EXPLORING THE SECOND WAVE OF DEMOCRATIC PEACE CRITIQUES: COMMONALITY, TERRITORIALITY, AND SELECTIVITY By Johann Park A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Political Science 2011 ABSTRACT EXPLORING THE SECOND WAVE OF DEMOCRATIC PEACE CRITIQUES: COMMONALITY, TERRITORIALITY, AND SELECTIVITY By Johann Park This dissertation consists of three related but distinct essays on the second-wave debates on the democratic peace. I explore three important caveats remaining in the extensive democratic peace literature: the potential confounding effect of the Cold War, the lack of contentious issues between democracies to manage peacefully, and fallible democracies only choosing weak enemies in crises. The first essay seeks to help resolve the ongoing debate concerning whether the democratic peace was purely a Cold War (CW) Phenomenon. Realists have attributed the democratic peace to common interests induced by the CW, but so far nothing has been directly tested against this claim. I use more direct measures of CW preferences to test if the effect of joint democracy is rendered statistically insignificant by introducing these measures. Conversely, liberals have suggested that the pacifying effect of democracy would be strengthened over time. I also test this claim by incorporating newly available post-CW data. Against realist expectations, the results show that joint democracy promotes peace independently of the CW security interests. Furthermore, the effect for democracy is estimated to be larger in the post-CW era although the difference between the consistently negative effects across the two periods is not itself significant in the main analysis. But all of the alternative analyses for robustness checks produce the significant results. In the second essay, I probe the possibility that the effects of territory and democracy are contingent upon each other. While the contingent effect is theoretically plausible, no study looks directly at the interaction between these two variables in affecting armed conflict. I conduct two kinds of statistical analyses. One is a replication analysis of Gibler’s (2007) work on stable border peace, where Gibler argues that democracy and peace are both symptoms, not causes, of the removal of unstable borders from the agenda of issues between neighbors. Questioning both his theory and empirics, I show that Gibler’s results are not replicable. Instead, I find that even controlling for border variables, joint democracy continues to be an important factor in reducing armed conflict, while Gibler’s border variables lack robust independent effects. The other is based on MID data and the Huth and Allee’s territorial claim list. The results from the interactive models are consistent with the conclusion that the pacifying effect of democracy holds for both territorial dyads and nonterritorial ones, and the effect gets stronger in the territorial context in spite of the imperatives toward militarization created by territorial conflict. The third essay studies how democracies select their foes in interstate military conflicts. Focusing on war outcomes, existing studies provide only indirect tests. I directly look at factors that effect selective initiation of conflict in the first stage of military conflict. Reframing democratic selection of conflict in terms of winnability and justifiability, I theorize that democratic accountability motivates leaders to choose armed conflict with justifiable causes as well as high probability of victory. The argument is tested by considering target countries’ relative military capabilities, geographical constraints, comparative human rights conduct and level of democracy, and commercial relations. The results show that democracies tend to attack weaker foes to a greater degree than do autocracies, countries with bad human rights and undemocratic practices, and those without important economic relations. Copyright by Johann Park 2011 All rights reserved. To my lovely daughter Juwon v PREFACE My dissertation is about democracy, war, and peace. The first time I heard about democratic or Kantian peace was three days after I came first to the States for my graduate study at the University of Missouri-Columbia in 2001. One of my senior Korean colleagues in the Political Science department, Hanseung Cho, was helping me buy necessities and furniture for settlement before I got my driver’s license. He, my wife, and I were sitting for a fast meal at the McDonald’s on Stadium Blvd. near Columbia Mall. Hanseung, an IR major, asked, “Have you ever heard about Kant’s democratic peace thesis?” Then, I never thought it would be my dissertation topic. At Mizzou, I learned about the democratic peace from both a critic (Patrick James) and a proponent (Joseph Hewitt). My first impression of the democratic peace was negative. For me as an international student from the Korean Peninsula, the area probably affected most by the 19th and 20th great power politics, it sounded like a disguised phrase of American imperialism. I had been very skeptical for the idea for a while. My first academic publication with Pat James and Whan Choi was a critical analysis of the democratic peace. Unintentional things in the 2004-2005 school year had made me a Michigan State Spartan from a Mizzou Tiger. I learned about the democratic peace again from Michael Colaresi. He directed me into more fundamental and in-depth readings about the subject. Also, Steve Kautz’s political philosophy course helped me acquire philosophical foundations of liberal ideas of international relations. This training and my own scientific probes have changed my view on liberalism and its ideas on international peace. vi With respect to human nature, I am a realist rather than an idealist or a liberal. Human beings are inherently selfish and egocentric. My skeptical view on human beings culminates when it comes to politicians. Political leaders and elites are ambitious, clever, and strategic much more than the rank and file. They possess desires for political influence, mental and physical aptitudes for political positions, and tactics to achieve and maintain political power weathering intense competitions. They are bold and brazen enough to use their political power for their political goals under the disguise of national interests and to make political decision that would influence a great number of lives. Indeed, the human history is fraught with ambitious, selforiented, and brutal political leaders. With respect to political institutions and foreign policy, I am a liberal more than a realist. Ambitions must be counterchecked by ambitions if ambitions cannot be removed. And this should be done through fair and representative institutions. Only institutionalized competitions, counter-balancing mechanisms, and empowered public evaluation can ensure peaceful transitions of power, political security of people, promotion of public goods, enhancement of national interests, and prevention of tyrannies. The beauty of the democratic peace theory is that we do not need to assume that leaders and citizens in democracies are more peace-loving and otheroriented than those in autocracies. Still, national interests and public goods such as peace, development, victory, international reputations, etc. can be conceived of by selfish democratic leaders and citizens. Democratic institutions allow political competitions to provide policies that can satisfy encompassing interests of divergent constituents because politicians and their parties need 50% of the public votes plus one for political power. Also, democratic leaders must interplay with empowered political rivals, actively monitoring free media, vociferous independent experts, directly affected stakeholders, and enthusiastic foreign policy zealots. The vii different way-outs of democratic politics should make different implications for foreign policy and international relations. Democratic leaders should be more deliberate and selective than autocratic leaders in making foreign policy because the former must be concerned more about the public consent than the latter. The completion of my Ph.D. degree in the States has been much longer than I would care to admit. It would be nice to be able to claim that it just took that long to experience all the necessary things as an independent scholar. This dissertation also has taken longer than I expected. It would be great if I can claim that it just took that long to think it through. My nine and half year-long journey would not be possible without helps from Him, many institutions, and individuals. I cannot complain about the longer graduate study in part because I have never been without financial aid for tuition fees and stipends. Above all, I know there have been always the pillar of cloud and the pillar of fire to protect and guide my life even when I did not recognize. The University of Missouri-Columbia provided me with the first opportunity to study politics and international relations in English. I met a lot of people there both good and bad. I would like to give thanks to Byoung-Kuen Jhee (political science), now professor at Chosun University, who taught me how to drive during the exceptionally hot days of the 2001 summer in Columbia. He also provided a lot of good advice and resources. My debts are also to Jaechul Lee (political science), now professor at Dongguk University, for buying me tons of beer, Deahun Han (sociology) and his wife Eunjung Jung for treating me to a lot of dinners, Whan Choi (political science), now professor at University of Illinois-Chicago, for providing many constructive critiques, and one selfish guy for giving me the unexpected, unwelcome opportunity to learn about how to survive with an additional reduction of finance that was already much less than the minimum cost of living. I was fortunate to meet one strong man in my first semester to viii whom I should give my sincere thanks. That is Junhan Lee, now professor at University of Incheon (my alma mater). He was a post doctoral fellow at Mizzou, who had just gotten just his Ph.D from MSU. In the first semester, I was totally unable and overwhelmed by the amount of assigned readings and writings, so I almost decided to quit my graduate study. Junhan Lee taught me that I had to read all the assigned readings if I wanted to eat, sleep, and live. He said, “If you do not do all of your work, then do not eat and sleep until you finish your work.” These words still ring true in my ears. Many other Missouri Tigers deserve my thanks; Hanseung Cho, Jiho Jang, Wang-Sik Kim, Brian Bough, Hojin Han, all of whom I met in the political science department. I am also grateful to two Missouri senior professors, Minion KC Morrison (political science, now at Mississippi State University) and Edward E. Brent (sociology). They both provided role models as teachers and wrote recommendation letters when I transferred to Michigan State University. I also owe a debt of gratitude to Dr. Catherine Holland. She ensured the second semester of funding at Mizzou that was in peril. Here at East Lansing, I have also met a lot of good friends. Sangchoon Jeon was a great consultant of statistics with real understanding during my coursework, Jiyoung Jang helped me with mathematics in many occasions, and Changkuk Jung has always been with me throughout my graduate program at MSU since its initiation. I owe special thanks to Petra Hendrickson. She has proof-read the entire manuscript of this dissertation and even this preface. Without her, there must be much more awkward sentences and phrases. All the grammatical errors and typos that still remain must belong to my faults. I also want to thank my other fellow graduate students: Shane Singh, Judd Thornton, Kristopher Grady, Matthew Kirwin, and Rob Carroll. I am very thankful to my home department, many of its faculty members and staffs. Professor Saundra Schneider accepted me to the Ph.D. program with the generous 5 year ix financial support when I decided to leave the Mizzou program. She indeed made me realize that a graduate student can openly talk about his concerns with a director of graduate studies. She has been always friendly and considerate to me. Because of her, I have always been able to feel safe and warm. Professor Steve Kautz is also another example of a good graduate director with thoughtful advice and warm consideration. He was extremely helpful for me not to be out of the track when I was unexpectedly without my dissertation supervisor. Indeed, I cannot think of graduate directors better than Sandy Schneider and Steve Kautz. I also would like to thank other teachers like William Jacoby, Kenneth Williams, and Eric Chang. William Jacoby is the best example of how to learn and teach statistics with insight and effect. Kenneth Williams provided me with an idea about how to teach a huge-size introductory course in political science. He also was very generous to give out his older daughter’s bikes to my daughter. Eric Chang gave a lot of hard and time-consuming statistical homework that taught me how the basics and fundamentals could be the most difficult. My graduate life at MSU could have been much tougher without the administrative help from Karen Battin and Kelly Fenn. Karen spent many hours doing paperwork for my program and sending out my job applications. Kelly has made sure that my stipend is delivered on time. Several of my friends are MSU faculty members. Especially, my family and I owe to an excellent and warmhearted couple, Seunghyun Kim and Jaemin Cha, both professors at the School of Hospitality Business. Without them, the cold Michigan weather would be much colder. I have been blessed to have two of my great graduate mentors, Patrick James and Michael Colaresi. They make a good combination of encouragement and reality check. Pat, now at University of Southern California, was my advisor when I was at Mizzou. Things made both him and me leave for USC and MSU, respectively. Because of Pat, I was able to feel secure then in a x white male dominating place, though with the presence of many internationals. Without him, I could not have survived the first and second years in my graduate studies at Mizzou. Pat has always been a man of encouragement. His encouragement has kept motivating me to read, write, and work more. I admire his energy, openness, and intellectuality. Mike has been my advisor throughout my program at MSU. He has helped me get through my entangled academic world. I have often been in awe of how much a young scholar can be knowledgeable with deep insight, understanding, and creativity. His intellectual sharpness and precision has pushed me to think harder, deeper, and righter. Without his support and input, this project would never have come to fruition. My thanks are also due to the other members of my dissertation committee. Valentina Bali has taught me how to conduct rigorous quantitative research. She also has provided writing suggestions for the entire draft of this dissertation. Any errors still present remain my responsibility. Cristina Bodea has taught me Political Economy. Without her, I would have missed that there is another stream other than International Security in my own field, International Relations. Matt Zierler, at James Madison College within MSU, has provided me with timely advice and care. Besides the committee members, I have benefited from other distinguished scholars who have read and commented on parts of this dissertation at the various stages. Those are David Carter, John Vasquez, Allan Stam, Dan Reiter, Bruce Russett, John Oneal, Michael Tomz, Douglas Lemke, Mark Souva, Erik Gartzke, Sara Mitchell, Paul Hensel, Brandon Valeriano, Douglas Gibler, Christopher Gelpi, Joanne Gowa, and Michael Mousseau. I am grateful to my teachers at my alma mater college, University of Incheon: Kwangha Jung (now professor emeritus), Jaeseok Lee, Kisuk Cho (now at Ewha Womans University), xi Gupyo Lee, Hochul Lee, and Beon-Shik Shin (now at Seoul National University). They made me love the study of politics and international relations and dream about studying abroad to get the degree in America. Finally, I have been blessed with support, encouragement, warmness, and all the sacrifices my family has provided me with. I know my moms (my own and my wife’s) have never stopped praying to God for my successful graduate study. My dads (my own and my wife’s) have been supportive. Especially, I appreciate the love, consideration, resources, and support my father-in-law has never failed to offer to me. I hope I can be a father-in-law like him for my daughter’s future husband when she becomes a lady. Throughout my graduate career, there have been a lot of variables in terms of people, places, resources, and many other factors. There is only one constant: my wife, Mitaek. I really appreciate all of her love and patience that I do not necessarily deserve. She continues to make me a better person than I have been. The specific ideas for this dissertation research began to develop in the Fall Semester of 2005 when my beloved daughter, Juwon, was conceived. She is turning five in the coming June. Although she has been taking a lot of my research time, those have been blessing years. She has provided peace and happiness for me and helped me find my home. There may be no greater joy than I observe she grow in wisdom and statue, and in favor with God, women and men. To her, my first born, I dedicate the first full-blown fruit of my academic work. Johann Park Hidden Tree, East Lansing, Michigan March 11, 2011 xii TABLE OF CONTENTS List of Tables………………………………………………………………..…..xiii List of Figures………………………………………………………………....xiii CHAPTER 1 The Second Wave Critiques of the Democratic Peace Interests, Issues, and Victory.................................................………...……….……1 The democratic peace and its first wave critiques...................................…....3 The Second Wave Critiques and Research Questions….……………..….….6 Variants of the Democratic Peace Theory……………….………...……….22 The Plan of the Dissertation…………….….…………………..….….…....27 CHAPTER 2 Forward to the Future? Democratic Peace after the Cold War.....................................................................29 Explaining the Democratic Peace…..……………………………….……..31 Hypotheses: Post-Cold War Democratic Peace.……………….………......37 Research Design…………………………………………………..……..…43 Results and Analysis……………………………………………….…........51 Conclusion…………………………………………………...………….….61 CHAPTER 3 Bordering on Peace or Pacifying the Disputatious Borders? Democracy, Territory, and Armed Conflict………………………………….…...59 Territorial Issues and the Democratic Peace……………………………….65 Bordering on Peace? Replication Analysis for Gibler’s 2007 ISQ Study..............................................................................................................73 Pacifying Disputatious Borders? Interactive Analysis of Democracy and Territory........................................................................................................98 Conclusion………………………………………………………………..122 CHAPTER 4 Picking the Fight, Winnability or Justifiability? Democracy and the Selection of Armed Conflict……..…………………………126 xiii Democracy and Wise Selection: Moving beyond Victories and Capabilities……………………………………………………………..…129 Selection Mechanisms: Winnability vs. Justifiability………………….....136 Hypotheses………………………………………………………………..141 Research Design…………………………………………………………..153 Results and Analysis……………………………………………………...158 Conclusion……………………….………………………………………..167 CHAPTER 5 Conclusion Debates, Findings, Limitations, and Implications...……………………………..170 Issues and Problems in the Debate………………………………………..171 Summary of Analyses and Findings………………………………………172 Limitations………………………………………………………………..174 What I Have Learned……………………………………………………..175 Implications……………………………………………………………….178 REFERENCES…..……………………………………………………..……....184 xiv LIST of TABLES Table 2.1 Predicting MID Onset, the entire period, 1816-2001…………………..51 Table 2.2. Joint Democracy Effect on MIDs during and after the Cold War……..54 Table 2.3 Probability Changes in Risk for Annual MID onset according to Joint Democracy………………………………………………………………….……..56 Table 2.4. The Robustness Checks for the Time-Variant Pacifying Effects of Joint Democracy………...………………………………………………………………59 Table 2.5. Probability Changes in Risk for Annual MID Onset…………………..60 Table 3.1. Replication of Model 4 of Table 2 in Gibler (2007): The Effects of Borders on the Likelihood of Joint Democracy, 1946-1999…..….81 Table 3.2. Replication of Model 4 of Table 3 in Gibler (2007): The Effects of Borders and Joint Democracy on Conflict, 1946-1999…………...87 Table 3.3. Multicollinearity Checks………………………………………………91 Table 3.4. Separate Analyses of MID Prediction by the Border Variables and Joint Democracy, respectively……………………………………………………...…..93 Table 3.5. Probit Analysis of the Effects of Democracy and Territory on MIDs, 1919-1995………………………………………………………………………..113 Table 3.6. Four Categories by Democracy and Territory………………………..113 Table 3.7. Interactive Effects of Democracy and Territory on MID Onset, 19191995.......................................................................................................................115 Table 3.8. Interactive Influence of Territory and Democracy on the Probability of a MID……………………………………………………………………………...120 xv Table 4.1. Contingent Effects of Winnability Factors on MID Initiation, 19811999……………………………………………………………………………...159 Table 4.2. Contingent Effects of Winnability and Justifiability Factors on MIDS, 1981-1999………………………………………………………………………..161 Table 4.3. Contingent Probability Changes in Risk for Annual MID initiation by Statistically Significant Variables according to Challengers’ Regime Type……165 xvi LIST OF FIGURES Figure 2.1. Time Trend of Global Average Democracy Level, 1816-2001……….38 Figure 3.1. Predicted Probability of MID and Net Effects of Lowest Democracy……………………………………………………………………..….95 Figure 3.2. Interactive Effects of Lowest Democracy by Land Contiguity……....96 Figure 3.3. Types of Dyads and Probabilities of MID Onset……………………118 Figure 4.1. Democracy, Accountability, and Choice of Armed Conflict………..138 xvii CHAPTER ONE THE SECOND WAVE CRITIQUES OF THE DEMOCRATIC PEACE: INTERESTS, ISSUES, AND VICTORY This dissertation consists of three related but distinct essays that explore and address the important theoretical and empirical gaps found in the second wave of the democratic peace debates. Democracy and war (or peace) have been two of the most important concepts in the study of politics. An extensive body of literature has written about the relationship between democracy and war over the last five decades (e.g., Babst 1964; Small and Singer 1976; Rummel 1983; Doyle 1986; Maoz and Abdolali 1989; Bremer 1993; Cederman 2001; Oneal and Russett 2005; Harrison 2010). This dissertation research seeks to deepen our understanding of the relationship between these central concepts that encompass a broad range of IR perspectives such as realism, liberalism, and issue-based approach. I will identify, explore, and address three important caveats that are informed from the lingering debates of the democratic peace among the paradigms. It is informative to identify what has been generally accepted with respect to the democratic peace. The following five phenomena have been found to be empirical regularities about the relationship between democracy and peace (or war): (1) in aggregate, democracies are no less prone to violent conflict than nondemocracies (the counter monadic democratic peace); (2) however, democracies do rarely fight each other (the dyadic democratic peace); (3) once involved in disputes with other democracies, democracies are more likely to resolve interdemocratic conflict peacefully as compared to pairs of autocracies or mixed dyads (the dyadic democratic peace process); (4) democracies tend to win the wars they have fought, and 1 especially, this victorious effect is more apparent when they are initiators (the democratic war success and selection effects); and (5) democracies tend to suffer fewer battle deaths and fight shorter wars, especially when they initiate (the selection effects and casualty sensitivity). Roughly speaking, sets of democratic norms and institutions have been proposed as theoretical underpinnings to explain these empirical phenomena. Norms of peaceful conflict resolution socialize democratic leaders to adopt reciprocity in foreign relations and avoid extreme recourse to unilateral concessions and violence. Democratic institutional features such as separation of powers, political competitions, and regular elections restrict the autonomy of leaders from inadvertently embarking upon military adventure abroad. Although numerous studies have contributed to our understanding of the unique conflict patterns of democracies, there remain several well-known loopholes, both theoretical and empirical, in the democratic peace scholarship. These loopholes refer to at least three caveats made by the second wave critics concerning the potential confounding effect of the Cold War, the lack of territorial issues as a source of interstate violence between democracies, and democracies’ willingness to fight weaker enemies. Three scholars—Joanne Gowa, Douglas Gibler, and Michael Desch—stand out for raising important skepticisms against the democracyconflict causal relationship that have yet to be answered properly from the democratic peace scholarship. Each critic makes a distinctive caveat against the veracity of the democratic peace that this dissertation research seeks to address: 1. The (dyadic) democratic peace is limited to the Cold War period (Gowa 1999). 2. A lack of contested boundaries, neither norms nor institutions, promotes democratization and explains the separate peace among democracies (Gibler 2007). 2 3. No direct evidence exists for the causation from democratic carefulness to victorious war outcomes (Desch 2008). This dissertation project aims to offer both theoretical and empirical keys to the puzzles instigated from these caveats. The importance of these critiques can be best captured in terms of context, by discussing where they are placed in the evolution of the democratic peace literature. Therefore, the next section of this introductory chapter will review the first wave of democratic peace critiques. The third section will explore the second wave critiques that motivate the research questions of this dissertation project. The fourth section will discuss the four variants of the democratic peace explanations that provide the theoretical framework for this dissertation’s working hypotheses. The fifth and last section will provide the plan of this dissertation as an overview. The Democratic Peace and Its First Wave Critiques It can be said that the democratic peace literature has advanced both empirically and theoretically at least through two stages. The first stage involved explanations of the monadic and dyadic democratic peace propositions (1−3). The second stage involved findings relating to the monadic democratic selection success and casualty sensitivity propositions (4 and 5). But such development has been painted with many theoretical and empirical skepticisms, and critiques have tended to cluster into two waves. In the first wave, realists challenged the idea of the democratic peace most actively. This was a natural response from the realist camp, since the democratic peace would undermine the traditional realist claim for the primacy of power and systemic factors: “the important elements 3 of international politics can be explained without differentiating states by regime types” (LynnJones 1996: xi). Some disputed the very existence of the separate peace between democracies both on theoretical and empirical grounds. Layne (1994) and Spiro (1994) argued that the structural constraint-based explanation of the democratic peace is weak because they cannot explain the first regularity: democracies are as war-prone as autocracies. These critics regarded the normative explanation as a promising candidate that might explain the first three regularities. Spiro pointed out that few empirical studies had examined the democratic peace in terms of normative liberal causes. In this vein, to test the normative explanation, Layne selected four cases of “near miss” crises in which democracies came close to war but avoided it. Based on his review of these four historical records, he concluded that the factors that drove the near misses were not shared democratic norms but traditional realist factors such as calculations of national interests, strategic concerns, military capabilities, and military threats. Spiro also offered an unfavorable empirical finding that the observed number of democratic wars that could have been caused by chance is substantially greater than 50%. He argued that the statistical significance found for the democratic peace is a statistical artifact of serially correlated data. However, these critical studies had serious methodological problems. Layne inappropriately selected on the dependent variable, not considering “all the dogs that did not bark” (Russett 1995: 167). Spiro relied on an inappropriate probability model (Russett 1995; Hewitt and Young 2001), and alternative studies showed that the relationship still held even controlling for the temporal dependence in the data (Beck, Katz, and Tucker 1998; Maoz 1998; Russett and Oneal 2001). Gowa (1995; see also Farber and Gowa 1995, 1997) argued that both normative and structural explanations are unpersuasive. She suggested that norms of peaceful conflict resolution are not unique to democracies; other types of states also have incentives to resort to such norms 4 because wars are costly options of resolving disagreements (cf., Fearon 1995). In contrast to Layne and Spiro, Gowa regarded the checks-and-balances system of democracy as the core explanation for the democratic pacifism. However, she argued that the structural explanation is undermined by the well-known propensity of democracies to wage wars against nondemocracies. It should be noted that Gowa did not negate the existence of the dyadic democratic peace per se. Rather, she challenged the notion that democracy is the prime mover. Gowa argued that the democratic peace is an artifact of the Cold War by presuming that democracies shared security interests exerted by the Cold War bipolar politics. Farber and Gowa (1995) provided an empirical finding that is more plausible than other realists’ evidence. They found that the relationship between peace and democracy was not statistically significant except for the 19461980 period. Students of the democratic peace have responded to Farber and Gowa, but their work still remains “the most significant realist critique of the democratic peace” (Gelpi and Griesdorf 2001: 634). For this reason, I include Gowa’s critique in the second wave of democratic peace critiques as discussed below. Gowa’s follow-up studies (Farber and Gowa 1997; Gowa 1999) and related studies by other scholars (Gartzke 1998, 2000, 2007; Werner 2000; Cederman 2001) give us interesting puzzles on the presumed time-varying effect of democracy. Another set of critiques active in the first wave debate concerns reverse causality between peace and democracy. Thompson (1996) argued that the democratic peace is something like “putting the cart before the horse.” Some realists share Thompson’s concern. For instance, a structural realist, Patrick James, in his study with other colleagues (James, Solberg, and Wolfson 1999), holds that there is inverse causality between democracy and peace, with the causal arrow running from peace to democracy. But the reverse causality critique at the monadic level has not 5 seriously impaired the dyadic democratic pacifism. In effect, Reiter (2001) shows that peace does not affect the joint level of democracy. Even at the monadic level, Mousseau and Shi (1999) find that nations are about as likely to become more institutionally autocratic as they are to become more democratic in the periods before the onset of wars. Now even Thompson (see Rasler and Thompson 2004), one of the first critics of the reversed causality, admits that democracy promotes peace as well as peace nurtures democracy. In sum, the first wave of democratic peace critiques made important theoretical and empirical arguments against the democratic peace. However, these critical arguments have yet to provide extensive, systematic empirical evidence while most of their empirical critiques have serious methodological problems. The Second Wave Critiques and Research Questions The second wave debate on the democratic peace has extended to include the fourth and fifth regularities. The theoretical and empirical issues in this debate concern the loopholes found in discovering and explaining democratic war success and selection, territorial issues as a source for both democracy and peace, and preference similarity as an explanation for peace among democracies. As the first step to explore and address the three loopholes, I begin by asking three general questions that provide the focal points of this dissertation research. The first question is about the presumed-time varying effect of joint democracy: does the democratic peace persist or even strengthen after the Cold War? The second question is related to the effect of joint democracy vis-à-vis territorial issues: does controlling for territorial issues render the effect of joint democracy insignificant or at least do territorial issues limit its pacifying effect? The third 6 question concerns the selective decision mechanisms of democracies at international conflict: do democracies initiate armed conflict for different reasons than nondemocracies? As demonstrated below, these three questions are informed from and implicated by the second wave democratic peace critiques. Common Interests and post-Cold War Democratic Peace Does the democratic peace persist or even strengthen after the Cold War? Although Gowa’s original critique discussed above would belong to the first wave critique, it remains fairly relevant to the current democratic peace scholarship, motivating the first research question. First, many proponents still acknowledge Gowa’s (Farber and Gowa 1995, 1997; Gowa 1999) analysis of the Cold War-driven democratic peace as the most “formidable” or “cogent” challenge against the democratic peace scholarship (Gelpi and Griesdorf 2001: 634; Russett and Oneal 2001: 60). To date, many studies on the democratic peace still take Gowa’s argument as their null hypothesis (e.g., Gelpi and Griesdorf 2001; Schultz 2001b; and Huth and Allee 2002). Second, both Gowa’s analysis and proponents’ responses (e.g., Oneal and Russett 1999b) are far from a complete investigation due to the lack of a direct measure for Cold War effects and the limited temporal domain that does not encompass the post-Cold War era. Third, and most importantly, Gowa’ critical argument provokes some interesting questions about the contemporary post-Cold War international system like the research question specified above. If realists like Gowa are right, we would no longer observe the reduced conflict propensities in inter-democratic state relations. If liberals like Cederman, Oneal and Russett are right, we may expect a more pacific interstate system in the post-Cold War era where the world has evolved to be more democratic than ever. 7 Many realists have suggested that the democratic peace was caused by the US superpower that had strategic incentives to protect fellow democracies in Europe against the Soviet power and other Eastern European communist regimes during the Cold War (e.g., Mearsheimer 1990a; Rosato 2003). Therefore, for these realists, democracies had reasons to cooperate under the collective security umbrella built by the US, but such reasons were motivated not by democracy but by common security interests (Gowa 1999). The realist critique is hard to ignore for two reasons. First, the Cold War, in fact, was a long-lasting conflict between the Western and Eastern Blocs that had dominated the interstate system for the last half of the 20th century. Undeniably, Western Bloc countries were representative democracies at that time. So the established democracies coincided with the Western Bloc members. It is plausible to argue that security interests shared among states reflected at least partly their bloc membership. Second, Gowa (1999) adds empirical evidence to that plausibility. She finds that the dyadic democratic peace did not exist during the pre-Cold War era and that democracies were more likely to ally each other than other combinations of dyads during the 1946-1980 period. Now the Cold War is gone, and its influence has been phasing out quickly. The Western democracies lost their common enemy (i.e., the communist threat) that allowed them to share security interests. Various other sources, such as modernized terrorist groups, nuclear-capable rogue states, Somali pirates, and global warming, have arisen to pose new serious security threats. However, these concerns are spread widely but thinly across the global society that is beyond the boundaries of the Western democracies. Additionally, interstate alliance systems do not appear to be formed and structured around the current global issues, unlike the Cold War rivalry. Hence, the postCold War era may not observe the separate inter-democratic cooperation to a substantial degree. 