that regime type is a primary determinant of conflict behavior (Morgan & Campbell 1991, Maoz & Russett 1993, Bueno de Mesquita & Siverson 1995, Ireland & Gartner 2001, Leblang & Chan 2003, Reiter & Stam 2002, Filson & Werner 2004, Lai & Slater 2006, Pickering & Kisangani 2010, Weeks 2012). While analyses based on regime type have arguably contributed to both our overall understanding of state conflict behavior and improved our ability the predict future empirical patterns of behavior, regime type is a broad categorization that overlooks and in some cases obscures domestic level variation that may affect a state s conflict behavior. The measure is simply too unrefined to account for variations such as political institutions, organizational behavior, and bureaucratic procedures. As a consequence, little is know about the effects of political institutions particularly those that have veto authority over conflict initiation, escalation, and resolution process on a state s conflict behavior. My dissertation seeks to clarify the relationship between domestic institutions and interstate conflict by examining the link between veto players and interstate conflict initiation, duration, and outcome. Veto players are individual or collective actors whose agreement is required for a policy decision (Tsebelis 1995). Tsebelis predicts that increasing the number of relevant veto players, decreasing the congruence of their policy preferences, or increasing the internal cohesion of each veto player increases overall policy stability. I assert that this variation in policy stability should, subsequently, result in variation in the likelihood of state initiating militarized conflict, the subsequent duration of that conflict, and the overall outcome. All else equal, states with higher levels of policy stability should, arguably, find it harder to initiate/militarize interstate conflicts as the decision to militarize often reflects a definable change in the status quo policy position (see Baumgartner & Jones 2010, Allison & Zelikow 1999, Halperin & Clapp 2006). Once a decision 2

11 has been made to militarize, the interjection of veto players policy should make it more difficult to reach negotiated settlements, which subsequently should increase the duration of wars fought by states with higher levels of veto player constraint. Finally, the presence of veto empowered independent actors should increase the likelihood of victory by improving a state s ability to accurately select vulnerable targets. 1.1 Interstate Conflict Understanding the relationship between veto players and conflict behavior begins with an understanding of interstate conflict. Violent interstate conflicts arises from the presence of an underlying state of discord between two or more states (Keohane 2005). In an anarchic and competitive interstate system, states achieve and maintain internal stability and prosperity as a result of the domestic and foreign policy actions executed by their governments. When a state s policy goals interfere with the policy goals of another state, this interference creates a latent potential for conflict; the more salient and discordant the policy goals, the greater the potential. 1 Latent conflicts become active when constraints and possibilities align in ways that either allow, entice, or force states with discordant policy goals to interact: internal political pressures may force states to actively pursue a discordant policy goal that, in turn, elicits a policy response from the affected state; emergent technology may create a shift in the balance of power that allows a previously disadvantaged state to overcome previously insurmountable obstacles; troops positioned in some strategic location may interact in ways that reignite a latent, yet unresolved dispute. If neither state is willing or able to adjust or defer its policy goals, war may ensue. However, war should be viewed as a process rather than a singular event (Schelling 1980, Bremer 1992). Wars are simply interstate conflicts that, after 1 While an underlying state of discord (either real or perceived) is necessary for conflict, the existence of discord does not always result in active conflict. Time and resource limitations may prevent states from actively pursuing a particular policy; when that policy goal is discordant, states may be unable to pursue a particular policy goal due to the affected state s resistance. 3

12 militarization, 2 reached a certain level of severity 3 before they were eventually resolved. In this framework, discord is a necessary precursor for conflict and militarization is a necessary precursor for violent interstate conflict, whether they be wars, or conflicts of lesser severity. If states truly pursue violent conflict as a means of achieving political goals, then it stands to reason that conflict decisions are governed by a process that is both competitive and political. The notion that foreign policy decisions are largely immune to partisan politics (see Wildavsky 1966), though popular amongst modern political pundits, has been largely invalidated by evidence indicating the presence of coherent, stable, ideologically-driven partisan divisions both amongst political elites and within the public at large (see Holsti 2009, Wittkopf 1990). In addition to being political, the decision making process is likely to be institutionalized. Whether formal or informal, states choose to institutionalize their authorization out of necessity. First and foremost, a well designed authorization process helps government leaders retain control over their military forces. The hierarchical structure of military forces centralizes command and control authority in the hands of a few functionally unitary leaders. With this concentration of authority comes the constant threat of coup, particularly when the state possess substantial military capability. 4 When properly implemented, an institutionalized authorization process helps mitigate the threat of coup by making military commanders dependent on civilian leaders for their command authority. Whether by controlling the military budget, controlling the military appointment process, 5 or other means, institutionalization effectively interjects civilian leaders into the military chain of command and thereby limits the ability of military leaders to issue legitimate orders without civilian approval. When properly institutionalized, authorization processes are difficult to circumvent. The more 2 In conflict, militarization occurs when one state directs a threat, display, or use of force (short of war) towards the government, official representatives, official forces, property, or territory of another state (Jones, Bremer & Singer 1996). 3 International relations scholars use the term war to differentiate high intensity interstate conflicts from those of lesser severity (see Singer & Small 1994). 4 Military training conditions operators to unquestioningly follow the legal orders of their commanders, who, likewise, follow the commands of their appointed leaders. 5 If officers derive their authority to command from appointment rather than a personal connection to their subordinates, their ability to command becomes dependent on their ability to retain the favor of the appointment authority. 4

13 transparent and respected the authorization process, the better its ability to control the authorized use of military power. Institutionalization also helps government leaders maintain more precise control of their military forces when they are engaged in combat. While the finality of total warfare allows military leaders to pursue broad objectives with high degrees of autonomy, low-scale, limited conflicts require more precise interaction between government and military leaders. In the absence of such interaction, military leaders may pursue actions on the battlefield that inadvertently derail the negotiation process. However, maintaining effective command and control of fielded military forces is a difficult operation made only slightly more manageable by the modern advances in communication technology. A byproduct of institutionalized authorization procedures are the creation regularized, verifiable channels of communication that ensure the orders issued to field commanders are verified. While the specifics vary from state to state, all states have some developed institutionalized process for authorizing and directing the use of its armed forces. The institutionalization of the force authorization process should affect all phases of conflict from initiation to outcome. Given that militarization is a prerequisite for violent conflict, the likelihood of a state initiating militarized action is a function of the shape of the force authorization process. In that same vein, the subsequent intensity, duration, and outcome of that conflict is also governed by the institutional constraints placed upon the commander-in-chief of the armed forces. Commanders who share their authority over the armed forces with other independent, equally empowered agents face an entirely different conflict environments than similarly provisioned yet unconstrained leaders. In addition to facing obstacles to their desire to militarize, these leaders may also have to account for the tactical choices they make during the conflict. While unconstrained leaders are given a blank check, more constrained leaders will be expected to account for the resources they expend; expanding the conflict may require additional authorization. Similarly, the constraints placed on a leader s ability to initiate militarized action also affects that leader s ability to end conflicts via negotiated settlement. Here, the terms of the settlement are bound to be subjected to a ratification process (see Putnam 1988). 5

14 States decide to start and end conflicts particularly those that are limited in scope using institutionalized decision making processes. The veto player argument presented here extends the notion of war as a continuation of policy by incorporating the mechanics of policy change into the analysis of conflict behavior. Veto players are individual or collective actors whose agreement is required for a policy decision (Tsebelis 1995). Increasing the number of veto players, decreasing the congruence of their policy preferences, or increasing the internal cohesion of each veto player increases policy stability. 1.2 Veto Player Framework The foundation of the veto player framework is the notion of status quo win-sets (Tsebelis 1995, Tsebelis 2011). He defines veto players as individual or collective decision-makers whose agreement is required to change the status quo. Once a status quo has been established, changing that status quo requires the cooperation of all the relevant players. Player preferences for new policies are conceptualized exclusively in terms of the ideological "distance" between the proposed policy and the players ideal point; players prefer proposals that are ideologically closer to their ideal than the status quo. The win set is the set of policy options that are closer to the relevant players respective ideal points than the status quo and thus are outcomes that can potentially beat the status quo. The size of this win-set approximates the likelihood of policy change within a state. Larger win-sets imply more viable policy options and thus higher levels of policy mutability. Conceptualizing policy stability in terms of veto-players results in three propositions. Tsebelis predicts that increasing the number of veto players, decreasing their congruence (the distance between the veto players preferences), and/or increasing their cohesion (the level of internal differences within each veto player) increases policy stability and thus decreases the executive s ability to implement policy changes. The number of veto players in a political system is a function of the institutional structure. With each additional actor, the likelihood of their preferences converging 6