8 It is possible to make some theoretical and empirical defense of the time-inconsistent effect of joint democracy. Maturity may be the key to that issue. Mansfield and Snyder (1995, 2005) have argued and found that democratizing states tend to experience more external conflicts than stable regimes. They attribute this inconvenient finding to the two barriers unique to the countries in the initial stages of expanding political participation: (1) the weakness of democratic institutions; and (2) the resistance of social groups who would be damaged in a process of democratization. Other scholars have disputed Mansfield and Snyder’s argument and finding, arguing that the regime instability, rather than transition toward a particular regime type, is the main problem. These scholars find that autocratization is as conflict-generating as democratization or even more so (Enterline 1996, 1998; Maoz 1998; Ward and Gleditsch 1998; Braumoeller 2004). Most recently, Narang and Nelson (2009) challenge Mansfield and Snyder’s empirics in terms of data, measurement, model specification, and case selection. In any case, it is apparent that democracy needs times for maturing to be a real factor that pacifies state interactions. A more recent study by Cederman (2001) gives us a hint on this issue. Cederman reinterprets the democratic peace as a dialectic and macrohistorical learning process, highlighting Kant’s emphasis on human learning. For Kant, peace can be perpetuated as a result of progressive learning processes. Cederman finds that the conflict propensities of democratic dyads have steadily fallen since the nineteenth century, and that the pacifying effect of joint democracy is strengthened as it matures. As such, realists and liberals have divergent expectations about the post-Cold War system. Testing the effect of joint democracy for the post-Cold War period will provide the focal point in resolving the debate over the democracy-peace relationship. It is timely to that purpose, now that about 22 years have passed since the winding down of the Cold War system. The updates from 9 the two main data sources, the Correlates of War and Polity projects, jointly allow a test of 12 post-Cold War years (1990-2001). In Chapter 2, I point out that nothing has been directly tested regarding the contrasting claims between realists and liberals. I investigate the presumed time-variant effect of joint democracy, based on a set of period-specific probit models for the 1950-1989 period and the 1990-2001 period, respectively. Also, using the information from the COW alliance data, I employ more direct measures of the Cold War preferences than extant research to test if the effect of joint democracy is rendered statistically insignificant by introducing these measures. Evidence suggests that joint democracy promotes peace independently of the Cold War security interests. Furthermore, against realist expectations, the effect for democracy is estimated to be larger in the post-Cold War era. However, the difference between the consistently negative effects across the two periods is not itself significant in the main analysis, although all of the alternative analyses for robustness checks do produce the significant results. The Democratic Peace vis-à-vis Territorial Issues Does controlling for territorial issues render the effect of joint democracy insignificant or do territorial issues at least limit its pacifying effect? The second wave debate on the democratic peace includes non-realist skeptics. These recent participants in the conversation are concerned with how issues such as territory add a nuance to the well-established effects of regime types (James, Park, and Choi 2006; Gibler 2007; Hutchison and Gibler 2007). They suggest that the democratic peace is epiphenomenal to the stability of borders and wanes in the context of territorial issues. For instance, Gibler (2007) suggests that democracies have few territorial issues and rarely fight militarized interstate disputes (MIDs) with each other because they have settled 10 1 borders with their democratic neighbors. He finds that inclusion of geographic measures for border stability wipes out the statistical significance of a measure for joint democracy. He goes on to make a bold claim that both democracy and peace are symptoms, not causes, of stable borders. In other words, the separate peace between democracies is in fact the by-product of stable borders that encourage the emergence of democracy. In line with this argument, using the territorial claim data of the Western Hemisphere (Hensel 2001), James et al. (2006) find that there are few systemic effects of democracy that lessen the violent escalation and encourage the peaceful settlement of disputes over territory, presumed to be the most contentious issue among others. They argue that territorial disputes can wash out the effects of the democratic peace. Taken together, these studies suggest that the key to peace is not democracy but territory. If true, these skeptics’ arguments and findings seem to overturn the well-established findings on the dyadic democratic peace that have been accumulated over the last five decades. However, there are reasons to have serious reservations about the plausibility of these skepticisms themselves. In case of Gibler (2007), he conceives border stability in a too-static manner based on geographic constraints exogenously imposed upon states. In Gibler’s framework, states, leaders, and citizens are characterized as incapable of acting away from geographically structured constraints. This scheme is not only of little practical value to Gibler’s model as a social theory but also is unrealistic. International borders are institutions constructed as a result of mutual acceptance between neighbors to advance mutual benefits (Simmons 2005). While certain geographical features may have some impacts in drawing mutually acceptable boundaries, their influence should not be absolute. For instance, in Gibler’s own analysis, none of his border variables has a consistent and robust effect across model specifications. Therefore, 1 MIDs are defined as instances in which at least one state took militarized action against another state in the form of a threat, display or actual use of force (Jones, Bremer and Singer 1996: 195). 11 it is unclear whether the significance for the joint democracy measure disappears due to the true causality of the border variables on democracy and peace or due to the model’s inefficiency caused by adding many irrelevant variables that would inflate the standard error for joint democracy. Besides, Gibler’s geographical measures for border stability are at best proxies. Even his intellectual godfather, John A. Vasquez, the founder of steps-to-war theory, identifies this use of proxy measures as a main weakness of Gibler’s analysis. A much better measure than Gibler’s proxies would be a direct indicator for the presence of interstate disagreements over a piece of territory. Since his 2007 study published in International Studies Quarterly, Gibler has used militarized territorial disputes as a direct measure of territorial threat in his subsequent, related studies that attempt to show that territorial threat at the interstate level influences the values and characteristics of democracy at the domestic level (Hutchison and Gibler 2007; Gibler 2010). Though this effort may be regarded as a significant progress (Vasquez 2009), the evidence based on the independent variable measured as militarized territorial disputes can add little (if any) to Gibler’s “bordering on peace” thesis. This is because in his 2007 analysis, Gibler uses onset of MIDs as the dependent variables to prove that his proxy border variables render joint democracy to have no independent impact upon military conflict. The points made so far necessitate the use of nonviolent territorial disputes as an appropriate measure for territorial threat and unstable borders. In effect, related data for this purpose are available. Both Hensel’s (2001) Issue Correlates of War (ICOW) project and Huth and Allee’s (2002) list provide the information on territorial disputes at the claim level that precede militarization. These datasets allow revisiting Gibler’s critical analysis with a more appropriate measure of territorial threat. In my own analysis, I opt for Huth and Allee’s data because its spatial domain covers the entire globe. 12 In addition, Gibler’s analysis is fraught with other methodological flaws. To state briefly, first, Gibler fails to apply adequate statistical models for his interactive border variables whose effects are presumed to be contingent upon land contiguity. By dropping constituent terms in the equations, the presumed interactive effects are not assessed properly and clearly in Gibler’s analysis. Second, Gibler fails to use the industrial standard methods for the time-series crosssection (TSCS) data in which temporal correlations and clustering on dyads would hamper appropriate hypothesis testing. Although acknowledging Beck, Katz, and Tucker’s method for BTSCS (binary times-series cross-sectional) data, Gibler does not follow their suggestions to use robust standard errors and cubic splines to account for possible clustering and nonlinear serial correlation. Third, Gibler measures joint democracy inconsistently. When predicting the observation of a jointly democratic dyad by his border variables, he uses a dichotomous measure. But when predicting MID onset by joint democracy and his border variables to compare their independent causal power, he uses a continuous measure for joint democracy based on the weakest link assumption (Russett and Oneal 2001). All of these weak points and methodological problems will be detailed in the replication section of Chapter 4. Also, James et al.’s (2006) critical analysis on the democratic peace vis-à-vis territorial issues contains several weaknesses. First, they are almost silent about an explanation for why the effect of joint democracy is canceled out in the context of territorial claims. Second, their empirical domain is limited to the territorial claims in the Western Hemisphere. Third, their data are based on a selected sample consisting of dyads already involved in territorial claims. Therefore, it is unclear that James et al.’s null finding is whether because the pacifying effect of joint democracy is conditioned by territorial issues or because only conflict-prone democracies are included in their data due to their selection criteria. 13 Taken together, the territorial issue-based claim for the spuriousness and ineffectiveness of the democratic peace needs further investigation because of its accompanying theoretical and methodological flaws. I think that the relationship among democracy, territory, and armed conflict is different from those suggested by the above studies. The effects of democracy and territory on armed conflict may be contingent on each other. There are two possibilities for this conditional effect. For example, existing theories of the democratic peace argue that democracies will not use force to seek territorial expansion because acquired territory can only serve a small fraction of broad democratic constituencies (Lake 1992; Bueno de Mesquita et al. 2003) or because democratic institutionalization decreases nationalist and irredentist claims (Kacowicz 1995: 265). Thus, the pacifying effect of democracy should be stronger on territorial issues than other types of issues. However, the territorial explanation of war and conflict, outlined by Vasquez and his coauthors (e.g., Vasquez 1993, 2009; Senese and Vasquez 2003, 2008), tells us a different story. According to this perspective, territory is something that, by nature, every human being inherently pursues eagerly. Also, territory involves both symbolic and tangible values that are very important for national integration, security, and development. Given this high salience of territorial issues, when democracies are involved in territorial disputes, they may reveal few differences from nondemocratic states. In Chapter 4, I conduct two separate analyses. First, I revisit and replicate Gibler’s analysis. Since Gibler’s data are nowhere available unlike the author’s note on the first page of his 2007 article and he has been unwilling to help, I have to use my best guess for some variables for which Gibler provides unclear descriptions in reconstructing the replication data by myself. My replication analysis, even without any methodological corrections, shows that joint 14 democracy has a significant effect in abating interstate conflict regardless of the inclusion of the border variables that Gibler supposes to precede the causal power of joint democracy. In contrast, many of the border variables generate insignificant or inconsistent effects. Second, questioning Gibler’s measures for territorial issues and James et al.’s design in testing the democratic peace in terms of territorial issues, I conduct an alternative interaction analysis based on Huth and Allee’s territorial claim list for the entire globe. Even if Gibler’s “bordering on peace” thesis were true, the proposed causal precedence of territorial threat does not necessarily rule out the possibility that joint democracy can still pacify interstate interactions in the territorial context. Second, I conduct an interactive analysis on the democracy-territory linkage. James et al.’s finding carries a partial story at best because of the limited sample and lack of explanation. My interactive analysis goes beyond James et al. (2006) to provide both theoretical and empirical reasons why the democratic peace does or does not hold up for territorial issues. Interactive hypotheses will be derived deductively via a model of issue specification from both the territorial explanation and the democratic peace theorizing. In order to compare the effects of democracy conditional upon the presence and absence of territorial issues, all interstate dyads are considered, interacting regime types of all dyads with territorial claims. The use of all dyads enables us to not only avoid the possible selection bias but also assess the statistical uncertainty for the possible interactive effects of democracy and territory. The results based on MID data and Huth and Allee’s list from 1919-1995 reveal that the pacifying effect of joint democracy stands up to both territorial dyads and nonterritorial ones in spite of the imperatives toward militarization created by territorial conflict. However, territory still appears to increase the likelihood of armed conflict between two democracies. With respect to the interactive effect, joint democracy appears to have a larger effect in the territorial context than the nonterritorial context, meaning that the gap 15 between democratic dyads and nondemocratic ones in the probability of MID onset gets wider in the sample of dyads with territorial claims than those without territorial claims. Democracy and Wise Selection Does political accountability lead democracies to wiser selection of interstate conflict? Further, what constitutes a wise selection of conflict? The third essay of this dissertation attempts to answer these questions in order to shed light on the linkages between conflict choice, political institutions, and beneficial foreign policy outcomes. The realist-liberal debate on the democratic peace has expanded to include the questions of conflict selection and outcome in relation to regime types. David Lake (1992) offered a seminal investigation. He found that democracies had won an overwhelming proportion of wars they had fought. He offered an explanation that focuses on the high power status of democracies accrued through the growth-generating effect of democracy (see also Olson 1993; Bueno de Mesquita et al. 2003; Schultz and Weingast 2003). He argues that unlike autocracies, democracies are constrained from rent seeking and thus tend to create fewer economic distortions, generating greater national wealth and devoting more resources to national security. However, although well-established democracies tend to be wealthy countries and high income is strongly associated with the survival of democracy, there is little consensus about the impact of democracy on economic growth (e.g., Barro 1997; Przeworski et al. 2000; Doucouliagos and Ulubaşoğlu 2008). Besides, democracy remains a statistically significant factor for victory even controlling for military capabilities (Reiter and Stam 1998a; 2002). Among proponents of democratic victory, a consensus seems to have emerged from emphasizing careful and selective decision-making processes in democratic institutions when 16 initiating the use of force (Reiter and Stam 1998a, 2002; Bueno de Mesquita et al. 1999, 2003; see also Siverson 1995; Schultz 2001b). In other words, various democratic accountability-based explanations commonly argue for democratic selection effects that democracies tend to win wars because they purposively select only easy and winnable wars. Greater political accountability of democratic leaders is commonly proposed to be the key for the wiser conflict selection. In addition, political competitions and the marketplace of ideas that characterize democracy are thought to help democratic leaders better estimate the chances of favorable war outcomes. Proponents have accumulated ample evidence for these selection effects in terms of materialistic capabilities, physical costs, and outcomes of war. To summarize briefly, as compared to nondemocracies, democracies are found to be more victorious especially when they initiate wars (Reiter and Stam 1998a, 2002; see also Gelpi and Griesdorf 2001), suffer fewer battle deaths and civilian casualties (Siverson 1995; Valentino, Huth, and Croco 2010), meet fewer violent responses from their targets for the military threat they initiate as compared to other states (Schultz 1999; 2001b), be less likely to participate in and escalate armed conflict than nondemocracies when there is an military disadvantage (Bueno de Mesquita, Morrow, Siverson, and Smith 2004), and fight shorter wars, especially those they initiate (Bennett and Stam 1996, 1998; Bueno de Mesquita, Koch, and Siverson 2004). As with other democratic peace-related issues, the selection effects argument has generated a lively debate between proponents and realist critics. Most of the realist critiques have focused on Reiter and Stam’s seminal work that is thought to provide the most explicit argument and finding related to the democratic selection effects. Realist critiques reveal four caveats regarding the selection effects argument. First, Reiter and Stam and other proponents have failed to consider that autocratic leaders also are constrained from engaging in a risky and failing war 17 that would cost not just decision makers’ political life but biological one in autocracies (Desch 2002, 2008; Rosato 2003; see also Chiozza and Goemans 2004). Second, the proponents have overlooked the possibility that democratic decision-making processes that involve variegated but uninformed public opinions can lead to poor foreign policies (Kaufman 2004; Cramer 2007; Downes 2009). Third, equally neglected is that for realist reasons (e.g., coping with threat to territorial integrity or political interdependence), democratic leaders can resort to deception to engage the nation into wars against highly undefeatable enemies (Schuessler 2010). Fourth, proponents’ evidence is based on flawed research designs that are incapable of direct testing of their causal mechanisms (Desch 2008; Downes 2009). Desch’s work, Power and Military Effectiveness, deserves more space here since it provides the most comprehensive critiques that also closely pertain to my third essay. He makes a case that democracy “hardly matters” for explaining victory and loss in war. He makes three unique points regarding the selection effects and democratic victory. First, Reiter and Stam’s data are full of mismatched war cases. According to Desch’s yardstick, 54 of 75 war cases are “gross mismatches” in which democratic war participants enjoyed at least 2:1 material capabilities. After excluding these cases, the relationship between democracy and victory is found to be reduced significantly. His approach is useful to discern the military effectiveness of democracy against the effect of military power. But it should be noted that the fair match analysis cannot be an appropriate test for the selection effect argument. The fact that democracies have been involved in a disproportionate number of mismatched wars may reflect the very selectiveness of democracies in choosing wars. Second, the democratic initiator variable has one of the smallest marginal effects while other realist variables like terrain and material capabilities have the strongest effects. However, this approach is too simplistic and misleading. 18 Comparing the substantive effects of independent variables based on marginal effects is inappropriate because a marginal effect of an independent variable in the nonlinear regression models depends on the values of other independent variables as well as the slope coefficient (Long 1997). Besides, it is inappropriate to isolate the marginal effect of a multiplicative interaction term of two variables while its component terms are missing in the equation, and the logit or probit model makes this much more difficult (Ai and Norton 2003; Braumoeller 2004; Brambor, Clark, and Golder 2005; Berry, DeMeritt, and Esarey 2010). A better method is an analysis of probability changes (King, Tomz, and Wittenberg 2000; Berry, DeMerritt, and Esarey 2010). In their response to Desch’s (2002) initial critique, Reiter and Stam (2003: 177) report, “a democratic initiator (POLITY score 10) has a 95 percent chance of winning, whereas a mostly autocratic regime (POLITY score of -7) has a 62 percent chance of winning….” Third, Reiter and Stam’s analysis does not provide a direct test for the causal mechanisms that link democratic selectivity to war victory. Desch correctly points out that the selection effects argument starts from the peacetime decision making processes and goes to the initiating stage of armed force, but the proponents’ analyses focus on countries already involved in wars. One way to address this is to use a two-stage selection model, but such models are notoriously difficult to implement correctly and highly sensitive to identifying assumptions on error correlation and model specifications (Sartori 2003; Clark and Olmsted 2009). Besides, it is hard to find support for one of the core tenets—the public punishes leaders vis-à-vis conflict outcomes—from the second stage data that are highly biased in favor of low-cost wars (Schultz 2001a). Desch argues for a case study of detailed process tracing, rather than statistical modeling, as an optimal way to test for the selection effects argument. He finds that according to his own standard, there are only ten cases that democracies initiated wars. However, this strategy is also problematic to disprove the 19 democratic selection effects. Desch makes the same error that he charges against the proponents since he himself focuses not on the peacetime decision making stage but on the wartime stage. Because of this error, he fails to look at “dogs that did not bark.” If there were only ten cases of wars initiated by democracies, such rarity may suggest that democracies have been prudent enough to initiate such a small number of wars in history. As Desch suggests, the selection effects argument apparently includes the decision not to initiate risky conflict in the first place. An appropriate approach should take one step back and look at the decision to initiate armed conflict against particular opponents. This can be done with a large-n statistical design that includes all available directed dyadic years as units of analysis to consider peacetime decision making before initiation. In sum, both proponents and critics provide incomplete answers to the question of democratic selectivity with important theoretical and empirical flaws. As mentioned above, despite its substantial amount, the evidence provided by proponents is indirect at best because of their inattentiveness to the initiation stage of conflict while their accountability-based arguments explicitly refer to selective initiation of conflict. Also, their explanations are silent about the cases that well-established democracies do engage or take the initiative in difficult wars and crises like two of World Wars, Vietnam War, Cuban Missile Crisis, and even the current war in Iraq. The critics’ arguments and methods are not also without criticisms. Whereas they raise a valid theoretical concern for the possible vulnerabilities of democratic foreign policy making that is driven by ignorant and impulsive public opinion, they often resort to a single or a few cases deliberately selecting on the dependent variable (Kaufman 2004; Cramer; Desch 2008; Schuessler 2010). Their statistical testing is indirect like that of their academic counterparts by only examining states that are involved in wars (Desch 2008; Downes 2009). Most importantly, 20 it remains as a historical fact that democracies have won a large majority of war they have fought. While the critics accept that fact (Desch 2008; Schuessler 2010), they never offer a theoretical explanation for why it is only a correlation, not a causal relationship. In Chapter 4, I address these theoretical and empirical weaknesses in existing research on the linkage between democracy and deliberate conflict initiation. I argue that extant research is too much outcome-oriented to confirm or disconfirm democratic selectivity because of either fascination or antipathy toward the idea of victorious democracies. Because of this inadequate orientation, direct testing for democratic selectivity has been delayed. I aim to test the very assumption of democratic selectivity that precedes its “effects” (i.e., easy fight and victory). If democracies are really smart in initiating conflict, they must be less likely to initiate armed conflict against strong and difficult foes to a greater degree than nondemocracies. Another related loophole I seek to address is that the unbalanced focus on war outcomes results in an underspecified definition of smart conflict decision. Indeed, both of the proponents and critics are silent about what factors constitute democratic selectivity. If not, nothing but materialistic military capabilities is regarded as an important criterion for selective conflict initiation. I explore other possible factors that effect selective initiation of conflict in the first stage of military conflict. Drawing upon the accountability explanations, I reframe democratic selectivity regarding conflict initiation in terms of justifiability as well as winnability. I theorize that political accountability motivates leaders to choose armed conflict with justifiable and agreeable causes as well as high chances of victory. This argument is tested empirically by considering target countries’ comparative military capabilities, geographical obstacles, human rights performance, and commercial relations. The statistical results based on initiation of MIDs show that democracies tend to attack weaker foes to a greater degree than do autocracies, that 21 democracies tend to fight with presumably unjust countries with bad human rights practices, and that democracies are unwilling to fight with economically important trading partners. Variants of the Democratic Peace Theory In order to answer the three general research questions, I will develop a set of related hypotheses in each essay drawing upon various explanations of the democratic peace. Therefore, it is helpful to provide a theoretical orientation in this introductory chapter. Let us briefly summarize the variants of the democratic peace theory. There are largely two kinds of explanatory frameworks—normative and institutional— for the democratic peace. First, the normative explanation focuses on norm externalization and mutual trust that reduces the likelihood of dispute escalation between two democracies (Doyle 1986; Russett 1993; Dixon 1994; Owen 1994; Raymond 1994; Simmons 1999). Inter-democratic disagreement is expected to be resolved peacefully to a greater degree than other interstate disputes because of norms of peaceful conflict resolution unique to democracy. Democratic leaders are socialized to use peaceful methods for domestic political competitions such as mediation, negotiation and compromise. When some conflicting interests arise between two democracies, these common experiences at the domestic level allow both sides to trust and respect each other. Another strand of norm-based explanations corroborates this conclusion. Owen (1994) argues that the foreign policy of a democracy is more likely to conform to the true interests of citizens, who prefer freedom and peace. By definition, democracies perceive each other as pacific and trustworthy, which leads to a reduction of suspicions and dampens the security dilemma. 