15 decreases. As Tsebelis (1995, 2011) notes, the size of the win-set of the status quo of N +1 players is a subset of the win-set of the status quo of N players and thus increasing the number of veto players cannot increase the size of the win-set, and thus cannot increase the likelihood of policy change. Decreasing congruence the commonality in player preferences increases policy stability by decreasing the size of the win-set. As the preferences of the veto players become more and more dissimilar, the likelihood of their positions on a particular policy issue aligning decreases. Cohesiveness comes into play because veto players are often collective actors. Cohesiveness is a measure of how well the veto player s ideal point represents the preferences of the majority of actors that make-up the veto player. Changes in ideological distance between the individual actors that make-up the collective veto player, the number of actors that make up the veto player, and/or the institutional structure (primarily defined by the decision rules) alters the range of options alternatives to the status quo that the group would succeed under the group s respective decision rule that the constituent players are able to agree on: this range of acceptable options comprises the veto players "win-circle" (Tsebelis 2011). More cohesive veto players have a smaller win circles and smaller wincircles lead to higher overall policy stability. 6 Another, way of explaining the affects of additional veto players is through the application of probability theory. Imagine a world in which two veto players are working together to make a policy decision: to either accept or change the status quo. If we simply randomize the actor s choices maybe they are both undecided and are awaiting the results of some event in order to make a final decision and require that they both have to choose change in order for change to win, then there is only a 25 percent chance that they will make the same decision and policy will change. If we increase the number of veto players to three, yet maintain the same parameters, their probability of overcoming the status quo and changing policy now reduces to just 12.5 percent. With each increase in the number of veto players the probability of unanimity decreases exponentially. In 6 Tsebelis finds that while decreasing cohesion (increasing the size of the win-circle) of a collective veto player does not always decrease stability (increase the size of the win-set the status quo), increasing cohesion always increases stability 7

16 a simple extension of this model, we can see that the probability of overcoming the status quo is a function of the number independent preferences. If we set up the same three player game, but allow for players to collude e.g. players A and B form a party and agree make the same decision the probability of change is the same as in a two player model: 25 percent. If we complicate the model further by allowing each one of the veto players to be a collective actor who via some pre-decided decision rule (unanimity, majority, etc) will make a unitary veto decision, we see that group cohesion also affects the probability of reaching an agreement. Using the same three player game as before, we will make one of the players a collective actor comprised of five members using a simple majority decision rule. The probability of the collective actor selecting change is 53 percent. This makes the probability of change for the group 14 percent. Modeling increased cohesion in terms of probability of a random draw is somewhat counterintuitive as cohesion is not a measure of the probability of change, but of how well change represent the preferences of the group. Increasing cohesion in this case is equivalent to increasing the number of votes required for change to win the collective players vote. A higher vote requirement means that, for change win, a higher percentage of the group supports it. If we change the decision rule of our five person group to a super-majority of 4, this decreases the probability of change winning to 23 percent, and thus the probability of change passing is now just 6 percent. It should be noted that there are some limitations associated with using a veto player approach to analyze policy outcomes. As noted by Tsebelis, the model "makes predictions about the consequences of the number of actors involved without knowing exact number or locations" (Tsebelis 2011, p.26). In the absence of this information, the theory can only address "the rate of change or how much different institutional settings permit the change of the status quo." It can neither predict the direction of change, nor predict when change will occur. It can only tell us the likelihood of change in a particular state relative to other states that are more or less constrained. Additionally, the veto player framework is not an analysis of the bargaining process by which decision makers agree on policy (Choi 2010). While bargaining clearly affects the likely outcome 8