22 Second, the initial structural constraint model of the institutional framework emphasizes the difficulties and the costs imposed upon democratic leaders regarding use of military force. This model argues that democratic leaders are constrained from using force more than their autocratic counterparts. This is thought to be so because of the consensus from a broad spectrum of constituencies that democratic leaders need when making foreign policy decisions. First, the democratic citizens are assumed to abhor violent foreign policy because they must bear the costs of war (Kant [1795] 1957; Morgan and Campbell 1991). Second, the complexity of democratic processes and the requirement of consensus make the mobilization of resource for war and use of force slower in democracies than in autocracies (Russett 1993). Third, democratic institutions facilitate the mobilization of opposition. When incumbents undertake unpopular, failed, or costly policies, political opponents may do their best to unseat the officials from the government (Bueno de Mesquita and Lalman 1992). Taken together, these all make the use of force more costly and risky for the democratic leaders and send credible signals to other states of an aversion to using force. While the normative and structural constraint models appear to be effective in explaining the absence of conflict between democracies, they do not explain other related phenomena such as the overall war-proneness of and the successful war records of democracies. The normative explanation is not only tautological in logic (Bueno Mesquita et al. 1999) but also inconsistent with the historical records such as the imperialistic wars of European democracies, American interventions in fellow democracies, and frequent covert military actions conducted by democracies against democracies (James and Mitchell 1995; Rosato 2003). The structural constraint model cannot account for the empirical record that democracies engage in wars as much as autocracies. 23 Several recent efforts to incorporate these empirical patterns related to democracy have been made within the institutional explanation framework. In these efforts, the democratic accountability, public consensus, and political oppositions do not necessarily constrain leaders from use of force; rather these institutional features continue to affect and constrain leaders’ discretion and decision about use of force. In addition, the democratic public is no longer assumed to be more averse to violence. Therefore, the democratic institutional constraint does not necessarily make democratic leaders particularly dovish. Reiter and Stam (2002) provide a concurrent public consent model based on democratic institutions and active roles of voters. As in the constraint model, because of relative ease of leadership removal, democratic leaders are assumed to be more accountable for the policies they make than autocratic leaders. Their foreign policies tend to represent more closely the values and preferences of the people than autocrats’ foreign policies. Especially, the role of the people comes to the fore in this model. The model allows citizens to influence foreign policy decision making from the beginning to the end of conflict. In this model, both of citizens’ ex ante expectation and ex post evaluation matter for leaders’ decision making of war. Democratic states are thought to act on “contemporaneous consent of the people.” As a result, they tend to avoid starting wars expected to be difficult and costly. So when democracies go to war they win and suffer fewer casualties (see also Siverson 1995), and when casualties mount in wars, they seek to quit and settle for draws and losses (see also Bennett and Stam 1998). Bueno de Mesquita et al.’s selectorate theory (1999, 2003) also considers accountability and public evaluation important in decision making of armed force in democracies. However, it differs in the causal mechanism from Reiter and Stam’s model with a distinct view of the public constraints. In this model, voters are assumed to evaluate leaders’ performance mostly based 24 upon policy outcomes. Like other institutional explanations, the selectorate model of war starts from an overarching assumption of reelection-oriented leaders, whether democratic or autocratic. Distinct domestic institutional arrangements, however, generate different political incentives for leaders before, during, and after war. The model holds that political survival of leaders is a function of satisfying members in winning coalitions. In doing so, leaders must be able to adjust to the structure of their political systems. The structure refers to the size of a winning coalition relative to that of the relevant selectorate. The provision of public goods is desirable in political systems with a relatively large winning collation. Political systems with a relatively small winning coalition necessitate the provision of private goods. With respect to foreign policy, losing interstate war may be thought to be a foreign policy failure in both democracies and autocracies alike. However, the selectorate model suggests that the same foreign policy failure produces differing effects depending on the type of political regimes. Compared to autocracies, democracies characterized by a large winning coalition need greater confidence of victory before going to war and put greater efforts to win once engaged in war. By contrast, autocratic leaders with a small winning coalition can compensate relatively easily for foreign policy failures by devoting resources to private benefits for a narrow spectrum of essential supporters. In short, democracies are more careful, selective, and hard-working in starting and conducting military conflict as compared to autocracies. The observable implications include that democracies are unwilling to fight against fellow hardworking democracies, unhesitant to exploit weaker opponents, and prone to produce favorable war outcomes. Centering on the clear information properties of democracy, the information perspective also provides explanations for the separate peace between democracies and prevailing performance of democracies in conflict (Fearon 1994; Siegel 1997; Schultz 2001b; Partell and 25 Palmer 1999; Schultz 2001b; Choi and James 2007). This perspective argues that democratic accountability helps reveal true preferences and resolve conflicting interests before escalation. Hence the reason for frequent peaceful conflict resolutions between democracies is the ability to communicate clearly reducing informational asymmetries that often break down interstate bargaining. This perspective agrees with the concurrent public consent model that assumes people are attentive and influential contemporaneously to foreign policy issues. The key to this explanation is the assumption that international crisis is a public contest where the public watches leaders’ performances and assesses their foreign policy confidence. Within the informational perspective, two nuanced causal mechanisms have been proposed. For instance, Fearon (1994) suggests that the ability to generate high audience costs is a political leverage that democratic leaders take advantage of in interstate crises. The public is assumed to dislike seeing their government retreat in international crises. Therefore, leaders incur audience costs when they make empty threats that they will not carry out later. Interstate understanding of those costs during crises is the key to fathoming opponents’ true preferences about taking further military actions. Military threat that generates large audience costs will be likely to be regarded as genuine. Fearon emphasizes electoral institutions of democracy as a main source of high audience costs. High audience costs require leaders to be time-consistent, meaning that once they initiate a threat, they must carry it out when a target state reciprocates rather than backs off. As a consequence, democratic leaders are constrained from bluffing behaviors but this constraint makes their threat once made look genuine. Schultz (1998; 2001b) also makes a similar argument about democracy, bluffing, and credible signals. But he disagrees with Fearon (1994) on the sources of the clear information. For Schultz, it is not audience costs but transparency that characterizes democratic decision making 26 processes. He emphasizes the role of effective political oppositions within democratic systems that compete overtly for electoral support. Democracies find it difficult to conceal weak public support for war and misrepresent their preferences because of the squealing that strong opposition parties make when the government pursues unpopular military commitments. Conversely, if there is strong public support for war, the opposition parties confirm the veracity of the government’s resolve. In short, democratic competition does not simply constrain a government from use of force but these constraints lead a government to be more selective about making a threat or using force, which in turn improves the effectiveness of its military threats. The Plan of the Dissertation As mentioned above, this dissertation contains three related but distinctive essays. Each has different research questions with different theoretical and empirical focuses. Therefore, separate research designs, though similar in terms of the dependent variable (i.e., MID onset) and the main independent variable (joint democracy), will be employed for the separate analyses of the three essays. Each essay also contains its own detailed literature review, hypothesis generation, data and variable description. In Chapter 2, I assess the common interests-based critique put forth by realists against the democratic peace. The debate has not been resolved because of realists’ failure to provide a direct measure for common interests and liberals’ failure to provide explicit counter-evidence. I believe that it is opportune to explore the realist-liberal debate on the effect of democracy regarding the Cold War given the new release of the relevant data that allows statistical analyses for the 12 post-Cold War years. 27 Chapter 3 concerns the effect of democracy in relation to territorial threat. In order to address this issue, two separate statistical analyses are conducted. One replicates Gibler’s “bordering on peace” analysis. Given the inconsistent findings for his measures for territorial threat and methodological flaws in his design, the other one employs a different statistical model that takes into account the possible interaction between democracy and territory. I argue that based upon territorial claims as a measure for territorial issues and all dyad years as units of analysis, the interactive approach provides a fairer test for the democratic peace vis-à-vis territorial disputes. In Chapter 4, I explore the debate between realists and proponents about the relationship that links democracy to wise conflict selection and advantageous war outcomes. I reveal that because of their overemphasis on war outcomes, theories and empirics provided by both proponents and critics are underspecified in identifying factors that may actually underplay selection mechanisms of conflict initiation unique to democracy. I provide two corrections. First, my analysis includes peacetime decision making by directly looking at initiation military conflict that is short of war. Second, I explicitly identify two kinds of selection factors: one affecting the justifiability of conflict initiation and the other affecting its winnability. I conduct an interactive analysis to test democratic selectivity in terms of winnability and justifiability, based on all the directed dyads from 1981-1999 period for which all the measures are jointly available. In Chapter 5, I conclude with a summary of the empirical findings and discuss their practical and theoretical implications. I also acknowledge the limitations of this dissertation project for future research. 28 CHAPTER TWO FORWARD TO THE FUTURE? DEMOCRATIC PEACE AFTER THE COLD WAR This chapter addresses a critical problem instigated by the ongoing debate between realists and liberals on the underlying cause of peacefulness among democracies: whether and how the democratic peace persists after the Cold War. The research program of the democratic peace has been among the most prominent in the field of international relations (IR) over the last two decades. Numerous studies have looked at the relationship between democracy, peace and conflict, but not a single study concludes that democracy promotes violent conflict, although several reasons for skepticism have been raised (e.g., Gartzke 1998, 2007; Gibler 2007). Indeed, the overwhelming balance of existing evidence seems to confirm the proposition that democracies rarely fight each other (e.g., Dixon 1994; Russett and Oneal 2001; Bueno de Mesquita et al. 2003). However, there is no consensus among students of international conflict as to what really causes this empirical regularity. The debate is especially prevalent between two main camps in the field of IR. Realists argue that the democratic peace is an artifact of the Cold War. Put differently, these skeptics consider the separate peace among democracies as a by-product of the Cold War bipolar politics. In contrast, liberals regard norms and institutions of liberal democracy as the prime causes. For these proponents, the pacifying effect of joint democracy is not limited to a specific constellation of major powers arrayed within the world system from 1950 to 1989. It is hard to judge which side is more convincing based on the existing findings. It appears that democratic dyads had been more peaceful than other combinations of regimes from 29 the Congress of Vienna System to the demise of the Cold War (Maoz and Abdolali 1989; Russett and Oneal 2001). But, Gowa’s series of period-specific analyses (e.g., Farber and Gowa 1997; Gowa 1999; see also Box-Steffensmeier, Reiter, and Zorn 2003) show that the pacifying effect of joint democracy holds only for the 1946-1980 period. I take Gowa’s finding seriously to probe whether and how the effect of joint democracy changes in the post-Cold War era based on a similar research design. In fact, Gowa’s critique remains one of the most influential, as discussed in the next section. To take just one example, by Google Scholar’s count, articles and a book put forth by Gowa against the democratic peace—all of which are published in prestigious political science outlets such as International Organization, International Security, Journal of Politics, 2 and Princeton University Press—have been cited more than 700 times. A recent study by Cederman (2001) gives us a hint on this issue. Cederman’s study reveals that although time-varying across periods, the effect of joint democracy has been on average gaining strength over time. The presumed time-variant effect of joint democracy invites an investigation for the post-Cold War era. If realists like Gowa are right, the effect of democracy should have dissipated or at least substantially weakened after the Cold War. If liberals like Cederman, and Oneal and Russett, who built upon work by Kant, are right, the pacifying effect should be at work beyond the Cold War, and possibly be strengthened relative to conflict proclivities between democracies during the Cold War and before. New data have become available to provide a test of these competing claims. This information comes from new releases of relevant data on international conflict and national polity from the Correlates of War (COW) project and the Polity IV project which have been 2 scholar.google.com, counted on November 4, 2010 using the search word, Joanne Gowa. 30 updated to cover through the year of 2001 and the year of 2009, respectively. 3 Hence, this chapter includes jointly available information from these two projects on the time span from 1816 to 2001, which most importantly includes a decade of post-Cold War information. I examine the effect of joint democracy on interstate armed conflict with special focus on the postCold War era (1990-2001). A set of period-specific probit analyses show that democratic dyads are more peaceful than other combinations of regime types, even for the post-Cold War era. Specifically, the pacifying effect of democracy appears stronger in this period. Also, the measures for Cold War effects and common interests, though having statistically significant effects on militarized disputes, fail to wipe out or even reduce the pacifying effect of joint democracy during the Cold War era. Explaining the Democratic Peace Given the suspected time-varying effect of democracy, the central question is whether or not the democratic peace holds when controlling for the effects of the Cold War bipolar politics. Equally important is the question of whether the democratic peace holds beyond the Cold War era. It has been argued that two democratic states in a dyad should reveal a positive (negative) connection with peaceful (militarized) attempts to resolve conflicts of interests between them. Numerous theoretical and empirical studies have supported that connection. According to the structural constraint model of the institutional explanations, government structures of democracy, characterized by “checks and balances,” draw out the process of decision-making and impose 3 www.correlatesofwar.org and www.systemicpeace.org/polity/polity4.htm (accessed on November 4, 2010) 31 constraints on democratic leaders from using force in international disputes (Morgan and Campbell 1991; Bueno de Mesquita and Lalman 1992; Siverson 1995). When a conflicting interest arises between two democracies, the relatively plodding structural features allow them to find peaceful solutions before enough domestic support builds up in either state for the use of armed force. In addition, political opponents may do their best to mobilize opinion against incumbent leaders’ attempts to use force against other democracies. Other institutional explanations focus on deliberate and careful foreign policy making unique to democratic leaders. Democratic leaders are subject to public evaluation of their foreign policy decisions because of their desire to be reelected and of the relative ease of leader removal through regular elections. Further, a broad spectrum of constituencies such as the legislature, opposition parties, interest groups, and the public also increase domestic accountability in democracies (Fearon 1994; Shultz 2001b; Reiter and Stam 2002; Bueno de Mesquita et al. 2003). Since leaders who appear incompetent in interstate crises will be more likely to be punished in democracies, democratic leaders are reluctant to initiate a fight unless there is a decisive chance of victory, but once involved in war they will fight desperately to ensure victory. When it comes to democratic dyads, a selective democracy will not initiate conflict against other tough-fighting democracies (Bueno de Mesquita et al 2003). This is also consistent with the claim that democratic leaders lack incentives to engage in bluffing behaviors, and their moves in crises are likely seen as resolved and credible (Schultz 2001b). In a dispute between two democracies, these effects are reinforced to produce peaceful conflict resolution before violent escalation more often than in a dispute between other types of dyads (Fearon 1994). The normative explanation stresses norm externalization and mutual trust between democracies (e.g., Dixon 1994; Mitchell 2002). Peaceful settlements are more likely between 32 democratic disputants than other conflicting pairs because of their experience with mediation, negotiation, and compromise at the domestic level (Dixon 1994). These experiences socialize democratic leaders to act on the basis of norms of peaceful conflict resolution whenever possible with, for example, a greater emphasis on the rule of law (Simmons 1999). When two democracies are involved in a dispute with one another, they externalize their domestic political norms because of a conviction that the other state also persists in such norms. Put differently, this conviction makes democracies more likely to trust and respect each other even when conflicting interests arise between them (Owen 1994). In short, students of the democratic peace argue that, due to normative and structural constraints, democratic dyads get along peacefully with each other. Realists have raised important concerns, both theoretical and empirical, about the idea of democratic peace. The democratic peace would undermine the traditional realist claim of the primacy of power and systemic factors: “the important elements of international politics can be explained without differentiating states by regime types” (Lynn-Jones 1996: xi). However, their theoretical arguments against the democratic peace have yet to be exposed to an extensive, systematic empirical testing while most of their empirical critiques have serious methodological problems. Layne (1994) and Spiro (1994) argue that the structural model produces the monadic pacifism as well as the dyadic one (see also Gartzke 2007). They point out that the normative explanation would be a promising candidate that may explain the democratic war proneness with nondemocratic countries, but few empirical studies had examined the democratic peace in terms of normative liberal causes. To test the normative explanation, Layne selected four cases of “near miss” crises in which democracies came close to war but avoided it. Based on these four cases, he concluded that the factors that drove the near misses were not shared democratic norms but 33 traditional realist factors such as calculations of national interests, strategic concerns, military threats, and used threats. Spiro also finds that the probability that the observed rarity of democratic conflicts could have been driven by random chance is substantially greater than 50%. He argues that the statistical significance found for the democratic peace is a statistical artifact of serially correlated data (see also Mearsheimer 1990a). However, these studies follow research design strategies that undercut the ability to make inference from the data at hand. Layne inappropriately selected on the dependent variable, not considering “all the dogs that did not bark” (Russett 1995: 167). Spiro relied on an inappropriate probability model (Russett 1995; Hewitt and Young 2001), and alternative studies show that the relationship still holds when controlling for the duration dependence in the data (Beck, Katz, and Tucker 1998). Some realists share Thompson’s (1996) concern about the democratic peace, “putting the cart before the horse.” For instance, a structural realist, Patrick James, in his study with other colleagues (James, Solberg, and Wolfson 1999) holds that there is reverse causality between democracy and peace, with the causal arrow running from peace to democracy. But the reverse causality critiques at the monadic level have not seriously impaired the dyadic democratic pacifism. In effect, Reiter (2001) shows that peace does not affect the joint level of democracy. Even at the monadic level, Mousseau and Shi (1999) find that nations are about as likely to become more institutionally autocratic as they are to become more democratic in the periods before the onset of wars. Now even Thompson (see Rasler and Thompson 2004), one of the first critics of the reversed causality, admits that democracy promotes peace as well as peace nurtures democracy. The most cogent challenge from the realist camp is Gowa’s research with a renowned economist, Farber (Farber and Gowa 1995, 1997; Gowa 1995, 1999). Gowa (1995) argues that 34 both normative and structural explanations are unpersuasive. She suggests that norms of conflict resolution are not unique to democracies; other types of states also have incentives to resort to such norms because wars are costly options of resolving disagreements (cf., Fearon 1995). In contrast to Layne and Spiro, Gowa regards the checks-and-balances system of democracy as the core explanation for the democratic pacifism. However, she criticizes that the structural explanation is undermined by the well-known propensity of democracies to wage wars against nondemocracies. It should be noted that Gowa does not negate the existence of the dyadic democratic peace per se; rather, she challenges the notion that democracy is the prime mover. Gowa (1999) argues that the democratic peace is a Cold War phenomenon. What has driven the separate peace between two democracies is not democracy at all, but rather common security interests between democracies induced by the Cold War bipolar politics. Gowa (1999) provides evidence that no discernable pattern is found among democratic dyads in reducing armed conflict rates for the pre-World War I era (see also Maoz and Abdolali 1989). She shows that the pacifying effect of joint democracy holds only for the 1946-1980 period. Also, she finds that democracies have allied with each other more than non-democracies for this same period. These results generally support the realist explanation: the Cold War induced strong common or conflicting security interests among members of the international system according to the bloc they belonged to. Indeed, Gowa has provided a more plausible empirical and theoretical challenge than other realists. Proponents of the democratic peace have responded to Gowa (e.g., Oneal, and Russett 1999b; see also Oneal and Russett 2005), but her work still remains “the most significant realist critique of the democratic peace” (Gelpi and Griesdorf 2001: 634). Likewise, to date, many 35 studies on the democratic peace take Gowa’s argument as their null hypothesis (e.g., Schultz 2001b; Gelpi and Griesdorf 2001; Huth and Allee 2002). However, Gowa’s (1999) analysis is less than convincing on its own, because she never provides a measure for common interests that may wipe out the effect of joint democracy. While she argues for alliance as a good proxy measure for common security interests, Gowa does not use it as an explanatory variable. In fact, many statistical studies that include alliance as a predictor of conflict find that the statistical significance of the alliance measure is sensitive to 4 research design and specification choices (e.g., Bennett and Stam 2000b). Besides, the measures for alliance ties and other shared interests such as alliance portfolio similarity (Signorino and Ritter 1999), the compatibility of foreign policy preferences (Gartzke 1998, 2007), and regime similarity (Werner 2000) have failed to eliminate the correlation between democracy and peace (Oneal and Russett 1999b, 2005; Cederman 2001; Russett and Oneal 2001; Bennett 2006; Dafoe 2011). If the democratic peace is really a Cold War phenomenon, it should be better to consider states’ alliance behaviors in regard to the US (or USSR) as a proxy measure for the existence of common or conflicting security interests of the Cold War. This is one of tasks that I seek to do in this chapter. In addition, I will utilize a set of period-specific analyses similar to Gowa (1999), but including the post-Cold War era up to year 2001 in order to examine the suspected null effect of joint democracy beyond the Cold War period. Cederman’s recent study stands out for this issue. Reinterpreting the democratic peace as a Kantian macrohistorical learning process, Cederman highlights Kant’s emphasis on human learning processes that progress in liberal democracies. In envisioning a perpetual peace, Kant 4 James Lee Ray (2005) warns that it is not legitimate to include alliance ties in a statistical model aimed to test the causal impact of democracy on peace, since the inclusion of such an intervening variable can reduce the relationship to statistical insignificance. This is known as post-treatment. 36 (1957 [1795], cf., 1970 [1784]) stressed the gradual diffusion of the democratic learning process from a powerful, enlightened state to its neighbors. This diffusion is an essential step toward building a peaceful federal association of states, a “zone of peace.” Since progress through experiences and learning are a dynamic and dialectical process interrupted with several set-backs, it is not surprising to observe the insignificant effect of democracy in the periods earlier than the Cold War. Consistent with Kant’s vision, Cederman (2001) finds that the conflict propensities between democracies had been gradually falling from 1837 to 1992, and that the pacifying effect 5 is strengthened as their democracy matures. The idea of maturity and neighborhood in terms of the democratic zone of peace may explain why the effects of joint democracy have been found to vary according to different time periods of modern history. So, we cannot yet conclude that the democratic peace is a Cold War phenomenon as realists insist until a thorough examination over the post-Cold War period is conducted. If realists like Gowa and Mearsheimer are right, the pacifying effect of joint democracy should be gone or at least reduced after the Cold War period. But if liberals like Cederman and Oneal and Russett are right, the effect should continue and even be stronger. HYPOTHESES: POST-COLD WAR DEMOCRATIC PEACE The presumed time-varying effects of joint democracy invite another examination for the post-Cold War era. The post-Cold War era should be a fairer period to test the democratic pacifism for both realists and liberals. The demise of the Cold War is a landmark for the attenuation of the East-West divide among major powers—France, the UK, and the US on one 5 Intriguing is one of Cederman’s (2001) findings that the conflict propensities among autocracies also exhibit a pacifying trend over time, but with a slower rate compared to democracies. 37 side, and the USSR and China on the other side. From the perspective of realists, it is the EastWest rivalry that has produced a large number of disputes of democracies against nondemocracies and/or nondemocracies against nondemocracies. The end of the Cold War also came with a remarkable democratization trend over the globe. The increased proportion of democracies in the post-Cold War era ensures that democracies have democratic neighbors, reducing the realist concern that the rarity of inter-democratic conflict is a function of the sparse dispersion of democracies during most of the 19th and 20th centuries (Mearsheimer 1990a; Spiro 1994). For liberals, the increased proportion means a more democratic world system. Enough democracies are required for their peace-inducing effect to be fully effective (Kant 1957 [1795]; Cederman 2001). Figure 2.1. Time Trend of Global Average Democracy Level, 1816-2001 38 Figure 2.1 displays the trend of global average democracy level for the 1816-2001 period. I calculate the level of global democracy scores by dividing the sum of all countries’ Polity2 scores by the number of system members (i.e., countries) in a given year based on the Polity IV and the COW interstate system membership data. Overall, there is an upward trend of global democratization over time with several interrupted setbacks as Kant expected. Indeed, countries in the system appear to have followed the dialectical, teleological, and macrohistorical democratization trend. Especially, a sharp increase in the average democracy score around 1989 reflects the appearances of many new democracies after the Cold War. As we can see, the postCold War era is the most democratic in history. In what follows, I specify hypotheses drawn upon the conflicting arguments between the two paradigms. Realist Hypotheses As Gowa (1999) suggests, the Cold War generated a unique security interest pattern among countries. In fact, the Cold War was a long-lasting conflict between the Western and Eastern Blocs. At the top of each bloc there were two superpowers, the US and the USSR. Whether countries had conflicting or shared security preferences with each other largely depended on which bloc they jointly or separately belonged to. Furthermore, during the Cold War, almost all the democratic countries belonged to the Western Bloc, and few (if any) countries were democratic in the Eastern Bloc. Therefore, the democratic peace might be epiphenomenal to the security interest pattern that the Cold War bipolar politics exerted. This suggests that we may no longer observe the pacifying effect of joint democracy for the post-Cold War period. Other realist critics have provided similar expectations. For instance, Mearsheimer (1990a) argues that the bipolarity of the Cold War order, not the prevalence of democracy in the 39 region, had produced unprecedented stability and peace in Europe, central to the international system. It is the East-West enmity and balance of power that gave rise to such order. He predicts that the demise of the Cold War order would increase the chance of wars and major crises in Europe. He also attributes the separate peace among democracies to the fact that democracies had been few in most of the 1800s and 1900s, and thus few (if any) democracies were geographically positioned to have a high chance of fighting with each other. Recently, Rosato (2003) criticizes each of the theoretical stories of the democratic peace. He asserts that the democratic peace is an imperial peace based on American power and thus a post-World War II phenomenon restricted to Western Europe and Americas. He points out that the US had been the dominant power in these regions that had placed an overriding emphasis on regional peace. Certainly the presence of the Soviet threat accounts for such emphasis. In the absence of such threat, the US may not have the particularistic incentives to control these regions to the same extent as it did in the Cold War era. As mentioned in the previous section, realists have never explicitly controlled for the Cold War effects that could wipe out the effect of joint democracy once they are put together into one statistical specification. I employ a measure for Cold War bipolar rivalries based on states’ alliance making with the superpower leader of each bloc. If it is really the East-West divide that explains the peace between democratic countries and the uproars among other types of dyads, the presumed effect of joint democracy should be washed out with the control for the Cold War rivalries. Taken all together, the realist critiques give us the following two hypotheses. Realist Hypothesis 1: During the Cold War period, the effect of joint democracy will be “indistinguishable from zero” with the inclusion of a measure for the East-West Cold War divide. 40 Realist Hypothesis 2: After the demise of the Cold War, the pacifying effect of joint democracy, spuriously found in the Cold War period model without an appropriate control for the Cold War bipolar conflict, will be significantly reduced or “indistinguishable from zero.” Liberal Hypotheses Liberals do not dispute the swaying influence of the East-West conflict in the Cold War international system. Certainly, many regional disputes had been driven by the superpower competitions to protect and maintain interests in their preferred regions. However, proponents of the democratic peace do have their own reasons to believe that democracy is a prime factor to encourage peace in jointly democratic interstate relations independent of the Cold War politics. Normative and institutional characteristics of democracy have been proposed as rationales that underpin the dyadic pacifism. Liberal Hypothesis 1: The pacifying effect of joint democracy will still hold for both of the Cold War and post-Cold War periods regardless of the inclusion of a measure for the Cold War interests. One difficulty of this expectation is that the effect of dyadic regime types is found to be inconsistent with the democratic peace theory in the pre-World War I period (Maoz and Abdolali 1989; Gowa 1999). Some defenses for the critique of the time-inconsistent effect of democracy can be found in the liberal camp. The forefather of the democratic peace, Kant, reasoned that perpetual peace can be achieved as a consequence of progress through learning. Individuals will realize that war is both disastrous and immoral. This realization will be translated into state 41 realizations. But, the transfer of realization is feasible only in particular institutional and constitutional settings. Learning can be substantively developed and processed in liberal democracies. However, the monadic learning process is not sufficient for the perpetual peace if there are many autocratic neighbors. According to Kant (1957 [1995]: 13), war is the easiest thing for unconstrained autocratic leaders to declare with no “sacrifice of the pleasures of his table, the chase, his country houses, his court functions, and the like.” To achieve the perpetual peace, there must be a gradual diffusion of democratic learning process from a powerful, enlightened state to others to constitute the democratic zone of peace. Since such progress is essentially based on dynamic and dialectical learning, it is reasonable to expect that the pacific effect will be deepening over time as existing democracies mature and more democracies appear over time. In support of Kant’s conjecture, Cederman finds that the conflict propensities of democratic dyads have steadily fallen from the late 1830s to the early 1990s. Other studies also suggest that there are deepening and widening effects of the democratic peace (Gleditsch 2002a; Mitchell 2002; Mitchell, Kadera, and Crescenzi, 2008). The idea of maturity and neighborhood in terms of democracy can explain why previous analyses have produced time-varying effects of democracy. As the proportion of democracies increases and the experience of democratic interactions expands over time, the pacifying effect of democracy is reinforced. Recent research above the level of the dyad offers support for the idea of pacific democratic neighborhood (Gleditsch 2002a) or even system-level effects (Mitchell 2002), in which democratic norms become more pervasive as the fraction of democracies in the system increases, to the point where even the behavior of non-democracies begins to change in a more desirable direction (see also Mitchell et al. 2008). Therefore I hypothesize: 42 Liberal Hypothesis 2: In the more democratized post-Cold War era, the pacifying effect of joint democracy will not only persist but will be strengthened. RESEARCH DESIGN My analysis of the post-Cold War democratic peace is based on the standard conflict model found in the literature (e.g., Gowa 1999; Russett and Oneal 2001; Gartzke 2007), where the common dependent variable is militarized interstate disputes (MIDs), the units of analysis are dyad-years, and the measure for joint democracy is based on the Polity project. I try to maintain a fair balance between representative quantitative studies of both realist and liberal camps (e.g., Russet and Oneal 2001 and Gowa 1999). Gowa’s specification, in general, is simpler than that of Russett and Oneal. Unless Russett and Oneal’s schemes are thought to be fair improvements over 6 Gowa’s, I opt for the latter. My sample includes all possible dyads throughout the world as Lemke and Reed (2001) recommend. Although Russett and Oneal prefer the sample of politically relevant dyads as having a reasonable chance of conflict, they have often used the all-dyad sample as well (e.g., Oneal and Russett 1999a; Oneal, Russett, and Berbaum 2003). In contrast, Gowa never has used the limited sample, because nothing in the democratic peace theory implies for this limited set of dyads and the criteria for relevant dyads are unsuccessful to fully identify states at risk of conflict 6 As robustness checks, I also reexamine the data with the alternative specifications (i.e., politically relevant dyads, one-year lag, and dummy for joint democracy of well-established democratic dyads). These, in general, produce more supportive results for the liberal hypotheses than the ones reported in the main analysis. These results will be briefly discussed in the analysis section and reported in the web appendix. 43 7 (see footnote 5 in Gowa 2009). Russett and Oneal often utilize the one-year lag in order to ensure that a dispute to be explained does not affect the independent variables to explain. However, like Gowa, I do not use a one-year lag specification for all independent variables, because setting a lagging point not only is arbitrary but also necessitates a loss of information in the data that amounts to the lagging point. All the information on the dependent and independent variables are jointly available from 1816 to 2001. This is the most overarching time span of statistical analysis of the democratic peace to my best knowledge. I define the Cold War spanning over the 1950-1989 period and the post-Cold War over the 1990-2001 period. This way of disaggregating the data is in line with 8 Gowa’s method (1999). Yet, since her period-specific analysis did not go beyond 1980 due to the data availability at the time of her analysis, I decide the year for the end of the Cold War on my own. I regard 1989 as the year in which the Cold War ended. The Cold War came to an end with a set of dramatic events around 1989. Especially, from 1989 to 1991, its demise was manifest with the crumbling of the Berlin Wall, the collapse of Communist party dictatorship in Eastern Europe, and the disintegration of the Soviet Union. Also, in this year the sponsor countries of each bloc, the US and the USSR, jointly singed a defense pact. I use the EUGene software program (v.3.2) to generate data for both dependent and independent variables (Bennett and Stam 2000a). For the dependent variable, I use the MID data 7 It has been brought to my attention in revising an earlier version of this chapter that Gowa has a new manuscript on the post-Cold War democratic peace (December 2009). In this new study, Gowa (2009: footnote 5) explicitly and correctly points out this problem. In my sample, about 15% of all Cold War MIDs and all post-Cold War MIDs alike occurred between politically irrelevant pairs. 8 While Gowa used to code 1946 as the start year of the Cold War in her previous analyses (Farber and Gowa 1997; Gowa 1999), in her new study, she codes 1950 as the first Cold War year and 1990 as the first post-Cold War year in her new study. I follow this new coding herein. 44 (v.3.1) of the Correlates of War (COW) project. My dependent variable is onset of MIDs, defined as threat, displays, and actual use of armed force by one state against another. If there is a MID onset between two countries (i.e., a dyad) in a given year, it is coded 1, otherwise 0. Unless a new MID occurs, I treat ongoing MID years between the same disputants as missing following Bennett and Stam’s (2000a) advice: treating ongoing years as new dispute years violates the statistical assumption of independence and the decision to start a dispute can be “quite different” from the decision to continue it. Another difficulty in coding MID onset arises from late comers who joined a MID after the first day. I drop all joiner dyad years as long as no direct dispute occurs between these dyads, since “the process of making a decision to join an ongoing dispute is not the same as making a decision to initiate a dispute” (Bennett and Stam 2007: 69). Yet, including joiner dyads does not change the statistical results of this chapter substantively. I use the Polity IV dataset to compute the joint level of democracy for a dyad in a given year. Polity2 score is used to facilitate the time series analyses of this chapter. This score ranges from 10 (a fully democratic dyad) to –10 (a fully autocratic dyad). As it is a standard practice, I take the score of the less democratic state in every dyad as the continuous dyadic measure for joint democracy (Russett and Oneal 2001). I call this variable “LOWEST DEMOCRACY.” It should be noted that unlike Russett and Oneal, Gowa (1999) uses a dichotomous measure for joint democracy with the cutoff of polity score 6. While each has its own advantages and disadvantages, here I prefer the continuous index to the dichotomous dummy for the following reasons. First, joint democracy is the liberals’ variable, not the realists’. Therefore, in this balanced investigation, unless there are critically comparable disadvantages, it would be fairer to use the proponents’ measurement than the critics’. Second, and more importantly, while setting an arbitrary cutoff point is unavoidable and even understandable for the dichotomous 45 joint democracy measure (Jaggers and Gurr 1995), this arbitrariness may lead to critically different results for my period-specific analyses, especially the post-Cold War equation. The conventional cutoff is mostly Polity score 6 or 7, coding 1 for dyads in which both states have the threshold score or higher and 0 otherwise. The problem is that many new democracies appeared near after 1989, and the Polity project assigns 6, 7, or 8 to most of the new regimes. According to the logic of the Kantian learning, the newly independent democracies are hardly distinguishable instantly from their old regimes and other entrenched autocracies in foreign policy interactions due to the lack of elapsed time for democratic learning and the possible 9 instabilities associated with the regime transition. For instance, newly democratized countries’ improved polity scores (e.g., Haiti’s polity score 7 in 1990 and Zambia’s polity score 6 in 1991) may not immediately take on effects in pacifying interstate relations. Besides, I observe that old communist countries and the former Soviet republics account for the majority of the new 10 democracies around 1989 according to the Polity 4D data. The use of the continuous index cannot fairly settle all the questions regarding these difficult cases, but it is fairer than its 9 There is a scholarly consensus that regime transition tends to produce external conflicts. But it is unclear whether this trend is more apparent for democratic transition than autocratic one. For instance, Mansfield and Snyder (e.g., 2005) have argued and found that democratizing states are more likely to experience external wars and conflicts than other stable regimes experiencing no regime change whatever democratic or autocratic. However, as mentioned in the previous chapter, other scholars have criticized that autocratization is at least as much as conflictgenerating as democratization, pointing out the regime instability, instead of democratization, as the main problem (Enterline 1996, 1998; Maoz 1998; Ward and Gleditsch 1998; Braumoeller 2004). For a more recent critique that focuses on Mansfield and Snyder’s data, measurements, model specifications, and case selection, see Narang and Nelson (2009). 10 The examples include Poland (polity score 8 in 1991), Czechoslovakia (8 in 1990), Slovak Republic (7 in 1993), Macedonia (6 in 1991), Bulgaria (8 in 1990), Moldova (7 in 1993), Romania (8 in 1996), Russia (6 in 1992), Estonia (6 in 1991), Latvia (8 in 1991), Ukraine (6 in 1991), Belarus (7 in 1991), and Armenia (7 in 1991). 46 counterpart in assessing the post-Cold War effect of joint democracy. At least partly, the former recognizes the difference between “fledgling” democracies and “fully-fledged” ones. As mentioned above, the main reason that Gowa’s analysis is unconvincing is that it lacks a measure for the effect of the Cold War. I employ a set of dummies to control for the effects of the Cold War, based on the formal alliance data (v.3.0) of the COW project. Yet, I do not simply employ the usual alliance variable that codes 1 for any dyads sharing an alliance (i.e., defense pact, neutrality, or entente) and 0 otherwise. Though often appearing in the literature as a regressor, this simple alliance measure alone may not well capture the unique pattern of security interests driven by the Cold War. As mentioned above, the effect of the alliance measure is found to be not robust, and some raise a concern that this measure may be an intervening variable rather than a legitimate independent variable in the democracy-peace relationship (Ray 2005). With respect to the Cold War effects, we should take into account the contrasting patterns of alliance-making behaviors between two countries. In effect, it may not necessarily be that allies fight each other less frequently than do non-allies. In fact, non-alliance does not indicate the presence of conflicting security interests between two countries. Rather, two countries that have a high chance of military fighting sometimes need to jointly sign ententes, mutual neutrality, or defense pacts to prevent a possible full-scale clash between them. Besides, even conflicting rivals are likely to be allied, when a third party with increasing military capabilities begins to pose a serious threat to the security of both countries. Given that the term “Cold War” is a metaphor for the conflict between the Western and Eastern Blocs, I consider alliance behaviors in regard to the Cold War rivalry. I employ a dichotomous measure for the Cold War rivals (CWR) to capture the presence of conflicting interests between two countries. The coding process is as follows. I look at whether two 47 countries in a dyad belong to the different blocs through formal alliance with the Cold War super powers, the US and the USSR. CWR is coded 1 if two countries in a dyad are allied separately and differently to the bloc leaders, otherwise 0. The zero indicates absence of the conflicting security interests that the Cold War had imposed upon interactions between two states. More specifically, if one country in a dyad was allied with the US but not with the USSR while the other was allied with the USSR but not with the US, this dyad constitutes a CWR. The examples include such dyads as the UK vs. East Germany (1956-1989), Spain vs. Iraq (1972-1989), and Denmark vs. Poland (1949-1989). For those dyads that include one of the Cold War super powers, I code 1 for CWR if a non-superpower side was allied not with its dyadic superpower partner, but rather with the other super power. For example, CWR is coded 1 for the US-North Korea dyad and the USSR-South Korea dyad, since North Korea was allied not with US, but with the USSR, and South Korea not with the USSR but with the US. In my sample, CWR dyads that constitute less than 7.5% of all dyad years account for more than 15% of MIDs in the 195011 1989 periods. I also employ three other dummies based on the COW alliance information. First, realist critics’ main point is related to the observation that most democracies had been allies with each other as part of the US superpower alliance structure during the Cold War (e.g., Mearsheimer 1990a; Gowa 1999; Rosato 2003). This suggests inclusion of a dummy that indicates whether two allies shared the US as an ally. I call this variable “US ALLY.” I code 1 for these dyads and 0 otherwise Second, allied countries that shared the Soviet ally might share similar security 11 One difficulty involved in coding of CWR may arise with dyads in which two countries were allied contrastingly with different bloc leaders while they were allied with each other. This is a valid concern but does not pose a problem for this chapter’s analyses. In my data, for the Cold War period, such cases include only 5 dyad-years between Turkey and Afghanistan from 1951 to 1955. 48 12 interests. I call this dummy “SOVIET ALLY.” Lastly, I code 1 for purely-dyadic allies that were not under one or both of the super power alliance structures and 0 otherwise. I call this dummy “DYADIC ALLY.” Taken together, the results for these dummies are interpreted in relation to the excluded reference category of non-allied dyads that had no Cold War rivalry. Following Russett and Oneal (2001), I employ more extensive controls than Gowa (1999), including capability ratio and distance as well as contiguity and major power status. All of these are found to be important and robust factors for MIDs in many quantitative studies. Previous research shows that the greater the difference in capabilities between two states, the lower the likelihood that they will be involved in a MID. Therefore, I include the capability ratio for each dyad. I label this variable as “CAPABILITY RATIO.” I use the COW Composite Index of National Capabilities (CINC, v.3.02). As is standard, CAPABILITY RATIO is a natural logarithm of the higher/lower ratio between states A and B. Geographic proximity matters to MIDs. Following Russett and Oneal’s practices, I consider two related but different factors, contiguity and distance between two states. The first measure is “NONCONTIGUITY” that equals 1 unless two states share a land boundary or are separated by less than 150 miles of water. I also employ the natural logarithm of the great circle distance in miles between the capitals of the two states. I label this variable as “DISTANCE.” These two measures are not highly correlated in my sample (correlation = -0.47). I expect that the more distance between two countries, the less chance they have of military conflict because of geographical constraints in exercising military force over long distances. 12 It is possible that allied dyads shared not just one but both of US ally and Soviet ally. In the Cold War data (1950-1989), such cases include only 6 dyad-years between France and UK from 1950 to 1955. 49 The effect of distance to constrain militarized options may be less for the major powers with the land, sea, or air capability to exercise military power over long distances. Many empirical studies have found that a dyad that includes at least one major power is more conflictprone than a minor-minor power dyad. I create a dummy variable that identifies whether a pair of states consists of two minor powers. I label this variable as “MINOR POWER.” MINOR POWER is coded 1 for minor-minor power dyads and 0 otherwise. I expect that a dyad consisting of only minor powers is less likely to engage in MIDs than dyads of minor-major powers and majormajor powers. In addition, since the data set is structured in the form of a cross-sectional time-series, serial correlations across dyad years may contaminate the data analyses. Thus, I opt to use the peace year correction in order to test and solve for deterministic temporal dependence as suggested by Beck, Katz, and Tucker (1998). However, I use the adjusted peace years developed by Suzanne Werner (2000) to account for the timing of pre-1816 armed conflict. Also, I do not use Beck et al.’s cubic splines because of arbitrariness about setting knots. Rather, as suggested by Carter and Signorino (2010), I use polynomials of peace years (i.e., t, t2, and t3). Carter and Signorino’s Monte Carlo study shows that this method can perform as well as cubic splines and even better if knots are misspecified. The polynomial corrections are statistically significant in all models reported below, suggesting there is strong evidence for temporal dependence in the form of a nonlinear deterministic trend. Heteroscedasticity is another concern for the crosssectional time series data. I follow Beck et al.’s recommendation to use robust standard errors. Taken together, except for the Cold War interest measures, employed to account for the Cold War effect, all the variables and statistical processes employed are common in the literature. What is 50 new is the additional data, the controlling for the Cold War effects, and the specific focus on the Cold War versus post-Cold War marginal effects of joint democracy. RESULTS AND ANALYSIS In order to compare the effects of joint democracy over the different eras of the international system, I run a set of period-specific probit models for the entire (1816-2001), Cold War (1950-1989), and post-Cold War periods (1990-2001), respectively. I test the effects of joint democracy and all independent variables in two specifications, one without the Cold War dummies (Model 1 hereafter) and the other with (Model 2 hereafter). Table 2.1. Predicting MID Onset, the entire period, 1816-2001 Variables Model 1 (1816-2001) Model 2 (1816-2000) LOWEST DEMOCRACY -.0068*** (.0028)+ -.0080*** (.0027) CWR .3363**** (.0505) US ALLY .0847 (.0848) SOVIET ALLY -.3279**** (.0901) DYADIC ALLY -.1124** (.0630) CAPABILITY RATIO (logged) -.0552**** (.0140) -.0520**** (.0137) NONCONTIGUITY -1.012441**** (.0603) -1.0290**** (.0592) DISTANCE (logged) -.1081**** (.0237) -.1187**** (.0227) MINOR POWERS -.7624**** (.0534) -.7558**** (.0518) T1 -.0373**** (.0034) -.0371**** (.0033) T2 .0004**** (.0001) .0004**** (.0001) T3 -.0000**** (.0000) -.0000**** (0.0000) Constant .0708 (.1565)) .1303 (.1579) N 549220 536990 2 1489.94**** 1562.55**** Wald Chi 2 .2859 .2891 Pseudo R Log pseudo-likelihood -10838.883 -10640.398 Significance levels: *p<0.1, **p<0.05, ***p<0.01, ****P<0.001 (one-tailed). +Robust standard errors are in parentheses. 51 The Entire Period (1816-2001) In Table 2.1, the effects of all the independent variables appear significant in the hypothesized directions regardless of the model specifications. The results show that democratic dyads, skewed capability ratio, noncontiguity, long distance, and the absence of major power reduce the likelihood of MID onset. Also, alliance making in association with the system leaders appear to be important. The coefficient for CWR is positive and statistically significant even in the entire period (1816-2000). 13 This result suggests that two states tend to fight more often when they are contradictory in alliance making with the system leaders. The bloc power-induced interests appear to affect MID onset negatively only in the Soviet side. Unexpectedly, the coefficient for US ALLY is positive though not statistically significant. By contrast, SOVIET ALLY and DYADIC ALLY significantly abate MID onset. In relation to joint democracy, adding the Cold War alliance measures does not make any qualitative change in the pacifying effect of joint democracy. The overall results are supportive of liberals rather than realists. However, the conclusion has yet to be made for the Cold War and joint democracy variables because the data analyzed in Table 2.1 are unspecific to the Cold War and post-Cold War periods unlike the hypotheses. The Cold War Era (1950-1989) In the first two columns of Table 2.2, all the results are largely consistent with those of Table 2.1. Regarding the Cold War dummies, Cold War rivalries, Soviet allies, and purely-dyadic allies have significantly different conflict propensities than non-allies. In comparison, there is 13 The COW alliance data is available up to 2000. So the model with the alliance and CWR variables covers up to 2000. 52 little difference between US allies and non-allied dyads; Soviet allies, followed by purely dyadic 14 allies, are found to be the most peaceful; and Cold War rivals appear the most conflictual. Consistent with Gowa’s claim, alliance ties, especially the ones related to the Cold War bipolar politics, appear as a good indicator with which to tell whether two countries in a dyad share security preferences. However, the general results are not supportive of the realists’ critique against the democratic peace. First, allied dyads under the US alliance structure were not distinguishable from non-allied ones in conflict proclivities as evident in the second column. This suggests that the alliance structure of the US Bloc, thought to be formed and operated against the Soviet Bloc, was far from being the hidden driving force for the democratic peace. Second, the significant negative coefficient for LOWEST DEMOCRACY suggests that joint democracy discourages MIDs. The Cold War dummies, though important per se, do not outperform LOW DEMOCRACY, failing to make its effect inconsequential as shown in the second column. The findings largely carry two implications. First, when properly specified, the alliance information can be useful as an important indicator of common and conflicting interests that influence interstate relations. For instance, the CWR dummy helps reveal that the Cold Warproduced tension between the US allies and Soviet allies tends to produce armed conflicts. Also, the US ALLY and SOVIET ALLY dummies allow finding that whereas the security interests shared among the Soviet allies reduced armed conflicts among them, this was not the case with the US allies. Second, unlike the realist speculation, joint democracy was an effective conflict-reducing factor independently of the Cold War effects. In sum, realists are right to point out that the Cold 14 This statement is made with statistical basis. The rank order was computed by comparing the estimated coefficients for the Cold War dummies. I used the lincom command in Stata to test whether these differences are statistically significant. 53 War shaped interstate relations to an important degree, but wrong to suggest that the democratic peace is a byproduct of the Cold War bipolar politics. Table 2.2. Joint Democracy Effect on MIDs during and after the Cold War Cold War Post-Cold War (1950-1989) (1990-2001) Variables Model 1 Model 2 Model 1 Model 2 -.0123*** -.0137*** -.0155*** -.0146*** LOWEST DEMOCRACY (.0048)+ (.0048) (.0050) (.0053) .3107**** .3358**** CWR (.0769) (.0918) .0428 -.0492 US ALLY (.1146) (.1562) -.7524**** -.2942* SOVIET ALLY (.1560) (.1819) -.1999** .0480 DYADIC ALLY (.0863) (.1212) CAPABILITY RATIO -.0915**** -.0846**** -.0996**** -.0952**** (logged) (.0193) (.0191) (.0236) (.0237) -1.1561**** -1.2102**** -1.040**** -1.013**** NONCONTIGUITY (.0742) (.0724) (.0920) (.0905) -.1572**** -.1601**** -.2424**** -.2701**** DISTANCE (logged) (.0302) (.0298) (.0369) (.0375) -.9157**** -.8772**** -.8168**** -.8144**** MINOR POWERS (.0769) (.0749) (.0893) (.0890) -.0760**** -.0743**** -.0385**** -.0367**** T1 (.0081) (.0077) (.0060) (.0059) .0013**** .0013**** .0006**** .0005**** T2 (.0002) (.0002) (.0001) (.0001) -.0000**** -.0000**** -0.0000*** .0000*** T3 (.0000) (0.0000) (0.0000) (.0000) .8736**** .8647**** 1.0841**** 1.2313**** Constant (.2138) (.2326) (.2514) (.2643) N 290740 290740 144861 132631 Wald Chi 2 1076.53**** 1217.09**** 893.78**** 896.65**** 2 .3686 .3768 .3466 .3461 Pseudo R Log pseudo-likelihood -4481.9337 -4423.7066 -1781.713 -1649.3577 Significance levels: *p<0.1, **p<0.05, ***p<0.01, ****P<0.001 (one-tailed). +Robust standard errors are in parentheses. 