17 of veto games, this framework is focused on conflict proclivity. Ideally, many of the features of bargaining arrangements that would systematically affect conflict proclivity, such as institutional interdependence, preference similarity, legislative fractionalization, and coalition proclivity, are all captured using the POLCON measure of veto players. 1.3 Veto Power as Legislative Constraint Veto players gain veto power by virtue of their position in the policymaking process. While the details of the policy making process may vary, what defines the veto player is its ability to independently block changes to the status quo policy position. In systems where largely immutable features structure the policymaking process as an all-or-nothing vote, veto power is clearly defined and, subsequently, veto players are more easily identified. For example, in peace negotiations for multi-party civil wars, veto power arises from the ability to independently continue fighting (Cunningham 2006). Thus, identifying veto players is simply a matter of identifying the groups who are cohesive and powerful enough to continue fighting in the event that their goals are not satisfied. However, veto power can also exist in less clearly structured decision making settings where policy changes are not subjected formalized voting. Here, veto power arises not from a player s ability to directly influence the ratification process, but rather their ability to independently influence preceding phases of the policy formulation process and the subsequent policy implementation process. The bureaucratic agents responsible for implementing the new policy can use their administrative discretion and delegated authority to change the eventual effects of policy changes. However, the effectiveness of this soft veto power is largely a function of the bureaucratic agencies expertise and independence. 7 In interstate conflict, veto power arises from the ability to single-handedly stop the threat, show, 7 Agency expertise limits the ability of principals to effectively monitor their agents performance. Independent agents those that are somehow shielded from executive influence are also difficult to regulate 9

18 or use of military force in conflict. As discussed earlier, the final authority on the show and use of military force technically rests with military field commanders as they have direct command and control of troops in combat. However, institutionalization limits the ability of front-line military commanders to issue orders absent the proper authorization from higher ranking officials. Whether this authorization comes from a military dictator with unconstrained war making powers or a chief executive who shares military command authority with other independent government agents is largely determined rules embodied in a country s constitution. In democratic countries, the use of separation of powers as a means of instituting checks and balances on government power often extends to the conflict authorization process. While a nation s head of state typically serves as commander-in-chief of the armed forces, this power is checked in some way by legislative power. For example, in the United States, the President s ability to militarize disputes is mitigated by Congress constitutionally granted war declaration authority, its control over the budget, its ability to create legislation that affects presidential war-making authority 8. In non-democratic systems, heads-of state are notably less constrained in their ability to use military power as the legislatures inability to block executive policymaking carries over into the conflict militarization process. As the preceding the discussion suggest, the relationship between veto players and militarized conflict can be modeled primarily in term of the relationship between the legislative and executive branches of government, irrespective of the regime type or the political system in place. Increasing the number of independent legislative chambers, decreasing the ideological congruence of the legislature and the executive, or increasing the cohesiveness of each independent chamber of the legislature, should increase the stability of the status quo. This increase in stability should affect the likelihood of conflict initiation, the subsequent duration of the conflict, and the final outcome. The preceding discussion on interstate conflict, veto players, and legislative constraint on the executive suggests that the level of political constraint should affect the likelihood of conflict initi- 8 Two prominent examples of the U.S. Congress using its legislative power to ostensibly limit the President s foreign policy activities are the War Powers Resolution of 1973 and the Boland Amendment 10

19 ation, the duration of conflict, and conflict outcomes. 1.4 Veto Players and Conflict Initiation Why are some states less prone to initiate conflict than others? Expanding on previous work (see Choi 2010, Ireland & Gartner 2001), the following chapter examines the relationship between veto power and conflict initiation. I argue that states with higher levels of political constraint should be less likely to initiate militarized interstate disputes. Given that militarized conflicts arise from a latent state of discord between two states, interstate conflict can arise from a combination of purposeful acts and random occurrences. However, in order to introduce military force, leaders in states where the use of military force is constrained must first navigate the political and bureaucratic processes that grant them the authorization to use force. Whether formal or informal, the force authorization process can be modeled using the veto player framework. As the number, ideological distance between veto players, or the internal cohesiveness of collective veto players increases, the likelihood of a state initiating militarized action should decrease due to an increase in policy stability. Using logit analysis techniques, I analyze a data set that includes all states in interaction with the states in the international system for each year from 1817 to The findings suggest that states with higher levels of legislative constraint are less likely to initiate militarized action than their less constrained brethren. 1.5 Veto Players and Conflict Duration Chapter 3 examines the effect of domestic veto players on the duration of interstate conflict. Given that few conflicts end due to the outright military victory, the possibility of a negotiated settlement, by definition, shifts the focus in conflict from violent coercion to the resolution of information asymmetries. The literature suggests that interstate wars end when uncertainty and commit- 11