54 The Post-Cold War Era (1990-2001) The third and fourth columns of Table 2.2 contain the results for the post-Col War analysis. How does LOWEST DEMOCRACY perform in the post-Cold War equation? If realists are right, then joint democracy should be substantially less effective and weakened, or its effect may be vanished away. If liberals are right, the effect should hold and may be strengthened. The overall results in the post-Cold War equation remain intact, except for US ALLY and DYADIC ALLY with the reversed insignificant coefficients. CWR is still strongly significant at the 0.01 level while SOVIET ALLY becomes only marginally significant at the 0.1 level (one-tailed). These results indicate that even after its demise some remnants of the Cold War continue to affect interstate interactions. Again, LOWEST DEMOCRACY has a negative, significant effect on MID onset in the post-Cold War era. To compare its period-specific effects, joint democracy seems to have a higher influence in the post-Cold War era than the Cold War era. For instance, the coefficient for LOWEST DEMOCRACY changes from -.0123 or -.0137 in the Cold War equation to -.0155 or -.0146 in the post-Cold War equation. However, these coefficient changes are not statistically meaningful. Wald test performed by Seemingly Unrelated Estimation (SUE) (Weesie 1999)—using Stata commands, suest and test—fails to reject the null hypothesis of the same coefficients of LOWEST DEMOCRACY across the two periods (p = 0.5830 for Model 1 and p = 0.8918 for Model 2). Another set of tests (not reported) based on multiplicative interaction models also produce the qualitatively same results. Probability Changes Simple comparisons of the coefficients may not convey appropriate substantive meanings. More substantive discussion is followed in terms of discrete change: the change in the predicted 55 probability of the dependent variable according to a change in an independent variable of interests. I conduct the analysis of probability changes using the CLARIFY program (King, Tomz, and Wittenberg 2000). I focus on LOWEST DEMOCRACY to show the substantive effect of joint democracy, which is of the primary importance in this chapter. To do this, it is required to set a baseline probability against which to make comparison. I hold all continuous measures at their mean. For dummy variables, I set the dyad to be conflict-prone to see how joint democracy pacifies this conflict-prone dyad (CWR=1, other Cold War dummies=0, NONCONTIGUITY=0). MINOR POWER=0, and I then estimate the annual probability that this “typical” dyad would experience a MID onset. Table 2.3 Probability Changes in Risk for Annual MID onset according to Joint Democracy Cold War Period -1/2 sd. +1/2 sd. First Diff. % Δ of Prob. Dummies Absent .0810 .0720 -.0090 -11.1% (Model 1) (.0550, .1152)+ (.0486, .1025) (-.0165, -.0022) Cold War Present .1412 .1261 -.0151 -10.7 % (Model 2) (.0962, .1949) (.0843, .1748) (-.0259, -.0044) .0477 (.0314, .0701) -.0105 (-.0199, -.0036) -18.1% Present .1011 .0860 (Model 2) (.0615, .1522) (.0517, .1279) +The 95% confidence intervals are in the parentheses. -.0151 (-.0293, -.0034) -14.9% PostCold War Absent (Model 1) .0582 (.0374, .0854) Table 2.3 gives the differences in the probability of MID onset and the percentage increase or decrease in the annual risk of a typical dyad being engaged in a MID according to a change in its joint democracy score (as LOWEST DEMOCRACY moves from –1/2 standard deviation to +1/2 standard deviation around its mean). In the Cold War era, LOWEST DEMOCRACY decreases the probability of MID onset by about -.0090 based on Model 1 (without the Cold War dummies). 56 This probability change amounts to an 11.1% decrease and is statistically significant since the 95% confidence interval of the first difference does not include zero. The effect of DEMOCRACY LOWEST is still present even when the Cold War effects are taken into account. It amounts to a 10.7% reduction in the probability of MID onset. The confidence interval for the first difference (-.0151 in the second row) based on Model 2 (with the Cold War dummies) does not include zero. The end of the Cold War has not brought the end of democratic pacifism. Rather the joint democracy has dampened MIDs at a greater rate in the post-Cold War period. Joint democracy decreases MID onset by -.0105, an 18.1% reduction. Controlling for the Cold War effects, joint democracy reduces MID onset by -.0151, a 14.9% reduction. These first differences are statistically significant given that none of the relevant confidence intervals includes zero. In comparison, the effect of joint democracy becomes 1.6 times (Model 1) or 1.4 times (Model 2) stronger in the post-Cold War era. In sum, the results suggest that the pacifying effect of joint democracy still holds even after the Cold War. In fact, against the realist expectations, the peaceful effect for joint democracy is actually estimated to be larger in the post-Cold War era. True, the difference between the consistently negative effects of joint democracy across the two periods are relatively small and not statistically significant, but they trend in liberals’ predicted direction. In any case, we cannot conclude that the democratic peace is a Cold War phenomenon as realists hold. Rather, the democratic peace has proceeded forward to the post-Cold War era. Robustness Checks 57 I conduct three robustness checks with alternative specifications and samples found in the literature (see Tables 2.4 and 2.5). All of the alternative analyses generate more favorable results for the Kantian thesis. First, I reanalyze the data with politically-relevant dyads, defined as those that are contiguous (by land or separate by less than 400 miles of water) or include at least one major power. The consistently larger negative coefficient for LOWEST DEMOCRACY in the post- Cold War era than the Cold War era becomes statistically significant. In terms of probability changes, in Model 1 (without the Cold War dummies) while joint democracy decreases MID onset by 8.4% in the Cold War period, the equivalent probability reduction is 27.8% in the postCold War period. Similarly in Model 2 (with the Cold War dummies), the magnitude of probability reduction increases substantially from 13.6% in the Cold War era to 25.6% in the post-Cold War era. Second, I incorporate a one-year lag for all of the independent variables like Russett and Oneal (2001) to account for the possible reciprocal relationship between the independent and dependent variables. Again, this alternative specification produces results in favor of the Kantian thesis. While the results for all the control variables remain very similar to the previous ones, the coefficient for LOWEST DEMOCRACYt-1 increases in absolute number by about 2.2 times (Model 1) or 2.3 times (Model 2) in the post-Cold War equation. (In the Cold War equation, its coefficients are estimated similarly to the ones in the main analyses.) This coefficient difference is statistically significant at the 0.01 level in both Models 1 and 2. In terms of probability changes, LOWEST DEMOCRACYt-1 decreases MID onset by -11.9% (Model 1) or -10.4% (Model 2) in the Cold War era. This conflict-reducing effect becomes more substantial in the post-Cold War era with a 31% (Model 1) or 30.0% reduction (Model 2). This more favorable result is not the function of the sample that is different than the one in the main analysis due to the one-year lag. 58 Table 2.4. The Robustness Checks for the Time-Variant Pacifying Effects of Joint Democracy Robustness Checks Cold War Wald Test Coefficients Dummies Coeff. Δ Cold War Post-Cold War (Post – CW) Absent -.0068* -.0212**** -.0144*** 1.LOWEST DEMOCRACY (Model 1) (.0047)+ (.0057) (Relevant Dyads) Present -.0124*** -.0232**** -.0108* (Model 2) (.0052) (.0062) Absent -.0136*** -.0301**** -.0165*** 2. LOWEST DEMOCRACY (Model 1) (.0049) (.0058) (One-Year Lag) Present -.0140*** -.0328**** -.0188*** (Model 2) (.0049) (.0060) 3. Dummies++ Absent -.4783**** -1.2698*** -.7915*** (Model 1) (.1274) (.2410) Joint Democracy Present -.5128**** -1.4353**** (Polity 9 or 10) -.9225**** (Model 2) (.1320) (.2691) Absent -.2163**** .0380 .2543^^ Joint Democracy (Model 1) (.1070) (.0685) (Polity 6, 7 or 8) Present -.2354** .0285 .2639^^ (Model 2) (.1054) (.0727) *p<0.1, **p<0.05, ***p<0.01, ****P<0.001. ^p<0.1, ^^p<0.05, ^^^p<0.01, ^^^^P<0.001, wrong direction. +Robust standard errors are in parentheses. ++The two dichotomous dummies for joint democracy are included together. Therefore, the reference category is non-democratic dyads whose joint democracy score is less than 6. MID onset is regressed by the non-lagged independent variables using the same sample used for the lagged analysis. With this sample, the statistical significance for the enhanced effect of joint democracy is absent as in the main analyses. The apparent lagging effect in the post-Cold War era may be associated with new democracies, as discussed in the previous section. Joint democracies involving at least one new democracy may not take on the beneficial effects immediately. Setting a lagging point is arbitrary. I have tried different lagging points separately from one-year lag to five-year lag. In general, the results for variables are very stable across the lagging points. 59 LOWEST DEMOCRACY and other Table 2.5. Probability Changes in Risk for Annual MID Onset Panel A. Politically Relevant Dyads Period Cold War PostCold War Cold War Dummies Absent (Model 1) LOWEST DEM. LOWEST DEM. -1/2 sd. .0463 (.0320, .0626)+ +1/2 sd. .0424 (.0290, .0589) First Diff. % Δ of Prob. (Post – CW) -.0039 -8.4% (-.0093, .0014) Present (Model 2) .0762 (.0478, .1116) .0658 (.0405, .0989) -.0103711 (-.0197, -.0023) -13.6% Absent (Model 1) .0393 (.0245, .0566) .0284 (.0170, .0426) -.0109 (-.0184, -.0050) -27.8% Present (Model 2) .0922 (.0566, .1358) .0686 (.0378, .1073) -.0236352 (-.0385, -.0109) -25.6% Panel B. One-Year Lag for All Independent Variables Period Cold War PostCold War Cold War Dummies Absent (Model 1) -1/2 sd. +1/2 sd. .0904 (.0611, .1266) .0797 (.0524, .1127) First Diff. (Post – CW) -.0107 (-.0192, -.0026) % Δ of Prob. Present (Model 2) .1549 (.1072, .2083) .1387 (.0939, .1921) -.0162 (-.0289, -.0052) -10.4% Absent (Model 1) .0704 (.0459, .0983) .0485 (.0299, .0699) -.0220 (-.0331, -.0127) -31.2% Present (Model 2) .1282 (.0829, .1819) .0898 (.0564, .1336) -.0384 (-.0575, -.0229) -30.0% -.11.9% Panel C. Dummy for Well-Established Joint Democracy Joint Democracy Joint Democracy (Polity 9 or 10) (Polity 9 or 10) =0 1 Absent .0822 .0325 (Model 1) (.0545, .1173) (.0143, .0632) First Diff. (Post – CW) % Δ of Prob. -.0497 (-.0753, -.0264) -60.4% Present (Model 2) .1411 (.0981, .1910) .0583 (.0256, .1059) -.0828 (-.1212, -.0454) -58.7% Absent (Model 1) .0704 (.0468, .0987) .0039 (.0006, .0122) -.0665 (-.0933, -.0431) -94.5% Present .1033 .0045 (Model 2) (.0661, .1517) (.0005, .0142) +The 95% confidence intervals are in the parentheses. -.0988 (-.1457, -.0629) -95.7% Period Cold War PostCold War Cold War Dummies 60 Third, I employ a dichotomous measure for joint democracy with a relatively high threshold, polity score 9. As implied in the measurement section, the insignificant finding for the time-increasing effect of joint democracy might be due to the fledgling democracies that appeared in conjunction with the winding-down of the Cold War. Rather, according to the logic of the Kantian learning, the time-strengthening effect of joint democracy is expected for interactions among well-established mature democracies. Then, it is interesting to ask whether the pacifying effect of joint democracy between well-established democracies has strengthened in the post-Cold War era? For this reason, I define jointly democratic dyads as those in which both countries have polity score 9 or higher, coding 1 for these dyads and 0 otherwise. I also include a dummy for dyads whose joint polity scores fall between 6 and 8. So the reference category is nondemocratic dyads with joint polity scores less than 6. In the results, the coefficient for well-established joint democracy is estimated significantly larger (more than 2.5 times) in the post-Cold War era than the Cold War era in both Models 1 and 2. In terms of probability changes, well-established joint democracy decreases MID onset by 60.4% or 58.7% in the Cold War era and by 94.5% or 95.7% in the post-Cold War era. The difference is much more conspicuous in terms of relative risks: the probability of MID onset between nondemocratic countries is about 2.5 times that of well-established democracies in the Cold War era but that difference in relative risks amounts to about 20 times in the post-Cold War era. CONCLUSION This chapter reveals that realists’ skepticism of the democratic peace is not compelling. The dummies for the Cold War rivals and allies are important to take account of the Cold War effects. But these interest measures do not cancel out the peace-inducing effect of joint 61 democracy. Besides, the results of this chapter suggest that if the effect of joint democracy is time-variant, it has evolved in a positive direction for a more peaceful inter-democratic system. The results of this chapter and their theoretical implications provide some directions for the future research regarding the deepening and maturing aspects of the democratic peace. Future research may take into account the maturity of joint democracy and the democratic neighborhood in testing the democracy-peace relationship. The years of joint democracy (i.e., year counts of how many years a pair of countries (i.e., a dyad) has been able to be jointly democratic) will be a good candidate measure for that. Also, the proportion of democracy in a given year could capture the pacifying effect that democracy may generate when democracies begin to live with more democratic neighbors. The previous research has been too stuck into the simple measures of joint democracy to reveal its presumed temporally and spatially dynamic effects. Taking account of maturity and proportion of democracies in the world may be able to better explain the multifaceted features of the democratic peace. 62 CHAPTER THREE BORDERING ON PEACE OR PACIFYING THE DISPUTATIOUS BORDERS? DEMOCRACY, TERRITORY, AND ARMED CONFLICT This chapter joins a growing literature that examines the interactive linkage between two of the most important factors in international politics, democracy and territory, that affect conflict and war. Its purpose is to compare and contrast these two factors against each other and identify their possible interaction that conditions each other’s causal effects on interstate conflict. Doing so should be interesting and worth a thorough investigation on its own ground. These two factors are contradictory as one is considered most highly as a facilitator of interstate cooperation and peace and the other as a catalyst of interstate conflict. Democracy, on the one hand, is a well-established characteristic that pacifies interstate relations. The proposition that democracies very rarely fight each other is often referred to as “closest thing we have to an empirical law in international relations” (Levy 1988: 661-2). There are numerous pieces of statistical evidence for the separate peace between democracies, not only for war but also lower levels of interstate conflict such as Militarized Interstate Disputes (MIDs) (e.g., Dixon 1994; Russett and Oneal 2001). Territory, on the other hand, is the ‘rising star’ among candidate explanations for interstate conflict processes. Clearly important, it has yet to be incorporated thoroughly into the democratic peace scholarship as what may be the most frequent and dangerous issue contested by states. It is important to take into account issues and their salience to explain interstate conflict (Diehl 1992: 333). Thus it becomes interesting to assess whether the dyadic democratic 63 peace holds up even for a difficult issue like territory since territorial issues stand out as the most contentious (Senese and Vasquez 2008). The following question provides the focal point: do territorial issues mute out the pacifying effect of joint democracy (i.e., a pair of jointly democratic states) due to their exceptional disputatiousness? In seeking to answer this question, I first revisit and attempt to replicate Douglas Gibler’s study (2007), “Bordering on Peace: Democracy, Territorial Issues, and Conflict,” published in one of the well respected journals in the field of International Relations (IR), International Studies Quarterly (ISQ). In this study, Gibler finds that the democratic peace is epiphenomenal to territorial issues. In the new sequel of a well-cited book, The War Puzzle Revisited, a representative scholar of the territorial explanation of war, John A. Vasquez (2009: 370) regards Gibler’s study as “one of the few studies that successfully wipes out the statistical significance between joint democracy and peace.” However, I suspect that Gibler’s data and research design may be prone to disguising the true interaction between democracy and territory, generating the unfavorable results against the democratic peace. Gibler conceives territorial issues in an extremely static manner, employs measures for joint democracy inconsistently, and fails to utilize properly the standard methods for statistical interaction (Braumoeller 2004; Brambor, Clark, and Golder 2006) and the binary time-series cross-section data (Beck, Katz, and Tucker 1998; see also Carter and Signorino 2010). In addition, his dataset is available neither on his personal website nor on the ISQ data archive page, in violation of the journal’s replication policy. In fact, my replication generates somewhat different findings than Gibler’s. Unlike Gibler’s analyses, my replication analysis reveals that the various measures for border stability do not perform well in predicting any of joint democracy and armed conflict. By 64 contrast, democracy remains as a solid factor that inhibits interstate conflict even in the statistical model that contains Gibler’s seven border-related variables. Second, I go on further to propose an alternative interactive model that is, I believe, more appropriate than Gibler’s model to test the effect of democracy vis-à-vis territorial issues. Since none of Gibler’s three main measures for territorial threat has a robust effect, I use an 15 alternative measure based on Huth and Allee’s (2002) list of territorial claims. This measure is often used by other scholars to measure the existence of territorial issues between two nations and always produces robust significant effects on important dependent variables of interest such as military conflict and bilateral trade (e.g., Senese and Vasquez 2003, 2008; Simmons 2005). I find that, although the presence of territorial claims increases the probability of MID onset even for democratic dyads as well as nondemocratic counterparts, the pacifying effect of joint democracy holds regardless of the presence of territorial issues and even this effect is strengthened within the territorial context This chapter will unfold as follows. The second section reviews the relevant literature linking the democratic peace and territorial conflict. The third section provides the replication of Gibler’s (2007) analysis. The fourth section formulates interactive hypotheses between democracy and territory regarding their possible contingent effects in affecting armed conflict. Data description and analysis for the interactive model appears in the same section. The fifth and concluding section will discuss implications and contributions of this present study as well as suggestions for future research. 15 Huth and Allee 2002 refer to these territorial claims as territorial disputes (short of military actions). For present purposes, I label them as territorial claims in order to avoid confusion with militarized disputes, my dependent variable, listed within the MID data. This practice is consistent with other studies of territorial issues (Hensel 2001; Senese and Vasquez 2003, 2008; James, Park and Choi 2006). 65 TERRITORIAL ISSUES AND THE DEMOCRATIC PEACE The past two decades have observed the ascendancy of the democratic peace literature that focuses on the connection between domestic political systems and international peace. Well confirmed are the findings that democratic dyads produce fewer militarized disputes with each other than do other types of dyads, and that democracies tend to be victors in international conflict once involved. To explain these empirical patterns, proponents have based their theoretical reasoning on democratic norms and institutions (e.g., Maoz and Russett 1993; Dixon 1994; Reiter and Stam 2002; Bueno de Mesquita et al. 2003). Recent participants in the conversation about the democratic peace are concerned with how contentious issues like territory may be intertwined with the effects of regime type (James, Park, and Choi 2006; Gibler 2007; see also Hutchison and Gibler 2007). These studies suggest that the democratic peace decreases considerably in the context of territorial issues and further is a spurious result of stable borders, absent of territorial threat. For instance, James et al. (2006) argue that territorial issues can wash out the pacifying effect of joint democracy. Building on Hensel (2001), they use the Issue Correlates of War (ICOW) territorial claim data on the Western Hemisphere to investigate the relationship between and among democracy, conflict management and territory as an issue. The results they obtain create a puzzle for the democratic peace: in territorial claims, joint democracy shows little impact on choices among strategies to resolve those claims peacefully or violently. While James et al. (2006) demonstrate how territorial issues can reveal nuances within a largely accepted body of evidence–the democratic peace–their analysis is incomplete. They are relatively silent about an explanation and only focus on testing whether democratic dyads can settle territorial claims more peacefully than other combinations of states. Their spatial domain is 66 16 limited to the Western Hemisphere. Furthermore, their sample is a selected one only consisting of dyads already involved in territorial claims. In order to compare the effects of democracy conditional upon the presence and absence of territorial issues, all dyads must be considered, interacting the regime types of all dyads with territorial claims. In this present study, I address these shortcomings. In the fourth section, going beyond James et al. (2006), I provide both theoretical and empirical reasons why the democratic peace does or does not hold up for territorial issues. An explanation will be derived deductively via a model of issue specification from democratic peace theorizing; in addition, a more overarching list of territorial claims–Huth and Allee’s (2002) territorial claim (nonmilitary) list–allows extension of the spatial domain of James et al. (2006) over the entire globe for the 1919-1995 period. My data analysis focuses on the possible interaction of the two most prominent factors today, territory and democracy, in relation to international conflict. Regarding the possible spurious relationship between democracy and peace controlling for territorial issues, Gibler (2007) poses a seemingly more formidable challenge. According to him, the separate peace between democracies is epiphenomenal to the stability of borders between them. He argues that democracy and peace are both symptoms, not causes, of the 17 removal of territory from the agenda of issues between neighbors. He finds that stable borders increase the likelihood that a dyad will become jointly democratic. Perhaps, his most important finding is that the effect of joint democracy on military conflict becomes insignificant when 16 When feasible, James, Park and Choi’s (2006) study should be extended over the entire globe, since their observations and tests are limited to the Western Hemisphere. This reflected the availability of ICOW data at the time. 17 According to the territorial explanation of war, resolving territorial issues represents a significant step away from escalating to militarized conflict and even war (Gibler 1997; Senese and Vasquez 2008; Vasquez and Valeriano 2008; Vasquez 2009). 67 controlling for his proxy measures of stable borders. Gibler (2007: 529) provides a territorial logic behind his argument and findings: “stabilized borders” pacify interstate relations and reduce “the need for militarization and centralization in one state,” which “also tends to demilitarize and decentralize the neighboring state.” Hence, “the absence of direct territorial threat” (Gibler 2007: 515) provides peaceful external environments auspicious for the emergence of democratic dyads. Indeed, for Gibler, the democratic peace boils down to “a stable border peace” with joint democracy having no independent impact on the likelihood of interstate dispute (see Figure 1 of Gibler 2007). However, Gibler’s arguments and findings are not convincing for the following reasons. First, although various hostile external environments such as rivalries, armed conflicts, and lack of interdependence may encourage centralization and deter democratization (Tilly 1975; Rasler and Thompson 1989; Colaresi and Thompson 2003), Gibler considers only territorial issues as a valid kind of external threat. Besides, the reverse causality challenge against the democratic peace—that the causal arrow runs from peace to democracy—posed by other critics (e.g., Thompson 1996; James, Solberg, and Wolfson 1999) has failed to seriously impair the dyadic democratic pacifism). In effect, Reiter (2001) demonstrates that peace does not affect the joint level of democracy. Also, in a monadic sense, Mousseau and Shi (1999) find that a nation is about as likely to become more institutionally autocratic as it is to become more democratic in the periods before the onset of wars. Now even Thompson (see Rasler and Thompson 2004), one of the first critics of the reverse causality, admits that democracy encourages peace, not just that peace facilitates democracy. Second, Gibler measures territorial issues too statically and indirectly. His main border variables rely on unchanging or little-changing geographical characteristics such as the 68 difference in the percentages of mountainous terrain, ethnic division alongside a border, and sharing of the same colonial heritage. 18 A border that divides an ethnic group, lacks topographical distinction, and/or was drawn by a colonial master might be ad hoc, arbitrary, and thus often contested. But these indicators are at best proxies that “do not directly measure the mutual acceptation of an [existing] border” (Vasquez 2009: 371). Besides, these proxies are based on the geographical characteristics that are too passively imposed upon contiguous states, allowing little room for those states to get over and coordinate on territorial issues. Gibler also employs other time-varying proxy controls for the effect of border strength, such as the interaction terms of capability ratio, peace years, civil war onset in at least one country in a dyad, and dyad duration with land contiguity. He states that these variables affect “the legitimacy of previously drawn borders” (p. 520). But these variables are much more indirect as Gibler terms them as “controls.” Furthermore, it is unclear why these controls would pertain only or more to contiguous dyads than noncontiguous ones while in the literature they have often served as general control variables that affect interstate conflict regardless of the contiguity of dyads. A more direct measure must pertain to states’ voluntary and intentional actions. As Gibler (2007: 529) himself clearly states, “borders are international institutions” defined consensually in the field of IR as rules, procedures, and norms designed to converge expectations and constrain behavior (North 1981; Krasner 1983). To borrow from Wendt (1992), what Gibler misses is that institutions are “what states make of them.” Here comes to the fore states’ voluntary and intentional interactions such as “agreement” and “coordination.” States institutionalize borders by showing “clear agreements over jurisdiction” (Simmons 2005: 824). Neighboring countries have incentives to settle borders, though that might lack a geographical demarcation not only 18 Gibler (2007: 519) assumes the same colonial experience between neighbors as an indicator of poor border definition lacking the geographical focal points. 69 because “settling territorial disputes and mutually accepting borders produce peace” as the territorial explanation suggests (Vasquez 2009: 371; see also Gibler 1997; Senese and Vasquez 2008; Vasquez and Valeriano 2008), but also because they generate economic and political benefits (Simmons 2005; Gibler 2007). As Simmons (2005: 827) argues, when borders are clearly accepted between the neighbors, they become jurisdictional institutions that produce joint gains (Keohane 1984), especially beneficial for “economies on both sides of the border” (p. 823). Conversely, less institutionalized borders are those over which one state often challenges the other state’s jurisdictional authority over a piece of territory in question. When disagreements over borders arise and are unresolved, these may drastically increase external territorial threats that can mar “the prospects of economic development and diversification” (Gibler 2007: 516) and the legitimacy and development of domestic institutions “within a clear physical domain” (Simmons 2005: 827). Therefore, a direct measure should be something that indicates whether an actual territorial dispute occurs between two states. Third, a related problem is that while there are indeed many proxy measures employed to approximate the border stability and strength, only one main variable, ethnic border, and two controls, peace years and civil war onset, appear to have a robust, significant impact on the likelihood of both joint democracy and MID onset in Gibler’s analysis. Other measures are often insignificant or significant in the opposite direction. In addition, Gibler’s statistical analyses are contaminated with wrong numbers and wrong interpretations. To take one example, his maximum number of observations for the simplest nondirected dyadic analysis for the 19461999 period in Table 1 is 537,653. Gibler terms the number of observations as “Number of Contiguous Dyads.” While this term should be a typo, the numbers cited above even go over the possible maximum number (518,368) of nondirected dyad observations generated from the 70 19 EUGene program (Bennett and Stam 2000a). Considering that the variables, such as capability ratio, GDP, and joint democracy, employed in the table have missing values, the actual numbers of observations should be much smaller. Also, Gibler’s simplest models in Table 2 (1946-1989) and Table 3 (1946-1999) have the same observation number despite the different time frames. Gibler also incorrectly interprets the result of capability ratio. He argues that a border of equal powers is unstable and it hampers democratization. In Table 2, capability ratio has a significant positive impact on the likelihood of joint democracy. Gibler interprets this as supportive of his argument, but his capability ratio is a measure for power parity, not power preponderance, given that it is measured as the ratio of weaker power/stronger one, not vice versa. Therefore, the estimated positive coefficient sign suggests that equal powers are more likely to be jointly democratic than unequal dyads, which is definitely inconsistent with Gibler’s argument that borders of unequal powers tend to be stable. Fourth, Gibler measures joint democracy inconsistently. To show joint democracy is endogenous to the border-related variables with no independent impact on MIDs, he regresses the former on the latter in Table 2 and goes on regressing MID onset on both joint democracy and the border variables in Table 3. When employed as the dependent variable, joint democracy is measured as a dichotomous variable, while it is measured as a continuous-index variable when employed as an independent variable. If his null-finding for the pacifying effect of joint democracy were from a consistent measure, either dichotomous or continuous, his claim and analyses would be much more convincing. 19 Based on Gibler’s (2007) statement on the data construction, the possible maximum number is obtained with the following options from the EUGene program: nondirected dyad years for the 1946-1999 period, excluding any ongoing years of MIDs and any joiner dyads. Including the ongoing dispute years if and only if there occurred a new conflict increases the observation number to 518466, which is still less than Gibler’s number. 71 Taken together, these theoretical and empirical flaws, found in Gibler’s (2007) study that proposes one of the most direct challenges to the well-established democratic peace scholarship, give us reasons to question its main arguments and replicate its findings. My replication analysis will be followed and reported in the third section. Is it true that territory is the actual driving force behind the democratic peace or at least an interacting force that abates it? If that is the case, one important implication is that joint democracy may have little independent impact on settling territorial issues peacefully. Studies of the democratic peace generally have shown little consideration about this possibility. Yet, some studies do offer some favorable evidence to the democratic peace even in a territorial context (e.g., Mitchell and Prins 1999; Huth and Allee 2002; Allee and Huth 2006). Mitchell and Prins (1999) discover that territorial MIDs between well-established democracies (with a Polity score of 10) are very rare and even then tend to be minor. Given that territorial wars have been fairly stable in occurrence for a long time (Vasquez 2001: 153; Hensel 1999: 128-129), Mitchell and Prins’ findings may give credence to the democratic peace. However, how many fully democratic countries have shared borders? It is still unclear whether the fewer armed conflicts on territory between solid democracies reflect their unwillingness to use force against each other or the possibility that countries need to remove territorial issues out of agenda in order to be fully democratic and peaceful. In this vein, Huth and Allee (2002) reveal that in territorial claims democratic dyads are less likely than nondemocratic ones to escalate their territorial claims into territorial MIDs while frequently relying on nonviolent settlement efforts such as talks and negotiations. Similarly, Allee and Huth (2006) find that democratic dyads, compared to nondemocratic ones, are about three times more likely to use legal procedures in settling territorial claims. 