20 ment fears are reduced enough to allow for the development of an overlapping bargaining space (Tomz 2007, Wittman 1979, Fearon 1995, Reiter 2009, Smith & Stam 2004, Goemans 2000, Stanley & Sawyer 2009). However, expectation updating process may suffer from inefficiencies. Given that war termination agreements can be conceptualized as two-level games international negotiations that are subject to some form domestic ratification domestic politics are a likely source of this inefficiency (Filson & Werner 2002, Filson & Werner 2004, Bennett & Stam 1998). In this framework, the relevant players in the domestic ratification process can be conceptualized as veto players. The main argument is that increasing the number, incongruence, or cohesiveness of the relevant players in the level two ratification process makes it harder for states to negotiate a settlement. Using survival analysis techniques on war-level data, I examine the effects of veto players on interstate war duration. The findings suggest that, counter to theoretical expectations, increasing the number, incongruence, and/or cohesiveness of veto players actually decreases the length of war. This suggest that rather than creating inefficiency, the presence of veto players enhances the information updating process. Granting independent agents veto power over the use of military force potentially increases a state s ability to accurately assess battlefield derived information by countering the tendency for the inefficiencies that arise from "groupthink" (see Allison & Zelikow 1999), culpability (?), and political entrapment (Iklé 1971, Palmer, London & Regan 2004). 1.6 Veto Players and Conflict Outcomes The preceding evidence that increasing veto player constraint increases the duration of war suggests that, rather than creating inefficiencies, veto players potentially improve the efficiency of the conflict negotiation process. The notion that veto players improve the accuracy of the assessment of battle-derived information suggests that veto players may also play a decisive role in outcomes of militarized disputes. Interstate conflicts end in one of four basic ways: a yield, a victory, a com- 12

21 promise, or a stalemate (Ghosn, Palmer & Bremer 2004). A state s ability to achieve a preferred conflict outcomes is a function of its relative military capability, its relative commitment to the issues at stake, its ability to accurately assess these factors when choosing targets, and its ability to effectively utilize their advantages once militarized conflict ensues (Fearon 1994b, Gartner & Siverson 1996, Partell & Palmer 1999, Reiter & Stam 2002). Conflict outcome models assume that states initiate militarized conflict when they feel that they have the advantage: their capabilities and commitment combine to ensure that their adversary will either yield or be defeated. In states where the decision to initiate military conflict is subject to veto player approval, the inclusion of independent agents could improve a state s ability to accurately assess the capability and commitment of adversaries. In addition to diminishing the likelihood of groupthink, introducing additional independent actors increases accuracy of ex ante assessments, and allows players to more accurately assess the level of uncertainty in their estimates. Evidence from chapter 2 suggests that veto constrained actors are less likely to militarize because of the decrease in the probability of preference convergence. Here, I argue that when preferences do converge, they are less likely to be based on inaccurate assumptions. For state s to initiate conflicts based on inaccurate assessments, all veto empowered actors must collectively make an inaccurate assessment: if even one agent recognizes that failure is imminent, they can single-handedly derail force approval process. Increasing the number of veto points, decreasing their ideological congruence, or increasing the cohesiveness of collective veto players decreases the likelihood preferences converging on an inaccurate assessment. In addition to affecting selection, veto player constraint potentially affects execution. Assuming that veto players remain some ability to stop the use of the military even after initiation, this oversight power could improve tactical choices. Using a Heckman selection model, I analyze the the effect of political constraint on conflict outcomes given their effect conflict selection using the directed dyad data analyzed in Chapter 2. Evidence suggests that in addition to decreasing the likelihood of conflict initiation, political constraint increases the likelihood of victory. 13