72 Although Huth and Allee reveal important conflict patterns for democratic countries in territorial disputes, yet to be established are possible interactive effects of the two most important factors, democracy and territory, on armed conflict contingent upon each other. Like James et al. (2006), Huth and Allee (2002) also select on the dyads already involved in the claims over territory. As Huth and Allee find, the democratic peace may exist despite the salient presence of contentious territorial issues. However, it is unknown whether and how much the ability of democratic dyads to coexist peacefully is weakened by the disputatiousness of territorial issues. Conversely, it is unknown whether and how much the conflict-generating effect of territory is reduced by the democratic peace. I will address this issue in the fifth section based on appropriate interactive statistical models. BORDERING ON PEACE? REPLICATION ANALYSIS FOR GIBLER’S 2007 ISQ STUDY I start my investigation of a contrasting linkage between democratic peace and territorial conflict by replicating Gibler’s (2007) statistical analyses. After describing data construction process and measurement, the results for the replication analyses are reported and discussed. Research Design First of all, unfortunately, I have been unable to obtain Gibler’s data. His data is not available either on his personal website (http://bama.ua.edu/~dmgibler/) or the data archive webpage of ISQ (http://www.isanet.org/data_archive.html), against the journal’s replication policy. I first inquired the data availability of him in June 2009. Douglas Gibler excused that he had difficulty finding the exact dataset used in his study because it was made when he was changing his affiliation. Since then, he had repeatedly promised to make the data available or at 73 least to help me construct his data by myself until he finally confirmed he lost his dataset and refused to receive any queries about his data in May 2010. Given this situation, I decided to do what I believe as the best strategy for the data construction: recreating Gibler’s data based on what he says about his data in his article as closely as possible. When some descriptions are unclear and ambiguous, I make my best reasonable guesses with explanations. Gibler conceives a state’s border as either stable or unstable based on two kinds of factors. In other words, his border variables are categorized into two kinds. According to him, one includes time invariant geographical border salient variables (i.e., similar terrain, ethnic border, similar colonial heritages) that would pertain to the risk of territory to capture or occupation. The other refers to border strength variables (i.e., parity, peace years, dyad duration, and civil war onset) that would be related to the ability of states to defend their territorial integrity and challenge others. In addition, Gibler specifies land contiguity as a necessary condition for the border instability. For him, international borders must be contiguous by land to be fragile and unstable enough to pose territorial threat, which, in turn, deters domestic political development and facilitates interstate armed conflict. Geographic Border Salient Variables First, Gibler (2007) states that similar terrain provides few geographic salients that would help define borders clearly. Similar terrain is simply operationalized as “the [logged] ratio of percent mountainous terrain for each dyad, using the lowest percentage mountainous state as the numerator” (i.e., ln(smaller) – ln(larger)) to measure the lack of geographical focal point that clearly divides a border between two states (p. 520). For this variable, I use Fearon and Laitin’s data (2003), following Gibler (2007). Zero values are undefined for the log transformation and 74 denominator. Following Fearon and Laitin, I add 1 to the percentage monadic value of each state before log-transformation and subtraction. Second, according to Gibler, borders that divide ethnic groups are those least likely to be drawn upon coordinating geographic focal points. Yet, operationalizing ethnic border is not straightforward. Gibler (p. 520) states, “I identify internationally divided ethnic groups using the Minorities at Risk dataset, coding a dummy variable for dyads with minority groups that believe an imagined homeland includes both states in that particular dyad.” But this is a very ambiguous coding rule statement. As suggested in the citation directly above, the units of the Minority at Risk dataset (MAR, 2003) are ethnic groups. The data set contains the information on whether an ethnic group has “an imagined homeland” and whether the imagined homeland lies inside or exceeds the boundaries of the state in which the group currently resides in. But, in the case that an imagined homeland lies over more than one state’s boundaries, the dataset does not identify 20 the other states’ name. For this reason, I am unsure of how Gibler was able to identify ethnic borders that divide ethnic groups based on the MAR dataset. If there should be a way within the MAR dataset, it is possible to observe whether the same named ethnic groups reside in more than one country. For instance, the group, named “KURDS,” appears in multiple neighboring countries like USSR, Iran, Turkey, Iraq, and Syria. I code 1 for a dyad that has at least one same ethnic minority group in the MAR dataset, and otherwise 0. This identification method is crude and problematic because some groups appear in non-neighboring countries. For instance, although the “CHINESE” group also appears in multiple states like Panama, Thailand, Vietnam, Malaysia, and Indonesia, but there seems little connection between the Chinese in South East Asia and those in Panama. Another problematic case is the group named “INDIGENOUS 20 See the MAR Dataset Users Manual 0300703, page 20. 75 PEOPLES” that appears in many Latin American states and even Canada. This problem of likely unrelated, distant ethnic groups with the same name may not pose a serious problem for the subsequent analyses that follows, since this variable is operationalized as an interaction variable multiplied by land contiguity. The point is that if one wants to operationalize ethnic borders properly especially based on the MAR project, it will be likely a complicated and timeconsuming, though maybe not impossible, process that requires additional data sources, great caution and detailed explication. However, Gibler (2007: 520) describes its operationalization in just one sentence. Third, Gibler (2007) suggests that contiguous dyads that share the colonial experience from the same colonial masters tend to have poorly defined borders. These dyads are coded as 1 and 0 otherwise. For this variable, Gibler simply states that he uses Fearon and Laitin’s (2003) replication data. Fearon and Laitin’s dataset contains two relevant dummy variables that indicate whether or not a country is a former British colony and a French colony, respectively. I code 1 whenever both countries in a dyad were colonized by either the British power or the French power, and 0 otherwise. In addition to the three geographic variables, Gibler controls for the effects of wealth that might be strongly correlated with former colonial status and democracy. Wealth is operationalized as “the natural logarithm of the smaller per capita gross domestic product (GDP) in the dyad” (p. 520), also using the Fearon and Laitin data. Border Strength Control Variables Noting that the three geographic border variables are almost time-invariant while “borders are often flexible,” Gibler employs four time-variant control variables that he thinks 76 affect “the legitimacy of previously drawn borders” (p. 520). First, he expects that borders between unbalanced dyads are less likely disputed than those between dyads of near parity. Power parity is operationalized as “the capability ratio of the weaker state to the stronger state, using the… Composite Index of National Capabilities from the Correlates of War Project” (Gibler 2007: 520). Yet, it is unclear whether the author uses the conventional log transformation of the ratio (i.e., ln(the weaker CINC score/the stronger CINC score) = ln(the weaker)−ln(the stronger)) found in the literature (e.g., Russett and Oneal 2001). I do not use the natural log transformation because the raw ratio measure for parity produces the results more similar to Gibler’s analyses. Second, Gibler also uses two time-related variables, peace years and dyad duration that may be indicators of border legitimacy and stability. I use Beck, Katz, and Tucker’s (1998) BTSCS (binary times-series cross-section) program to generate the time since the last MID for peace years. Gibler does not include the then standard cubic splines to account more fully for the 21 time dependence without any justification. In addition, he does not use robust standard errors to account for possible dyadic clustering. This is inconsistent with his explicit endorsement of Beck et al.’s method for BTSCS data. However, because it is necessary to follow the original author’s approach as a starting point for any replication analysis, I do not consider the cubic splines or cubic polynomials for the replication. As for dyad duration, like Gibler (2007: 520) I use “a count variable for the number of years since the last system entry date in the dyad”. Third, Gibler includes civil war onset as a border strength control variable, because civil conflicts may harm the legitimacy of existing international borders in relevant international regions. Like Gibler, I use Fearon and Laitin’s dependent variable to code 1 for a dyad within 21 For an alternative method to the cubic splines, see Carter and Signorino’s (2010) proposal for cubic polynomial approximation. 77 which if at least one of the states was experiencing a civil war onset in a given dyad year, and 0 otherwise. Dependent Variables Gibler has two different dependent variables for his main two sets of analyses. First, in predicting observation of joint democracy, he uses Polity IV data (Marshall and Jaggers 2004). Following Gibler (2007: 521), I code 1 for a dyad in which “if both states have combined Polity IV scores (democracy-autocracy) equal to or greater than 6.” To facilitate the time series analysis, I use the Polity2 scores. Second, in predicting armed conflict onset, Gibler uses the MID definition from the COW project. Following him (p. 521), I include any MID onset as the dependent variable of armed conflict, and I exclude any joiner dyads. I also exclude any ongoing dispute years regardless of new occurrence of a MID. 22 Sample Domain Gibler uses all dyads for the 1946-1999 period including noncontiguous dyads as well as 23 contiguous ones with a contiguity control variable. The inclusion of noncontiguous dyads adds some ambiguities to the replication analysis. According to Gibler, land contiguity is a necessary condition for border instability and certain geographic conditions add to it. This means that the effects of the three main border variables and other border controls are contingent upon land 22 Inclusion of ongoing MID years if and only if a new MID occurs does not change the statistical results and conclusions of this present study. 23 Although Gibler includes contiguity in his statistical models, he never offers specific description about this variable. Since Gibler explicitly argues for land contiguity, I code 1 for a dyad if two states are contiguous by land and 0 otherwise. I generate this variable through the Eugene data generating program. 78 contiguity. But any explicit interaction terms do not appear in Gibler’s statistical tables. The question is how Gibler treats noncontiguous dyads with respect to the values of the seven borderrelated variables? Does he code 0 for these variables if a dyad is noncontiguous? In general, he is not crystal clear about this point except for the only one case that he explicitly states the term, “contiguous dyads,” for the same colonial master variable in the data description. It is understandable that authors can be ambiguous in their data description due to the often strictlyimposed page limit. But this ambiguity is much lessened when the authors make their data available or at least are cooperative for other researchers to construct replication data. Nothing is available in this present case. In passing, Gibler states, “the border variables serve essentially as interaction terms” in his models (p. 526) when he accounts for a discrepant finding for the effect of contiguity on joint democracy. To replicate Gibler’s results, I take two approaches. First, I code 0 for the values of the seven border variables when dyads are noncontiguous. Although producing the most similar results to Gibler in my data analyses, doing so is equivalent to omitting all the constituent terms in the statistical interaction model (Braumoeller 2004; Brambor, Clark, and Golder 2006). This approach can be appropriate if and only if the seven border variables do not affect interstate conflict among noncontiguous dyads. This is a strong theoretical assumption that needs empirical testing, and the empirical testing is easily conductible. But some of the border strength controls like peace years and capability ratio have been found as important factors that affect interstate conflict in general. Therefore, my second approach addresses this issue. I keep the original values of the border variables regardless of contiguity to use as constituent terms and multiply them with contiguity to use as interaction terms. It must be noted that the only control variable, GDP per capita measure (logged), will not be employed as an interactive term multiplied by land contiguity in my replication analyses. Nothing in Gibler’s 79 argument suggests an interactive effect for wealth. Besides, when the interactive term is employed, GDP per capita in my replication produces the opposite estimates with those of Gibler. I use Bennett and Stam’s (2000a) EUGene program to generate the basic template of my dataset and the values for all variables other than GDP per capita, civil war onset, mountainous terrain, colonial history, and ethnic border. Results and Analysis Gibler has two main analyses, with one predicting the likelihood of joint democracy by his border variables and the other predicting the likelihood of MID onset by a continuous joint democracy measure as well as the border variables. Each has four statistical models. I replicate the most comprehensive workhorse model in Gibler’s Tables 2 and 3, respectively. Predicting Joint Democracy Table 3.1 contains Gibler’s reported results, my replication results, and the results from the interaction model. Fortunately, I am able to produce relatively similar results to those of Gibler in terms of signs and statistical significance with a few exceptions.24 First of all, similar to Gibler’s result, the control for wealth (i.e., GDP per capita) is positive and statistically significant in my replication. In fact, this control variable produces the most similar results in most of my replication analyses that follow. As for the border strength controls, the estimated 24 The differences between Gibler’s estimates and mine might come from the ambiguity associated with the measurement method of ethnic border. Therefore, I also replicate Model 1 of Table 2 in Gibler (2007) that only contains the border strength controls without the ethnic border measure. But Gibler’s results are not replicable exactly in any case. 80 coefficients of peace years, civil war onset, and dyad duration are statistically significant and in the expected direction although those estimates are not the exactly same as Gibler’s estimates. Table 3.1. Replication of Model 4 of Table 2 in Gibler (2007): The Effects of Borders on the Likelihood of Joint Democracy, 1946-1999 Gibler’s result Replication Interaction Low GDP (logged) 0.899*** (0.023)+ 1.267*** (0.007) 1.110*** (0.007) Border Strength Controls Parity (weaker/stronger) -0.720*** (0.021) × Land contiguity 0.346^^^ (0.023) 0.202^ (0.120) 0.916^^^ (0.120) Peace years 0.024*** (0.000) × Land contiguity 0.017*** (0.000) 0.021*** (0.002) -0.001 (0.002) Civil war onset -0.664*** (0.045) × Land contiguity -1.110*** (0.305) -1.041*** (0.304) -0.432 (0.307) Dyad duration -0.001^^^ (0.000) × Land contiguity 0.008*** (0.001) 0.002*** (0.001) 0.003*** (0.001) Land contiguity -0.133* (0.077) -0.577*** (0.092) 0.122 (0.091) Border Salient Variables Same colonial master 0.534^^^ (0.017) × Land contiguity -0.222* (0.117) -0.447*** (0.111) -1.044*** (0.110) Ethnic border -0.034 (0.037) × Land contiguity -0.270*** (0.078) -0.578*** (0.091) -0.540*** (0.097) Terrain similarity -0.077*** (0.005) × Land contiguity -0.293*** (0.029) -0.021 (0.035) 0.030 (0.035) Constant -3.388*** (0.012) -2.575*** (0.007) -3.165*** (0.015) Number of dyads 364,769 379,821 379,821 LR chi-square 23,342.09*** 45,806.76*** 54,541.09*** Pseudo R-square 0.109 0.1581 0.1882 AIC NA 0.642 0.620 NA -45,691.134 -54,335.532 BIC' *p<0.1, **p<0.05, ***p<0.01 (two-tailed). ^p<0.1,^^p<0.05, ^^^p<0.01(two-tailed), wrong signs. +Standard errors are in parentheses. NA: information is not available. Consistent with Gibler’s argument, these results suggest that neighboring dyads of long peace duration, no civil war experience at least in one country, and lengthy coexistence are more likely to be jointly democratic. However, parity has an unexpected positive, significant coefficient in both of Gibler’s model and the replication model. Though Gibler (2007: 524) gives an incorrect 81 interpretation for it, this result is inconsistent with Gibler’s argument, suggesting that neighboring dyads of equal capabilities that are presumed to be conflict-prone tend to be jointly democratic. The border salient variables have consistent results except for terrain similarity. The same colonial master variable has a higher statistical significance in my replication than Gibler’s analysis. Having the same colonial master appears to decrease the likelihood of joint democracy. Also, the ethnic border variable appears to have a larger coefficient in my replication than Gibler’s analysis. This result suggests that having an ethnic group divided by an international border inhibits joint democracy. Unlike the other border stability variables, the result for terrain similarity is inconsistent. The significant negative effect of terrain similarity on joint democracy found in Gibler’s model becomes much weaker in my replication. The estimated coefficient is reduced about 15 times to be statistically insignificant. This result indicates that two neighboring countries that lack topographic focal points do not pose territorial threat enough to inhibit joint democracy. In sum, many of Gibler’s border variables appear to be associated with the likelihood of joint democracy. These results seem supportive of Gibler’s argument that stable borders are something close to a necessary condition for the emergence of joint democracy between neighbors. However, Gibler’s analysis is not limited to contiguous dyads by including noncontiguous dyads. For this reason, he seems to employ the interactive measures for border stability by setting zero values for all these variables when dyads are not contiguous while he controls the effect of wealth (per capita GDP) as a sole control for all dyads. This method is equivalent to predicting the likelihood of joint democracy between contiguous dyads with the border variables while predicting the likelihood between noncontiguous countries only with the 82 per capita GDP capital, while assuming that power parity, spells of peace and coexistence, civil war, colonial history, internationally divided ethnic groups, and terrain similarity do not affect democratization even for neighboring countries that are not directly contiguous by land. Even if this were the case, it must be subject to empirical testing. The third column of Table 3.1 contains the results for the multiplicative interactive model where the nonzero values of the border variables are preserved for noncontiguous dyads and thus their lower order (i.e., constituent) terms are included separately from the multiplied interaction terms in the equation. A Wald test of the assumption that all of these lower order terms are simultaneously equal to zero in this interaction model has a test statistic of 8390.87 with a pvalue far below 0.001. Further, the differences in the Akaike (AIC) and Bayesian (BIC) information criteria strongly support the interaction model. The BIC penalizes the additional parameters in the successive specification more than the AIC. Given the conventional guideline for the strength of evidence favoring the one model against the other suggests an absolute difference of over 10, the absolute difference of 8,644.398 in BIC' gives an extremely strong support for the interaction model against the replication model. In the interaction model, more of the border variables are estimated to have inconsistent effects with Gibler’s thesis. For instance, parity has the correct sign for noncontiguous dyads but the wrong sign for contiguous dyads, because the estimated coefficient for its constituent term is negative and smaller in the absolute value than the positive coefficient of the multiplicative term. Peace spell does not appear to matter more for contiguous dyads than noncontiguous dyads, because the coefficient for its interaction term is negative though statistically insignificant. Therefore, we can conclude that peace spell equally effects the observation of joint democracy regardless of land contiguity. Also, the presumed negative effect of civil war onset on joint 83 democracy is not found to be different between contiguous and noncontiguous dyads. Though consistently negative, the coefficient for the related interaction term is statistically insignificant. In addition, terrain similarity appears to have the expected negative, significant effect not for contiguous dyads but for noncontiguous dyads. Its estimated coefficient (-0.077 + 0.030 = 25 0.048) for contiguous dyads in Model 2 is not statistically significant as in Model 2. In sum, these inconsistent results suggest that many of Gibler’s border variables are, in fact, associated with joint democracy regardless of land contiguity. Therefore it is unclear how these variables are related to joint democracy in terms of “border stability.” Of the seven border-related variables, only three variables—dyad duration, same colonial history and ethnic border—find empirical support for Gibler’s argument. For instance, while dyad duration is negatively related to joint democracy for noncontiguous dyads, it appears to have a significant, positive impact for contiguous dyads (-0.001 + 0.003 = 0.002) at the 0.01 level. Also consistent with Gibler’s expectation, same colonial master has a significant, negative impact (0.534 - 1.044 = -0.510) on joint democracy for contiguous dyads at the 0.001 level, while its effect is significant and positive for noncontiguous dyads, as Gibler hypothesizes. Regarding ethnic border, its estimated negative coefficient (-0.034 - 0.540 = -0.574) for contiguous dyads is statistically significant at the 0.001 level while its impact is insignificant for noncontiguous dyads. Do both democratization and peace in a region require the stabilization of borders in that region to the extent that what is known as democratic peace is in fact a disguised phenomenon of stable border peace as Gibler (2007: 516) argues? Although they are not likely to be necessary conditions, settling territorial issues and stabilizing borders might contribute to democratization 25 I estimate the standard error for this coefficient for the hypothesis testing using the lincom post-estimation command in Stata 9 (e.g., lincom terrainsim + landterrainsim). 84 and peace to a certain degree. However, Gibler is not quite successful in identifying that contribution given that my interactive analysis reveals that the majority of Gibler’s variables are neither related to joint democracy nor require land contiguity to have a stronger expected association with joint democracy. Even if the three variables with supportive findings were enough to prove the stable border-democracy link, the association per se does not prove the stable border democracy causality. That link must be assessed under further scrutiny to prove that the democratic peace is spurious to the presumed stable border peace. If stable borders cause joint democracy and peace and if joint democracy has no independent impact on interstate armed conflict, then the significant pacifying impact of joint democracy will be washed out by the border variables present in the same statistical equation. Predicting Armed Conflict In predicting armed conflict, Gibler uses the 21-point Polity index based on the weak-link assumption (Dixon 1994; Russett and Oneal 2001) instead of the dichotomous joint democracy measure that he uses as a dependent variable. He justifies this inconsistency by arguing that the correlations between joint democracy and his border variables would introduce multicollinearity in the model that includes both kinds of measures as independent variables. But, according to Gibler’s logic, those correlations are the reflection of the fact that, if it is true, the pacifying effect of joint democracy is spurious to border stability. In other words, if the border variables predict joint democracy that has no independent effect on conflict, the correlations exerted by the border variables joint democracy link will force the latter’s effect to be gone in favor of the former’s. Besides, in my replication data, joint democracy is not even moderately correlated with 85 26 any of the border variables. Another of Gibler’s justification is that the continuous index is superior to the dichotomous measure (Gibler 2007: 527). While this claim itself is arguable, 27 Gibler should have used the continuous index in predicting joint democracy as well if he really believes so. However, I use the weak-link specification because my primary purpose in this section is to replicate Gibler’s analyses. The first column of Table 3.2 contains Gibler’s (2007: 528) original results reported in the fourth column of his Table 3. In Gibler’s original report, of the three main border salient variables, only ethnic border is statistically significant in predicting MID onset. However, in my replication, none of these variables has a statistically significant impact. Regarding the border strength controls, my replication produces similar results to Gibler except for dyad duration. Peace years, parity, and civil war onset continue to appear as important factors that affect armed conflict among contiguous dyads. Dyad duration becomes significant in my replication but the sign continues to be in the wrong direction. Are these control variables less or not important factors in affecting MIDs among noncontiguous dyads as Gibler suggests? These possible interactive effects are tested and reported in the third column of Table 3.2 that includes the relevant lower order terms. A Wald test supports for the inclusion of these constituent terms. Additionally, AIC and BIC information 26 The highest correlation that is really weak (i.e., 0.0545) appears between joint democracy and peace years interacted with land contiguity. 27 The dichotomous measure does cause a loss of information. However, its measurement loss is minimal and it would better reflect the reality in that the vast majority of states are either highly democratic or autocratic according to the polity scale. In defense of a dichotomous measure of democracy, Alvarez et al. (1996: 21-22) observe that “while democracy can be more or less advanced, one cannot be half democratic: there is a natural zero point.” 86 criteria suggests that the interaction model performs better than the replication model in Table 3.2. Table 3.2. Replication of Model 4 of Table 3 in Gibler (2007): The Effects of Borders and Joint Democracy on Conflict, 1946-1999 Gibler’s result Replication Interaction 1 Interaction 2 Lowest democracy × Land contiguity Low GDP (logged) 0.004 (0.003)+ -0.032*** (0.006) -0.033*** (0.006) 0.252*** (0.033) 0.331*** (0.038) 0.245*** (0.041) -0.051*** (0.009) 0.035*** (0.012) 0.235*** (0.041) -0.473^^^ (0.178) 0.769*** (0.234) -0.071*** (0.003) -0.015** (0.006) 0.610*** (0.193) 0.251 (0.252) 0.028^^^ (0.001) -0.023*** (0.001) 3.492*** (0.149) -0.492^^^ (0.178) 0.793*** (0.234) -0.071*** (0.003) -0.015** (0.006) 0.619*** (0.193) 0.236 (0.251) 0.029^^^ (0.001) -0.024*** (0.001) 3.712*** (0.168) Border Strength Controls Parity (weaker/stronger) × Land contiguity 0.511*** (0.108) 0.324** (0.153) -0.052*** (0.003) -0.087*** (0.005) 0.829*** (0.161) 0.884*** (0.163) Peace years × Land contiguity Civil war onset × Land contiguity Dyad duration × Land contiguity Land contiguity 0.001 0.005^^^ (0.001) (0.001) 3.360*** 4.530*** (0.099) (0.118) Continued on the next page. 87 Table 3.2. (cont’d) Gibler’s result Replication Border Salient Variables Same colonial master × Land contiguity 0.110 (0.107) -0.007 (0.105) 0.326*** (0.088) -0.130 (0.112) -0.031 (0.039) -5.863*** (0.067) 364,779 3,696.08*** 0.234 NA NA -0.025 (0.048) -6.815*** (0.057) 379,821 3,640.34*** 0.235 0.031 -3,511.861 Ethnic border × Land contiguity Terrain similarity × Land contiguity Constant Number of dyads LR chi-square Pseudo R-square AIC Interaction 1 Interaction 2 -0.248 (0.162) 0.219 (0.193) -0.127 (0.288) 0.016 (0.309) 0.230*** (0.043) -0.270^^^ (0.064) -5.782*** (0.110) 379,821 4,553.61*** 0.293 0.029 -4,335.205 -0.240 (0.162) 0.219 (0.193) -0.128 (0.288) 0.025 (0.309) 0.226*** (0.043) -0.260^^^ (0.065) -5.903*** (0.120) 379,821 4,562.00*** 0.294 0.029 -4,330.743 BIC' *p<0.1, **p<0.05, ***p<0.01 (two-tailed). ^p<0.1,^^p<0.05, ^^^p<0.01(two-tailed), wrong signs. +Standard errors are in parentheses. NA: information is not available. In the third column (Interaction Model 1), the results for parity and peace years are supportive of Gibler. Parity has a significant but opposite effect on MIDs for noncontiguous dyads. As the power gap between two states in a noncontiguous dyad get closer, an armed conflict becomes less likely. By contrast, parity facilitates armed conflict among contiguous dyads (0.296 = -0.473 + 0.769, significant at the 0.1 level), suggesting that borders would be subject to capture and recapture among equal powers. The conflict-inhibiting effect of peace years appears stronger for contiguous dyads than noncontiguous ones, consistent with Gibler’s expectation. However, the results for the other two controls are inconsistent with Gibler’s expectation. Civil war onset does not have a stronger positive effect on MIDs for contiguous dyads than noncontiguous ones. The unexpected positive effect of dyad-duration becomes 88 substantially weaker for contiguous dyads, but its coefficient sign still remains positive and significant for these dyads at the 0.001 level (0.005 = 0.028 + -0.023). In the interaction model, the border-salient variables continue to appear as insignificant factors in affecting the likelihood of MID onset. One interesting thing is found with respect to terrain similarity. The presumed conflict-generating effect of terrain similarity appears to be the case only for noncontiguous dyads. But, this effect is gone when it comes to contiguous dyads, inconsistent with Gibler’s argument. The estimated coefficient of terrain similarity is even negative, though statistically insignificant (-0.040 = 0.230 + -0.270). It is really unclear how terrain similarity can be a meaningful factor between distant states, but not bordering neighbors. This result may reflect the effect of geographic distance between two states in a dyad. It may be that the more similar terrain, the closer distance. Indeed, many statistical studies have found that geographical distance in addition to contiguity is a useful control in predicting armed conflict (e.g., Russett and Oneal 2001; Gartzke 2007; Choi 2010). To summarize, of the seven borderrelated variables, only two, parity and peace years interacted with land contiguity, produce the results supportive of Gibler’s argument for land contiguity as a necessary condition for border instability. However, these two interaction variables are not Gibler’s main border salient variables. Table 3.2 also contains the results for my key variable, lowest democracy (i.e., the continuous index measure for joint democracy based on the weak-link assumption). Unlike Gibler’s original results (see the first column of Table 3.2), my replication model (the second column) reveals that lowest democracy is an important inhibitor of interstate conflict despite the presence of the seven border-related variables. I probe the possibility that Gibler might use the multiplicative interactive term of lowest democracy × land contiguity instead of lowest 89 democracy. But this does not change my conclusion (unreported here). When the interactive lowest democracy alone is employed in the replication model of Table 3.2, its estimated coefficient becomes lower but still negative and significant at the 0.05 level. The interaction model in the third column also shows that the significant negative effect of lowest democracy holds even with the extensive controls of the border variables and their interactive terms. One may wonder whether joint democracy reduces MIDs for contiguous dyads controlling for the border effects. The second interaction model, reported in the fourth column (i.e., Interaction 2), conducts this test by employing a multiplicative interaction term of lowest democracy × land contiguity. Lowest democracy appears to have a larger negative effect for noncontiguous dyads than contiguous dyads, as one might expect. Yet, lowest democracy still appears to have a significant and negative impact on MID onset even for contiguous dyads (0.051 + 0.035 = -0.017) at the 0.05 level. I also test the pacifying effect of joint democracy in the model that includes both the dichotomous and continuous democracy measures (unreported in the table). While the results for the border variables do not change substantially, the dichotomous measure outperforms the continuous one, with the presence of the former forcing the latter’s coefficient to reduce substantially to the extent that the latter becomes statistically insignificant. In conclusion, I find no evidence that stable border peace is the actual driving force for the democratic peace. Even doubtful is the extent that stable borders measured in terms of static geographical characteristics contribute to the understanding of international peace and conflict. Robustness Checks 90 In general, the border variables do not perform well to Gibler’s expectation and fail together to outperform joint democracy, regardless of its measurement specification. Gibler (2007: 527) would raise a concern for the possibility that joint democracy “was serving as an instrument for the effects of these border variables.” I conduct two sets of testing. Table 3.3. Multicollinearity Checks Variable R-Square VIF Cond. Index Eigenvalue Dummy democracy 0.679 3.11 1 1.000 5.063 separate estimate 0.149+ 1.08 Lowest democracy 0.680 3.12 2 1.197 3.533 separate estimate 0.152+ 1.18 Low GDP 0.202 1.25 3 1.704 1.744 Parity 0.038 1.04 4 1.956 1.324 × Land contiguity 0.639 2.77 5 2.122 1.125 Peace Years 0.570 2.33 6 2.232 1.017 × Land contiguity 0.627 2.68 7 2.496 0.813 Civil war onset 0.039 1.04 8 2.585 0.758 × Land contiguity 0.068 1.07 9 2.781 0.655 Dyad duration 0.570 2.33 10 3.005 0.561 × Land contiguity 0.640 2.78 11 3.183 0.500 Same colonial master 0.071 1.08 12 3.412 0.435 × Land contiguity 0.381 1.61 13 3.632 0.384 Ethnic border 0.268 1.37 14 3.971 0.321 × Land contiguity 0.413 1.70 15 4.298 0.274 Terrain similarity 0.029 1.03 16 4.833 0.217 × Land contiguity 0.612 2.58 17 6.701 0.113 Land contiguity 0.849 6.61 18 7.013 0.103 19 9.004++ 0.062 +Estimated R-Square values when dummy democracy and lowest democracy are separately considered. ++Conditional Number = 9.004. First, I examine the possible multicollinearity between joint democracy measures and border variables. According to Menard (2002), if multicollinearity is a concern in regression, each independent variable must be regressed on all other independent variables in a standard ordinary least squares (OLS) regression to obtain R² or tolerance value. If R² values approach 91 0.80 or more, “multicollinearity may pose problems” (Menard 2002: 76). In Table 3.3, only one 2 2 of the R scores regarding land contiguity (R = 0.849) approaches the threshold. Yet, it must be remembered this variable has served as a moderator variable interacted with all other border2 related variables. The R scores of the two democracy measures are among the highest. But these scores are still less than the threshold value and are due to the correlation between themselves, 2 not with the border variables. In fact, the R value for these two variables, when separately examined, is 0.149 for the dichotomous measure and 0.152 for the continuous index. Therefore, the variations in the joint democracy measures are little explained by the border-variables. Multicollinearity is also checked through the Variance Inflation Factor (VIF) (also tolerance (i.e., 1/VIF)), which shows “how the variance of an estimator is inflated by the presence of multicollinearity” (Gujarati 2004: 351). If there is no multicollinearity, the VIF or tolerance will be 1. There are no formal criteria to decide whether a VIF value is large enough to affect the predicted values. As a rule of thumb, if a VIF is higher than 10 (or a tolerance is 0.1 or less), it may indicate multicollinearity. None of the VIF scores, as evident in the second column of Table 3.3, approach the threshold value. In addition, a multicollinearity check is conducted using eigenvalues and resulting conditional index. (I will not discuss the meaning of eigenvalues here, for that would take us in to topics in matrix algebra that are far beyond the scope of this study). The conditional index is derived from the eigenvalues (i.e., condition index² = maximum eigenvalue / minimum eigenvalue). A rule of thumb is that if the conditional index exceeds 30, there is severe multicollinearity. Some assert that the conditional index is the most reliable among available diagnostics (see Belsley, Kuh, and Welsch 1980). In the third column of Table 3.3, the highest value of the conditional index is 9.004, which is far less than the threshold value. 92 Taken together, these tests show that multicollinearity does not contribute to finding insignificant effects for border variables and significant effects for joint democracy variables. This conclusion is especially strong when we consider that the absence of multicollinearity problem is found in the model that includes the seven interaction terms. Table 3.4. Separate Analyses of MID Prediction by the Border Variables and Joint Democracy, respectively. Border variables Gibler Model 1 Replication Lowest democracy -0.005** (0.002) -0.042*** (0.006) Lowest GDP (logged) 0.188*** (0.039)+ 0.087** (0.028) 0.226*** (0.036) Border Strength Controls Parity (weaker/stronger) -0.418^^ (0.176) × Land contiguity 0.702*** (0.232) 0.105 (0.101) 0.148 (0.143) Peace years -0.072*** (0.003) × Land contiguity -0.014 (0.006) Civil war onset 0.571*** (0.193) × Land contiguity 0.279 (0.251) Dyad duration 0.028^^^ (0.001) × Land contiguity -0.023*** (0.001) Land contiguity 3.478*** (0.147) 3.988***(0.057) 3.816*** (0.080) Border Salient Variables Same colonial master -0.245 (0.159) × Land contiguity 0.226 (0.189) Ethnic border -0.127 (0.287) × Land contiguity 0.031 (0.308) Terrain similarity 0.241*** (0.042) × Land contiguity -0.273^^^ (0.063) Constant -5.564*** (0.101) -6.785***(0.053) -6.829*** (0.057) Number of dyads 383,659 504,376 379,821 LR chi-square 4572.30*** 3,631.25*** 3105.52*** Pseudo R-square 0.291 0.199 0.200 *p<0.1, **p<0.05, ***p<0.01 (two-tailed). ^p<0.1,^^p<0.05, ^^^p<0.01(two-tailed), wrong signs. +Standard errors are in parentheses. Second, another check for Gibler’s concern about instrumentality of the joint democracy variable is in order, running the regression model without it (see Table 3.4). If the instrumentality 93 posed a serious problem against their effects in the previous analysis, the border variables must perform better in this model. This is also the method Gibler (2007: 527) utilizes when he addresses the inconsistent results for the dyadic duration and colonial master variables. The first column of Table 3.4 contains the results for the border variables estimated without any of the joint democracy measures. The results are qualitatively the same as those in the third column of Table 3.3 except for peace years. None of the three main border salient variables is statistically significant. Of the four border strength controls, only the interactive effect of parity is in the direction of Gibler’s expectation. The significant interactive effect of peace years, found in Table 3.3, now disappears in the model without lowest democracy. Indeed, rather than disguising the effects of the border variables, lowest democracy does help find one of the border variables’ significant interactive effect. I have noted that there is an ambiguity in measuring ethnic border. Therefore, I also conduct additional robustness tests to check whether this ambiguity has hampered my replication analysis from producing the exactly same results as Gibler’s. One of Gibler’s models (i.e., Model 1 of Table 3 in Gibler 2007: 528) only includes four independent variables such as low per capita GDP, lowest democracy, parity, and land contiguity. As suggested in the third column of Table 3.4, the failure to replicate Gibler’s result may not be associated with the ethnic border variable. I still fail to produce exactly the same results even for the simplest model without the borderrelated variables. The coefficient for lowest democracy is underestimated by as much as 8.4 times in Gibler’s analysis compared to my replication. Similarly, my estimated coefficient for the lowest GDP per capita is about 2.6 times higher than Gibler’s. The other two controls, however, produce results similar to those of Gibler. 94 Probability Changes Figures 3.1 and 3.2 display graphically the estimated probability changes of MID onset according to changes in the lowest democracy variable. These marginal effects are calculated using King, Tomz, and Wittenberg’s (2000) CLARIFY program. 0 .005 Prob. of M ID O nset .01 .015 .02 .025 Figure 3.1. Predicted Probability of MID and Net Effects of Lowest Democracy -10 -5 0 Joint Level of Democracy Replication 95%CI for Rep. 5 10 Interaction1 95%CI for Int. 1 Note: The circle-scatter line is drawn from the replication model of Table 2 while the squarescatter line is from the first interaction model in the same table. The dashed lines are the 95% confidence interval for the replication model and the dot lines for the interaction model. For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of this dissertation. Figure 3.1 compares the net effects of lowest democracy across models. To do this, I set the dyad to be contiguous while holding all other aspects to be typical (i.e., continuous variables 95 at their mean and dummy variables at their modal category). The circle-scatter line and squarescatter line, which are drawn respectively from the replication model and second interaction model of Table 3.2, show that lowest democracy decreases MID onset regardless of model specifications. In terms of substantive meaning, dyads with the maximum level of joint democracy (i.e., lowest democracy = 10) are about 1.9 times less likely to experience a MID than dyads with the minimum level of joint democracy (i.e., lowest democracy = -10). Figure 3.2. Interactive Effects of Lowest Democracy by Land Contiguity .0002 Prob. of MID Onset .0004 .0006 .0008 .001 .0012 a. Noncontiguous Dyads -10 -5 0 Democracy Mean Probability 96 5 95% CI 10 Figure 3.2 (cont’d) .02 .025 Prob. of MID Onset .03 .035 .04 .045 b. Contiguous Dyads -10 -5 0 Democracy Mean Probability 5 10 95% CI With respect to the interactive effect, Figure 3.2 compares the effects of lowest democracy contingent upon land contiguity. Figure 3.2.a describes the effect of lowest democracy for noncontiguous dyads while Figure 3.2.b does so for contiguous dyads. I calculate the predicted probabilities based on the second interaction equation that models explicitly the interaction between lowest democracy and land contiguity. The values of independent variables are set at their conditional mean or modal category in each context (i.e., the mean or modal category of X1 given that dyads are noncontiguous or contiguous). As shown in both Figures 3.2.a and 3.2.b, as lowest democracy increases, the probability of MID onset decreases regardless of land contiguity. However, such pacifying effect appears to be larger for noncontiguous dyads than contiguous ones. In substantive effects, noncontiguous dyads with the maximum lowest democracy score are about 2.8 times less likely to experience MID onset than 97 those with the minimum lowest democracy score. The equivalent probability difference is reduced for contiguous dyads. MID onset is about 1.4 times less likely for contiguous dyads with the highest democracy score than those with the smallest score. In sum, all the results show that unlike Gibler’s estimation, lowest democracy is a solid pacifying factor in terms of both regression coefficients and predicted probabilities even after controlling for territorial threat. If there is anything about the borders that may constrain the effect of democracy to a certain degree, it is probably through land contiguity itself. However, although its effect is weakened, lowest democracy still reduces MIDs significantly even for contiguous dyads. Therefore, I do not find support for Gibler’s claim that the democratic peace is epiphenomenal to the stable border peace. It should be noted that nothing can be definitively said about the discrepancies between Gibler’s analyses and mine and about the sources of such discrepancies, unless I can attain and analyze the data that Gibler actually used in his 2007 ISQ paper. But it also must be pointed out that the discrepancies and related uncertainty do stem from Gibler’s inability to follow the journal’s replication policy. PACIFYING THE DISPUTATIOUS BORDERS? INTERACTIVE ANALYSIS OF DEMOCRACY AND TERRITORY Is the democratic peace epiphenomenal to territorial issues? Do territorial issues reduce the effect of joint democracy or vice versa? In this section, I examine the possible interaction of democracy and territory in impacting upon interstate conflict. I argue and find that both democracy and territory are solid causal factors for conflict processes, while recognizing the theoretical possibility that one may condition the other’s effect to a certain degree and vice versa. 98 I formulate several interactive hypotheses drawing upon existing explanations on territorial conflict and the democratic peace. Democratic Peace Explanations and Hypotheses Respective democratic peace explanations may produce different or similar expectations about the behavior of democracies when territorial issues are at stake. First, the normative explanation focuses on norm externalization and mutual trust between two democracies. Peaceful settlements are regarded as more likely between democracies than other conflicting pairs because of their experience with mediation, negotiation and compromise at the domestic level (Dixon 1993, 1994). These experiences socialize democratic leaders to act with, for example, greater emphasis on the rule of law (Simmons 1999). When two democracies are involved in a dispute with each other, they externalize their domestic political norms because of a conviction that the other state also abides by such norms (Maoz and Russett 1993). This belief makes democracies more likely to trust and respect each other when conflict arises between them. Second, the structure-constraint model centers on difficulties and costs leaders in democratic systems face in mobilizing for use of force (Rummel 1983; Morgan and Campbell 1991; Bueno de Mesquita and Lalman 1992). The complexity of the democratic process and requirement of consensus from a wide range of constituents make democratic leaders more deliberate and accountable in their decision-making than autocrats. This makes mobilization of resources for war and use of force slower and more costly in democracies. When conflict arises between democracies, structural features facilitate peaceful solutions (e.g., Russett 1993: 38-40). Neither the structure-constraint nor the normative explanation specifically discusses the effects of issues in general or territory in particular. If people in democratic societies become 99 particularly attached to a piece of land, public opinion may propel leaders towards bellicosity to recover or protect the coveted territory. Structural and normative constraints in democratic societies therefore may be attenuated. From the standpoint of the democratic peace, while military force in support of territorial claims is likely to “generate short-term domestic support, costly or failed attempts at military coercion will also mobilize domestic opposition” (Huth and Allee 2002: 71). Democratic dyads, in the long run, should pursue more peaceful settlement attempts even for territorial issues. Taken together, these two democratic peace explanations generate the same hypothesis with respect to democratic dyads and territorial claims: Hypothesis 1: Even when territorial issues are at stake, the rate of armed conflict will be still lower between democracies than between nondemocracies. Two other variants of the institutional democratic peace explanation are able to produce a more nuanced hypothesis with respect to salient issues such as territory. Third, among variants of the institutional explanation is the informational model. It carries a somewhat more nuanced view of democracies’ conflict behavior and focuses on the “informational properties” of democracy (Fearon 1994; Schultz 1998, 1999, 2001b; Lipson 2003). Democratic institutions help to reveal reliable information in a dispute due to the accountability and transparency generated by their political processes. International crises are public contests in which leaders’ performances are evaluated by their domestic audience. It is assumed that, once a crisis is underway, domestic constituencies do not like to see their leaders back off. Audience costs encourage political leaders to be consistent in their posture: if they initiate conflict they should not back down, otherwise they can expect to suffer public wrath. 100 Leaders who bear high audience costs are less likely to bluff and more likely to prevail in international crises. Relative ease of removal of leaders and accountability in a democracy impose higher audience costs on its leaders. Thus democratic leaders can send more credible signals of resolve in a conflict than autocratic leaders. As a by-product, a dispute initiated between democracies becomes less likely to escalate. The credible information flow between democracies makes it easier for them to identify a settlement ex ante rather than incur the risks and costs related to use of force. Therefore, the disputes between two democracies (and possibly those initiated by democracies) are less inclined to militarization as I hypothesize above. The information perspective does not explicitly consider the role of issues under contention. However, an extension of the information perspective is possible with respect to the salience of a given disputed issue. The credible information flow between democracies is a function of the general public’s attention to international disputes and crises. Given the salience of territorial issues as discussed later, the pacifying effect of democracy may be even more evident in contentions over territory. Human territoriality, symbolic meanings and tangible values associated with territory may make citizens pay more attention to military conflict over territory and impose harsher punishments on unsuccessful leaders. Territorial issues between two democracies could increase the likelihood of armed conflict between them. Compared to autocrats, however, democrats are quite susceptible to public scrutiny and therefore should have even less incentive to bluff and misrepresent resolve with respect to a salient issue like territory. By contrast, for two autocracies, the inability to generate audience costs means that the very salience of territorial issues will make such dyads even more conflictual. This leads us to expect that the gap in conflict propensities between democratic and nondemocratic dyads will be wider for territorial disputes than nonterritorial ones. 101 Fourth and finally, the selectorate model, the latest variant of the institutional explanation examined here, also makes an interesting point about territorial issues – similar to the one from the information perspective, but through a different causal mechanism. According to Bueno de Mesquita, Smith, Siverson and Morrow’s (2003: 418-432) analysis of the winning coalition in a state, the smaller its size, the more likely it becomes that war aims will focus on private goods rather than public goods. The selectorate theory regards territorial issues as something closely associated with private goods in contrast to policy disputes regarded as something related to public goods. Therefore, a democracy will be even less likely to risk fighting with another democracy when the issue under contention is about territory. The leader with a large winning coalition “should not seek territory for its resources nor strongly resist demands for revision in such territories,” since resources gained by territory do not significantly increase the ability to maintain office through provision of public goods (Bueno de Mesquita et al. 2003: 413). By contrast, nondemocratic leaders with small coalitions can benefit from increases in state resources caused by territorial acquisition. They can use a larger resource pool to buy and consolidate coalition members’ loyalty, increasing their ability to hold office. Besides, losing a war itself may not be a fatal blow to nondemocratic leaders as long as they can compensate for the defeat with resources unspent for war efforts. Nondemocratic leaders’ willingness to gamble on war should be higher regarding territorial disputes and thus nondemocracies should engage in territorial expansion more often than democracies through use of force. Taken together, both the information perspective and the selectorate theory predict that the pacifying effect of democracy will get stronger in the presence of territorial issues. This anticipated effect is due to different causal mechanisms: salience of territorial issues for the former and lack of value for the latter. 102 Hypothesis 2: The difference of armed conflict rates between democratic dyads and other 28 combinations will be greater in territorial claims. The selectorate model predicts a low likelihood that territorial claims between two states with sizable winning coalitions will escalate into violence. However, this prediction depends on whether territory is a private good as per the following assessment. In contrast to the perspective of the selectorate model, for the territorial explanation of conflict and war, territory is something every human pursues for biological, materialistic and/or symbolic reasons (Senese and Vasquez 2008: 11). Territorial Explanation of War and Hypotheses Territorial explanation of conflict and war is a subset of the Issue-Based Approach (IBA) to the occurrence and escalation of interstate conflict (Mansbach and Vasquez 1981; Vasquez 1996; Hensel 2001). The IBA adds an issue-related dimension to the generic rationality assumption upon which both realism and liberalism are based (Legro and Moravcsik 1999): Decision-makers are described as rational but in the sense of maximizing satisfaction of the values that the issues represent. Outcomes of world politics will vary according to issues. According to the IBA, foreign policy decisions by leaders who seek reelection are affected by the salience of issues at stake. For instance, military actions against another country at the first sign of a dispute will be perceived as more risky when compared to other options. 28 Other studies provide the basis for a similar hypothesis. For instance, Kacowiz (1995: 265) argues that democratic institutionalization decreases nationalist and irredentist claims, and Lake (1992) argues that democracies will not use force to seek territorial expansion because territory serves the interests of a small fraction of broad democratic constituencies while distorting the economic structure of democratic society. 103 Although the threat or use of military force could generate a rally-around-flag effect over the short-term, such actions always entail the possibility of erupting into full-scale military confrontation. In addition, failed military actions often cost incumbent leaders their offices and sometimes their lives. What if, however, the issues at stake represent highly salient values? In the face of intensely competitive issues, “where each side is reluctant to give in,” such contests often become militarized, which increases the likelihood of full-scale war (Senese and Vasquez 2005: 608; see also Vasquez 1996, 2004). Therefore, the IBA suggests that the pacifying effect of joint democracy either may decrease significantly or not hold in contention over highly salient issues, especially territory as discussed below. The IBA also emphasizes that military conflict is a dynamic process. Existing studies do not contest that point, but few research designs including aggregate data look at how the nonviolent pre-stages of military conflict escalate to the level of militarized dispute in the first place. Research on the democratic peace, which considers phases of escalation, tends to look only at the dispute behavior of democratic pairs already involved in some level of militarized hostility. Generally, findings show that democracies are more likely to resolve their militarized conflicts through peaceful means such as mutual agreement and concession, negotiations, compromises, and third party arbitration or mediation before escalating to more violent levels of militarized coercion such as war (Dixon 1994; Raymond 1994; Mousseau 1998). Although these findings support and explain the absence of war between democracies, they do not necessarily account for another empirically confirmed phenomenon between democracies: the scarcity of lower-level military conflicts such as MIDs. The studies simply do not look at stages of conflict that precede the onset of militarized hostilities. Militarized conflict 104 does not appear without underlying contentions or disagreements over issues – put simply, there must be a dispute before the shooting starts. This study goes one step back to investigate conflict behavior for joint democracies in initial stages of disputes – that is, before militarization. The focus is on escalation processes, from non-military disputes over issues such as territory to militarized conflict. Among many issues, territory stands out as the most important underlying cause of militarized disputes and war. This is why many proponents of IBA focus on territory and provide what is now known as the “territorial explanation of war” (Vasquez 1993; Vasquez and Gibler 2001; Senese and Vasquez 2003, 2005, 2008; see also Huth 1996): “territorial issues are very salient to leaders and followers, so much so that, all other factors being equal, they will be willing to incur costs and take risks on territorial questions that they would not on other issues – such as regime or policy issues” (Senese and Vasquez 2003: 277). These assertions are backed up by two underlying arguments. First, the biological foundation of human territoriality predisposes humans to fight for territory (Vasquez 1993; Senese and Vasquez 2008). Division of the Earth into territorial units has been influenced considerably by human territoriality. In turn, it also makes states highly sensitive about their borders and any threat to their territory (Vasquez 1993:140-141; Senese and Vasquez 2008). The territorial explanation argues that, rather than being interwoven by similar political regimes, whether states can have long periods of peace largely depends on settled borders. Second, territory is politically salient because by nature it contains both tangible and intangible assets. Many territories have been the subject of dispute because they contained (or were believed to contain) valuable resources. Another tangible benefit of territory is its strategic 105 value. Certain territories provide access to the sea or other commercial routes (Huth 2000: 58). Some locations, like the Golan Heights and Dokdo Island of the East Sea between two democracies, Korea and Japan, contain defensible geographical features; these contribute to a state’s perceived power and security by allowing for advance warning of an impending attack and contributing to national defense. Symbolic or psychological reasons also can play a role. Territory lies at the heart of ethnic and national identity and cohesion, with the very existence and autonomy of an ethnic group or a state being rooted in its territory (Murphy 1990: 531; Huth 2000: 59). Thus territory frequently becomes contentious; both irredentism and secessionism can come into play when members of an ethnic, national, or religious group inhabit a neighboring state. According to the territorial explanation, territory is seen as the main cause of military conflict that may end in war, over and above regime type (Vasquez 1993; Senese and Vasquez 2008). In the context of territorial conflict, two democratic disputants still might be more capable than non-democratic disputants of resolving their disagreements nonviolently. This would be anticipated by the democratic peace explanation of conflict. However, the territorial explanation suggests that, at the very least, the pacifying effect of joint democracy may be reduced significantly vis-à-vis territorial issues. This is because contention over such highly salient issues leaves democratic leaders vulnerable to the prescriptions of realpolitik, thereby creating more chances to be drawn into a security dilemma (Senese and Vasquez 2005). Thus I derive my third and fourth hypotheses: Hypothesis 3: Claims over territory will increase the rate of militarized disputes between two democracies. 