22 Chapter 2 Initiation 2.1 Introduction Why are some states less prone to initiate conflict than others? Within the field of international conflict, the burgeoning consensus is that political institutions inhibit the ability of democratic leaders to initiate conflict (Morgan & Campbell 1991, Maoz & Russett 1993, Bueno de Mesquita & Siverson 1995, Ireland & Gartner 2001, Leblang & Chan 2003). In concurrence with this observation, many studies focus on the relationship between regime type and conflict proclivity (Reiter & Stam 2002, Filson & Werner 2004, Lai & Slater 2006, Pickering & Kisangani 2010, Weeks 2012) However, we know very little about the variation in the likelihood of conflict initiation among states sharing a common regime type. As noted by Ireland and Gartner (2001), "although great advances have been made in our understanding of the effects of democracy and international forces on conflict behavior, fewer studies have examined the factors that influence the probability of using force within a particular regime type." 1 1 Filson and Werner (2002) admit that domestic political institutions may inhibit a leader s ability to revise her demands. Bennett and Stam (1998) suggest that democratic states select shorter wars to avoid the problem of rapidly eroding public support. 14

23 Using Tsebelis formal theory of veto players, this chapter seeks to expand our understanding of conflict behavior by examining the effects of legislative veto players on the initiation of interstate conflict. As discussed in preceding chapters, increasing the number of veto players, decreasing the congruence of their policy preferences, or increasing the internal cohesion of each veto player increases policy stability by raising the level of political constraint. In this chapter, I argue that states with higher levels of legislative constraint should be less likely to initiate militarized action. While interstate conflicts arise from a mix of choices and circumstances, state policymakers make the deliberate choice to initiate militarized action. However, in order to take military action, state leaders must first successfully navigate the state-mandated authorization process for using military force. If that authorization process involves multiple players, states in which veto players are more numerous, less congruent, or more fractious should be less likely to take military action due to the increase in political constraint that accompanies any of these changes. Using an extant measure of political constraint the POLCON index (Henisz 2002) this study examines the relationship between political constraint and the initiation of militarized action in interstate disputes. Expanding on previous work on the relationship between legislative constraints and conflict initiation (e.g. Choi 2010, Ireland & Gartner 2001, Weeks 2012), I analyze a data set that includes all states in interaction with the states in the international systems for each year from 1817 to The unit of analysis is the directed dyad year. Using logit analysis techniques, this chapter examines the effects of veto players on the initiation of militarized action in interstate disputes. The findings suggest that states with higher levels of legislative constraint are less likely to initiate militarized action than their less constrained brethren. This chapter proceeds in four sections. The first section discusses how incorporating veto players improves existing work on conflict initiation. The second section describes my research design and discusses the difficulties associated with measuring domestic level veto players and then incorporating those measures into an analysis of conflict initiation. The third section presents my results. The final section concludes by presenting the associated implications of this finding. 15

24 2.2 Veto Players and Conflict Initiation Clausewitz (1832/1984) famously opined that war is the continuation of policy by other means, suggesting that the violence of warfare is not an end in itself, but a means by which states attempt to achieve broader political goals. If state states truly pursue violent conflict as a means of achieving political goals, then it stands to reason that states decide to wage war using the same (or similar) internal, decision making processes they follow when making other foreign policy choices. The veto player argument presented here extends the notion of war as a continuation of policy by incorporating the mechanics of policy change into the analysis of conflict initiation. Though there is some consensus regarding the predictors of conflict (see Bremer 1992, Oneal & Russet 1997, Bennett & Stam 2004), none of the widely recognized predictors of conflict adequately address the fact that states decide to start wars particularly those that are limited in scope using institutionalized decision making processes. Veto player theory suggests that the presence of veto players makes it more difficult to make substantive policy changes. Polities which choose to subject their decision to initiate violent conflict to a veto process should find it more difficult to take military action than those that are less constrained, especially when the veto players are more numerous, less ideologically congruent, and more internally cohesive The Predictors of Conflict The analysis presented in this chapter uses the general framework of utility theory (Bueno de Mesquita & Lalman 1992) while incorporating aspects of both the opportunity and willingness model of interstate conflict (Starr 1978) and the bargaining model of interstate conflict (Schelling 1980, Fearon 1995). Violent interstate conflicts arise from the presence of an underlying state of discord between two or more states (Keohane 2005). In an anarchic and competitive interstate system, states achieve 16

25 and maintain internal stability and prosperity as a result of the domestic and foreign policy actions executed by their governments. When a state s policy goals interfere with the policy goals of another state, this interference creates a latent potential for conflict; the more salient and discordant the policy goals, the greater the potential. 2 Latent conflicts become active when constraints and possibilities align in ways that either allow, entice, or force states with discordant policy goals to interact. If neither state is willing or able to adjust or defer its policy goals, war may ensue. However, war should be viewed as a process rather than a singular event (Schelling 1980, Bremer 1992). Wars are simply interstate conflicts that, after the introduction of military force, 3 reached a certain level of severity 4 before they were eventually resolved. In this framework, discord is a necessary precursor for conflict and the initiation of military action is a necessary precursor for violent interstate conflict, whether they be wars, or conflicts of lesser severity. In the opportunity and willingness model, the likelihood of interstate conflict is a function of the state level, dyadic level, and systemic level factors that affect the likelihood of states with discordant foreign policy profiles interacting (see Siverson & Starr 1990, Brochmann, Rød & Gleditsch 2012). Evidence indicates that factors such as geographic proximity (see Bremer 1992, Vasquez 1995, Brochmann, Rød & Gleditsch 2012), military capacity (see Kugler & Lemke 1996, Lemke & Werner 1996), economic interdependence (see Bremer 1992, Oneal & Russet 1997), alliances (see Leeds 2003), regime type (see Maoz & Russett 1993, Bueno de Mesquita & Siverson 1995, Gartzke 1998), and major power status (see Small, Singer & Bennett 1982) all significantly affect the likelihood of conflict. 5 Geographic proximity increases the likelihood of conflicts of interest 2 While an underlying state of discord (either real or perceived) is necessary for conflict, the existence of discord does not always result in active conflict. Time and resource limitations may prevent states from actively pursuing a particular policy; when that policy goal is discordant, states may be unable to pursue a particular policy goal due to the affected state s resistance. 3 In conflict, states initiate military action when they direct a threat, display, or use of force (short of war) towards the government, official representatives, official forces, property, or territory of another state (Jones, Bremer & Singer 1996). 4 International relations scholars use the term war to differentiate high intensity interstate conflicts from those of lesser severity (see Singer, Bremer & Stuckey 1972). 5 The distinction between opportunity and willingness is a subject of some debate within the literature. For example, some argue that geographic proximity affects both opportunity and willingness (see Vasquez 1995, Senese 2005) and 17

26 between states due to shared access to physical area and an unwelcome degree of interdependence. Power preponderance induces peace as it reduces uncertainty regarding the outcome of a military engagement. Major power status is arguably maintained through active foreign policy and thus, major powers are more likely to be involved in disputes. As alliances are indications of "security communities" among their members, they are expected to reduce the likelihood of war between members. In the case of democracy and economic development, though the direction of the effect is clear, the underlying reasons for those effects are less definite. Researchers use opportunity and willingness framework to link environmental and systemic factors to the behavior of decision makers by outlining the necessary conditions for war (Siverson & Starr 1990, Brochmann, Rød & Gleditsch 2012). The model predicts the likelihood of violent conflict by accounting for the primary the factors considered by decision makers consider when they decide to wage war. However, missing from this framework are factors that adequately account for the effects of variations in the bureaucratic decision making processes of states. Wars are not waged by individuals, but by conglomerates of organizations woven together by intentionally undefined statutes and broad ideological loyalties (Allison & Zelikow 1999). Even the leader of the most despotic, militaristic, autocratic regime must rely on an army of subordinates to effectively wage war. Fractionalized power coupled with split loyalties introduces the potential for political and ideological competition in the decision making process (see Weeks 2012). If the decision to take militarized action is made using an institutionalized, competitive process, e.g. a constitutionally mandated declaration of war, this process is likely to affect the general likelihood of militarized conflict. alliances affect willingness (see Brochmann, Rød & Gleditsch 2012) rather than opportunity. 18