106 Further, the pacifying effect may be significantly reduced to the extent that: Hypothesis 4: There is no difference between non-democratic dyads and democratic dyads in the rate of armed conflict. Territorial issues should be a hard case (or least likely case) for the democratic peace thesis in view of the territorial explanation (Eckstein 1975). But some proponents argue that democracies do not initiate fighting over territorial issues because the stakes from those conflicts are not large enough to distribute over a wide range of democratic constituencies (Bueno de Mesquita et al. 2003). I suspend judgment on this. If territorial issue specification is a difficult test, favorable results should provide strong support to the democratic peace thesis, i.e., what Levy (2002) labels the Sinatra inference, “if the democratic peace theory can make it with territorial issues it can make it with any other issues.” Research Design My hypotheses focus on how the effects of joint democracy differ between dyads with and without territorial issues. I examine all dyads (rather than dyads with territorial claims) from 1919 to 1995 where both measures for democracy and territorial claims are jointly available. The use of all dyads enables us to avoid the possible problems caused by the selected sample (see Senese and Vasquez 2003). It also allows us to examine the interactive effects of democracy and territory and assess the statistical uncertainty for such contingent effects. I use EUGene software to generate the basic template of my dataset (Bennett and Stam 2000a). My dependent variable is MID, coded 1 if there was a new onset of a MID in a given dyad year (otherwise 0) according to 107 the MID dataset (version 3.0). As Bennett and Stam recommend, I exclude ongoing MID years and ‘joiner’ dyad years unless a new direct onset of a MID occurs within the dyad. I measure territorial issues based on the information of territorial claims provided by Huth and Allee (2002). Huth and Allee identify 348 distinct interstate territorial claims, worldwide from 1919 to 1995. In some cases, the same dyads have been involved in multiple territorial claims. Huth and Allee’s list does not include claims regarding maritime boundaries: “The data set is restricted to only disputes over shared land borders, shared river boundaries, claims to sovereignty over offshore islands or islands in shared rivers, or a few military base cases that 29 raise issues of sovereignty over territory.” Since my units of analysis are all dyad-years from the interstate system (N = 498,689 for 19191995), I code 1 if a dyad is involved in at least one territorial claim, otherwise 0. I label this variable as “territorial claim.” I believe this measure is better than Gibler’s proxy measures that do not perform robustly in this present study. In Gibler’s (2007: 512) words, territorial threat to states refers to challenge to territorial integrity of states by other states. In turn, territorial integrity refers to jurisdictional legitimacy and authority of the government to rule within “a clear physical domain” (Simmons 2005: 827). As Simmons points out, clearly accepted borders as international institutions means the drastic reduction of external challenge to territorial integrity of both governments to produce joint gains. Conversely, frequent disputes over territory indicate the lack of clear mutual acceptance and institutionalization on borders. In this spirit, Gibler has used territorial MIDs, instead of the proxy measures of stable borders, as the measure for external territorial threat in 29 E-mail conversation with Paul Huth, July 2, 2009. 108 his later studies on democratic values (Hutchison and Gibler 2007; Gibler 2010). Vasquez (2009: Ch.10) acknowledges this effort as an important empirical improvement that is of theoretical interest that territorial disputes dampen democracy. However, employing militarized territorial disputes as territorial threat takes two steps forward while skipping the non-militarized dispute stage of territorial claims. In fact, territorial MIDs are a function of territorial claims (Vasquez and Senese 2003). Besides, since my dependent variable is MID onset, it is impossible to use territorial MIDs as an independent variable. By contrast, Huth and Allee’s criteria for territorial disputes fit well the notion of territorial threat defined above. They coded territorial disputes as cases of disagreements between governments over the location of a border, the refusal of one government to recognize the sovereignty of another over some portion of territory within the border of that country, or the refusal of one to recognize the independence and sovereignty of another country. Then a territorial disagreement involves one state’s claim over a piece of territory not accepting an existing boundary or its occupation and refusal to relinquish control (see Huth 1996: 20-21). Therefore, Huth and Allee’s definition of territorial disagreements touches upon the essence of territorial integrity, threat, or issues. Since my hypotheses refer to democratic dyads, not the joint level of democracy, I employ joint democracy as a dichotomous measure based on the Polity2 scores. This method also provides a direct test for Gibler’ argument. If the observation of joint democracy without an independent impact on MIDs is a function of the absence of territorial issues as Gibler argues, including the territorial claim variable will wipe out the statistical significance of the joint democracy variable. Following convention (e.g., Dixon 1994), joint democracy is coded as 1 if both states in a dyad have six or more as a score, otherwise 0. While this dichotomy does cause a loss of information, measurement loss is minimal because the vast majority of states are either 109 highly democratic or autocratic; in fact, the dichotomy might even be regarded as an improvement (Alvarez et al. 1996). Based on the standard statistical model of conflict (e.g., Bremer 1992; Russett and Oneal 2001; Senese and Vasquez 2003), five control variables are introduced that may influence states’ decisions to use force for settlement. All of the controls except for peace are operationalized through EUGene software (Bennett and Stam 2000a). First, I include a control for geographic influence on militarized conflict. Neighboring countries should have more opportunities and issues over which to fight militarily. Most importantly for this present study, neighboring countries tend to have more territorial claims and their claims may be even more salient (Vasquez 1995; Gibler 2007). Thus, it is necessary to control for this factor to attain the unbiased net effect of territory as well as the interactive effect with democracy. 30 “Noncontiguity” is coded 1 if two states are not contiguous by land or 30 As suggested, territorial claims do not all have equal importance (see Hensel 2001). To test the effect of democracy vis-à-vis territorial issues, it may be important to control for salience of a territorial claim if it is true that democratic dyads tend to have minor territorial issues. However, two difficulties arise. First, Huth and Allee’s (2002) data set does not include such a measure, although it does identify types of territorial claims according to whether or not economic, strategic, or ethnic values are entailed. Second, even if such a measure were available it is not easy to control for salience in this present study. To assess the interactive effect of democracy and territory, the universe of my data must become all dyads, not just those with territorial claims. Nevertheless, I ran a robustness check, differentiating types of territorial claims since certain types of territorial disputes may have more salient effect and the pacifying effect of democracy may be contingent upon certain salient types of territorial claims. I have done this first by running a probit model similar to the one in Table 1 to identify which types of territorial claims have the most effects on MIDs. Territorial claims can be categorized as having either no identified value or having economic, strategic, ethnic values, or some combination of those values. I found that while all kinds of territorial claims (even territorial claims lacking a specific value) significantly increase the probability of a MID, territorial claims have the strongest effects on MIDs when they include ethnic value and/or economic value. I regarded dyads as territorial (coded 1 for the territorial claim variable) only when the dyads had territorial claims with either or both of ethnic values and economic values. I conducted similar interactive analyses, reported in Table 3.7. The findings are consistent with my main results. 110 separated by 150 miles of water. In order to fully account for the effect of geography, I also employ distance between two countries. Distance between two states is measured as the natural logarithm of the great circle distance in miles between the capitals of the two states (or cities for the largest countries such as the US). I expect that the more distance between two countries, the less chance they experience military conflict because of geographical constraints in exercising 31 force over long distance. Second, the effect of distance to constrain militarized options may be less for the major powers with land, sea or air capability to exercise military power from far away. Many empirical studies have found that major powers are more conflict-prone than minor powers (e.g., Bremer 1980; Russett and Oneal 2001). Following these studies, I employ a variable that identifies whether both members of a pair of states are minor powers (“minor power”). Such a case is coded 1 and otherwise 0. I expect that a dyad consisting of only minor powers is less likely to engage in militarized attempts. Third, previous research also shows that, the greater the difference in capabilities between two states, the lower the likelihood they will be involved in a MID (Kugler and Lemke 1996). Therefore, the capability ratio for each dyad is included (Capabilities Ratio hereafter). Based on the COW Composite Indicators of National Capabilities (CINC) data (Singer, Bremer, and Stuckey 1972), the measurement for relative power in a given dyad is a natural logarithm of the higher/lower ratio between states A and B. 31 Some would argue against inclusion of geographic proximity controls that could exclude most current, important territorial claims between neighbors and produce disproportionate effects on the sample from seemingly low salience cases and old colonial cases. I consider this point and find that excluding the geographic controls does not alter my conclusions on either the net or interactive effects of democracy and territory. 111 Fourth, it generally is thought that allies are less likely to fight each other because they share common security interests (and also political and economic interests). Evidence supports this proposition (e.g., Bremer 1992; Russett and Oneal 2001). For this reason, a dichotomous variable is coded 1 for any dyad sharing an alliance (i.e., defense pact, neutrality, or entente) and zero otherwise based on the formal alliance data set of the COW Project (Gibler and Sarkees 2004). Fifth, and finally, since the data set is structured in the form of a cross-sectional timeseries, there may be time dependence among dyads across years. Thus I opt for the method of peace year correction to test and correct possible temporal dependence. However, in accounting for time dependence, I do not use Beck et al.’s (1998) cubic splines because of arbitrariness about setting the knots. Rather, as suggested by Carter and Signorino (2010), I use cubic 2 3 polynomials of peace years (i.e., t, t , and t ). Carter and Signorino’s Monte Carlo study shows this method can perform as well as cubic splines and even better if knots are misspecified. Results and Analysis My statistical findings can be summarized as follows: 1) territorial claims increase the probability of a MID regardless of the regime types of dyads; 2) although the conflict-generating effect of territorial claims is present even with democratic dyads, it appears weaker among democratic dyads than nondemocratic dyads; 3) joint democracy decreases the probability of a MID whether or not dyads include territorial claims; and 4) the pacifying effect of joint democracy appears stronger in the presence of territorial claims. Table 3.5 reports probit regression results for the independent effects of territory and democracy on MID onset for all dyads from 1919 to 1995. All of the control variables appear to 112 have statistically significant, negative effects on the onset of MID. As expected, territorial claims seem to increase the probability that an interstate dyad goes to a militarized dispute. The empirically well-established pacifying effect of joint democracy is confirmed again, even when controlling for the effect of territorial issues. This finding surely weakens Gibler’s argument regarding a potentially spurious relationship between peace and joint democracy observed in the absence of considering territorial issues. Table 3.5. Probit Analysis of the Effects of Democracy and Territory on MIDs, 1919-1995. Variable Coefficient Robust St. Error Joint Democracy -.5462**** .0926+ Territorial Issue .7725**** .0669 ln Capability ratio -.0810**** .0127 Noncontiguity -9310**** .0645 ln Distance -.1668**** .0239 Minor power -.9172**** .0550 Alliance -.1212** .0593 Peace Years -.0330**** .0040 Peace Years 2 .0004**** .0001 3 Peace Years Constant N .0000**** .5138**** .0000 .1588 498689 1967.31**** -7201.5592 .3623 2 Wald chi Pseudo log-likelihood 2 Pseudo R *p<.1**p<.05, ***p<.01, ****P<.001 (one tailed). +Robust standard errors, adjusted for clustering on dyad, are in parentheses. Table 3.6. Four Categories by Democracy and Territory. Joint Democracy Territorial Claim Yes No Sum Yes Democracy-Territory (551) Democracy-Nonterritory (43,332) 43,883 113 No Nondemocracy-Territory (3,896) Nondemocracy-Nonterritory (450,910) 454,806 Sum 4,447 494,242 498,689 With respect to the contingent effect of democracy and territory on the probability of MID onset, the interaction of two dichotomous variables, joint democracy and territorial claims, generates four categories, as displayed in Table 3.6: 1) democratic dyads with territorial claims (democracy-territory); 2) democratic dyads without territorial claims (democracy-nonterritory); 3) nondemocratic dyads with territorial claims (nondemocracy-territory); and 4) nondemocratic dyads without territorial claims (nondemocracy-nonterritory). The parentheses in each cell contain the number of dyads falling in that category. As shown, democratic dyads appear to have territorial claims no less often than nondemocratic dyads. 32 About 1.3% of democratic dyad- years contain territorial disputes (552/43895), while 0.87% of nondemocratic dyadic years do so (3944/455525). Similarly, about 12% of territorial claim years contain democratic dyads (552/4496), while 8.8% of nonterritorial claim years contain democratic dyads (43343/494924). Table 3.7 reports probit results for differences in the effects on MID onset between democracy and territory, along with other control variables. In interpreting the coefficient on the interaction terms, recall that in the probit model, the linear predictor, XB , is the inverse cumulative density function of the probability of MID onset, Φ ( XB) . Therefore, the probit model can be represented as Φ −1 ( M ID = 1) = α + β 1 D + β 2T + β pT D + 32 k ∑ j=3 x jβ j + ε i This finding and the subsequent findings on the effect of democracy on MIDs are consistent with Kinsella and Russett (2002). They find the effect of democracy to be greater for preventing escalation of low level disputes into MIDs than for the initiation of low level disputes. 114 k where T and D refer to territorial claims and joint democracy, respectively, and ∑ j=3 x jβ j refers to the sum of all other effects of control variables. The interactive coefficient, then, can be obtained by the derivative with respect to democracy or territory. Table 3.7. Interactive Effects of Democracy and Territory on MID Onset, 1919-1995. Robust St. Variable Coefficient Error -.4591**** .0881+ Joint Democracy .8018**** .0662 Territorial Issue -.3481** .2080 Joint Democracy × Territorial Issue ln Capability ratio -.0805**** .0127 Noncontiguity -.9263**** .0645 ln Distance -.1679**** .0239 Minor power -.9164**** .0551 Alliance -.1208** .0589 Peace Years -.0328**** .0039 2 .0004**** .0001 Peace Years 3 .0000**** Peace Years Constant N .0000 .5103**** .1585 498689 2 2086.03**** Wald chi Pseudo log-likelihood -7196.4223 2 .3627 Pseudo R *p<.1, **p<.05, ***p<.01, ****P<.001 (one tailed). +Robust standard errors are adjusted for clustering on dyad. With respect to the pacifying effect of joint democracy, it appears to exist in both territorial context and nonterritorial contexts. The contingent coefficients for joint democracy are obtained by ∂Φ−1 (MID = 1) = β1 + β pT = −.4591 + (−0.3481)Territory . ∂Democracy 115 First, in the absence of territorial claims (i.e., T=0), the above equation is reduced to β1 (i.e., - 0.4591) that refers to the coefficient for joint democracy in the sample of dyads without territorial claims. This coefficient is statistically significant suggesting that democratic dyads without territorial claims are less likely to experience MID onset than other regime types of dyads without territorial claims. Second, in the presence of territorial claims (i.e, T=1), the rate of MID onset is still lower for democratic dyads than for nondemocratic dyads. In effect, such pacifying effect is strengthened in the territorial context because of the addition of another significant negative coefficient, -0.3481, to the coefficient of joint democracy. The amplified coefficient (-0.8073 = -0.4591 + -0.3481) of joint democracy in the sample of dyads with territorial claims is statistically significant. In sum, the results suggest that the pacifying effect of joint democracy increases with the presence of territorial disputes. This interpretation might seem counter-intuitive, but one must be careful in interpretation of the interactive effect. The larger effect of joint democracy in the sample of territorial dyads does not mean that democratic dyads with territorial issues are more pacific than democratic dyads without territorial issues. In fact, as Figure 3.3 shows, the democracy-nonterritory category is the most pacific among the four categories. Rather, the result suggests that the difference in the probability of a MID between democratic dyads and nondemocratic dyads gets larger in territorial claims. We can do the same thing for the interactive effect of territory. It appears that the conflict-generating effect of territorial issues, which is central to the territorial explanation, holds regardless of regime type of dyads. The interactive coefficients of territory are obtained by ∂Φ−1 (MID = 1) = β2 + β p D = 0.8018 + (−0.3481)Democracy . ∂Territory 116 First, as for the sample of nondemocratic dyads (i.e., D=0), the coefficient for territory is 0.8010, which is statistically significant. This result suggests that territorial claims increase the probability of MID onset among nondemocratic dyads. Second, this pattern also is found in the sample of democratic dyads (i.e., D=1). The coefficient of territory (0.4537 = 0.8018 – 0.3481), although reduced by 0.3481, is still statistically significant at the 0.05 level for democratic dyads. 33 This suggests that contention over territorial issues does increase the probability of MID onset among democratic dyads. Indeed, the pacifying effect of joint democracy is limited by the presence of territorial disputes. However, the significantly-reduced coefficient suggests that the conflict-generating effect of territorial issues is greater among nondemocratic dyads than democratic dyads. Put another way, although territorial issues increase the probability of armed conflict between two democracies, joint democracy appears to weaken the effect of territorial issues. In sum, the findings reveal two things: 1) the pacifying effect of joint democracy holds regardless of the presence of territorial claims, but the effect gets stronger in a territorial context; 2) the conflict-generating effect of territorial issues holds even among democratic dyads, but the effect is reduced. So far, the interactive effects of democracy and territory have been examined in terms of the beta coefficients and their conditional differences. However, this approach has yet to be complemented by considering actual changes in the probability of a MID from the interaction to deliver substantive meanings of the statistical findings. Some assert that this is a necessary step for interactive analyses of nonlinear models (Berry, DeMeritt, and Esarey 2010; c.f., Ai and Norton 2002). Following Berry et al.’s (2010) recommendation, I conduct the analysis of 33 I compute the statistical significance for this interactive coefficient using the Stata command, “lincom.” 117 probability changes using the CLARIFY simulation program (King, Tomz, and Wittenberg 2000). I estimate the probability changes based on the interactive probit model in Table 3.7. To do this I set all continuous variables at their mean. For dummies, I hold their categories to be conflict-prone to see how territorial issues and joint democracy affect the typical conflict-prone dyad (i.e., noncontiguity = 0, ally = 0, and minor power = 0). .3 Figure 3.3. Types of Dyads and Probabilities of MID Onset .1 Prob. of MID .2 Nondem-Terr Dem-Terr Nondem-Nonterr 0 Dem-Nonterr 1 2 3 Rank of Disputatiousness 95% CI 4 Mean Prob. Drawn from the results of the probability change analysis, Figure 3.3 compares each type of dyad in terms of the probability of militarized conflict. Each type is located in a probability point with its confidence interval. In terms of rank order, democracy-nonterritorial dyads appear most peaceful, followed by nondemocracy-nonterritorial and democracy-territorial dyads with 118 nondemocracy-territorial dyads as most conflictual. The probability difference between nondemocracy-nonterritorial dyads and democracy-territorial dyads, however, is not statistically significant, because its 95% confidence interval does include zero (unreported). More importantly, the interactive effect of democracy and territory becomes visually much more conspicuous. The vertical distance (i.e., absolute probability difference) between nondemocratic and democratic dyads is much wider in the territorial context than the nonterritorial context. Indeed, joint democracy appears to make a larger change with contentious issues like territory. For a numerical presentation, Table 3.8 reports the probabilities of MID onset, the first differences of the probabilities, the equivalent percentage changes, and the difference of the first differences. First, let us compare the probability changes (i.e., first differences) that joint democracy makes in the absence and presence of territorial claims, respectively. In the absence of territorial claims, joint democracy decreases the probability of a MID by 0.0433107 (i.e., first difference of moving joint democracy from 0 to 1 while territorial dyad = 0), which is about a 61.2% reduction. Conversely, joint nondemocracy increases the probability of a MID by 157.5%. In the presence of territorial claims, joint democracy decreases the probability of a MID by 0.1785813 (i.e., first difference of moving joint democracy from 0 to 1 while territorial dyad = 1), about a 70.7% reduction. Conversely, joint nondemocracy increases the probability by 240.9%. The absolute value of the first difference appears larger in the presence of territorial claims: the difference of the two first differences is -0.1352707 (i.e., -0.1785813 − -0.0433107). All these values of first difference and difference of the first differences are statistically significant since the 95% confidence intervals do not include zero values. Therefore, we can conclude that the pacifying effect of democracy is stronger in reducing the probability of a MID in the presence of territorial claims. 119 Types of dyads Nonterritorial Territorial Nondemocratic Table 3.8. Interactive Influence of Territory and Democracy on the Probability of a MID Interactive Effects Change in Probability of a MID changes First diff. B−A % change (Diff. of first diff.) X from A to B (∆ Prob.) Joint Dem 0 1 Joint Dem 1 0 Joint Dem 0 1 .0708 .0275 (.0509, .0935) (.0159, .0430) .0275 .0708 (.0159, .0430) (.0509, .0935) .2527 .0741 (.2069, .3042) (.0307, .1396) -.0433 (-.0613, -.0278) +.0433 (+.0278, +.0613) -.1786 (-.2343, -.1091) +157.5% Joint Dem 1 0 .0741 .2527 (.0307, .1396) (.2069, .3042) +.1786 (+.1091, +.2343) +240.9% Territory 0 1 .0708 .2527 (.0509, .0935) (.2069, .3042) +.1819 (+.1442, +.2218) +256.9% -61.2% -70.7% The negative effect of (Joint Dem 0 1) is strengthened in the territorial context by: -.1353 (-.1880, -.0687) The positive effect of (Territory 0 1) is weekend in democratic context by: .2527 .0708 Territory -.1819 -72.0% 1 0 (.2069, .3042) (.0509, .0935) (-.2218, -.1442) .0275 .0741 Territory + .0466 -.1352707 +169.6% 0 1 (.0159, .0430) (.0307, .1396) (+.0041, +.1119) (-.1880, -.0687) Democratic .0741 .0275 Territory -.0466 -62.9% 1 0 (.0307, .1396) (.0159, .0430) (-.1119, -.0041) Note: The numbers in the parentheses are the 95% confidence intervals for the probability estimates and their changes. 120 Second, I compare the effects of territorial issues between democratic dyads and nondemocratic dyads. Among nondemocratic dyads, the first difference of territorial claims (i.e., moving territorial dyad from 0 to 1 (from 1 to 0) while joint democracy = 0) is 0.1819005 (0.1819005). This is about a 256.9% increase (72% reduction) in the probability of MID onset. Among democratic dyads, the first difference of territorial claims (i.e., the probability change made by moving territorial dyad from 0 to 1 (from 1 to 0) while joint democracy = 1) is 0.0466298 (-0.0466298). This is about a 169.6% increase (62.9% decrease) in the probability of MID onset. The positive probability change made by territorial claims appears to be lower in democratic dyads than nondemocratic dyads: hence, subtracting the former from the latter produces a negative difference of the first differences, -0.1352707 (i.e., 0.0466298−0.1819005). All of these first differences and the difference of the first differences are statistically significant. Therefore, we can conclude that territorial issues increase the probability of a MID both between democracies and between nondemocracies, but that effect is less prominent among democratic dyads. All of my analyses show a significant interactive effect between democracy and territory in influencing the likelihood of a MID. The significant interactive effect is in favor of the democratic peace thesis. Democratic dyads continue to interact peacefully with each other more than nondemocratic dyads whether or not there are contentious issues such as territory at stake. When territorial issues are under contention, the ability of democratic dyads to use nonviolent means for settlement seems to get more conspicuous vis-à-vis that of nondemocratic dyads. In line with territorial explanations, I have found that disagreement over a piece of territory significantly increases the likelihood that two countries go to armed conflict, which appears to be 121 the case even for democratic dyads. Yet, the conflict-generating effect of territory is reduced significantly in democratic dyads. I need to qualify the interactive finding, however, due to a potential problem associated with using all dyads. In my sample, all dyads without territorial claims are lumped together while some of those nonterritorial dyads may have other nonterritorial contentious issues. Therefore, it is less than clear whether the increased pacifying effect of joint democracy in the territorial context supports (a) the selectorate theory that democracies tend to devalue territorial issues compared to other issues or (b) the information perspective that joint democracy is a conflict management rather than disagreement-inhibiting factor. It would be ideal to differentiate these dyads from conflict-free dyads using appropriate dummies. However, currently there is no such data available. Although the ICOW project has collected information on various issues including maritime, river, and regime claims as well as territorial claims (it also plans to expand data collection to include issues like ethnic, linguistic, religious, and other identity-based claims), this data collection has yet to be completed over the entire globe. CONCLUSION Democracy matters in international conflict processes, based on the results obtained here for MIDs, even in the presence of territorial issues and regardless of research design, measurement, and model specification. To summarize, I found that Gibler’s critical analysis against the democratic peace is not replicable. Joint democracy remains as a solid pacifying factor that reduces interstate dispute. In fact, all of Gibler’s three main border variables do not significantly explain MID onset far from confining the effect of joint democracy. Of the four border controls, two variables produced supportive results for Gibler’s arguments. In any case, 122 none of these border-related variables was able to offset the pacifying effect of dyadic democracy. If there is anything important with borders, land contiguity reduces the effect of democracy to a certain degree. Yet, this conditioning effect of contiguity may not be limited for joint democracy. As Reed and Chiba (2010: 72) find that contiguous dyads behave differently than noncontiguous ones with many other observable variables such as alliance, trade dependence, capability ratio, major power status, etc.: “these behavioral differences account for a significant portion of the gap in the conflict probability between contiguous and noncontiguous dyads.” It also should be noted that contiguity per se does not denote the instability of a border or territorial threat while it constitutes the border itself. I used Huth and Allee’s identification of territorial claims as a measure for territorial issues that previous research found important to destabilize borders and facilitate interstate armed conflict. With this measure unlike Gibler’s measures, I found that territorial issues, which reign supreme in the steps-to-war explanation, do result in more MIDs even among democracies. Yet, the results show that the extraordinary power of democratic institutions to reduce conflict is present even under the most challenging conditions. Armed conflicts are less likely between democratic dyads than nondemocratic dyads even in the presence of territorial claims and such difference in the conflict proclivities is wider. It is especially interesting that the pacifying effect of joint democracy is greater for territorial disagreements in the sense that the gap in conflict propensities between democratic dyads and nondemocratic ones is wider for territorial than for non-territorial disputes. This result should be assessed with caution in that the nonterritorial dyads in my sample do not differentiate dyads with other serious issues from those that are relatively conflict-free because of data availability. 123 New directions for research are suggested by these interesting results in the nexus of democracy and territory and their qualifications. All of the data analysis should be replicated when, as noted above, the ICOW dataset that attempts to identify several outstanding issues at stake in interstate relations is complete (i.e., when it becomes available on a worldwide basis). Further tests for robustness of results are in order; for example, additional economic variables (e.g., Gartzke 2007; Mousseau 2009: 55-57) could be added to the research design. Another possibility concerns selection effects regarding the kinds of territorial conflicts experienced by democracies in the data set used here. Perhaps probing a subset of more difficult territorial disagreements--excluding, for instance, issues that are more likely ‘easy’ cases (however assessed here in footnote 30) or access to far-flung locations—would produce results more favorable to the IBA. Will the democratic peace thesis, which performs so well in the present research design, hold up under such tests? Another question concerns salience and (in)tangible issues. Recent studies establish that these more nuanced aspects will impact upon conflict escalation and management, so further analysis of democracy and territory in combination should take that into account. It could be revealing to probe further into how, once they are involved in MIDs, territorial issues affect democratic dyads regarding escalation into a full-scale war (i.e., using a research design similar to Senese (1997), Reed (2000) or Senese and Vasquez (2003)). The same can be said about the nascent research program that brings together territory, democracy and rivalry (Lektzian, Prins and Souva 2010). Interesting to ponder as well are potential internal effects from territorial conflict; for example, Hutchison and Gibler (2007) discover that territorial threats reduce political tolerance on a cross-national basis. 124 Potential causal processes embedded in the statistical findings deserve closer inspection through case studies. This is relevant to contemporary and future conflict management as well. For instance, what can be expected among Canada, Russia, the US and other states if climate change produces new sea routes in the high Arctic? At the other end of the Earth, sovereignty in Antarctica continues to produce controversy. And at many points in between, such as Gibraltar vis-à-vis Spain and Britain, tension either is intermittent or ongoing. The territorial explanation can inform us about why the democratic peace holds and why democratic dyads do not fight wars: (a) they do not have wars because democratic pairings of states tend to avert territorial MIDs; and (b) democratic dyads handle territorial claims better than do other dyads (Vasquez 2009). Thus it would be interesting to look at ‘deviant’ territorial claims—those that do not go to MID and inter-democratic MIDs that do not go to war—in a case study format. In closing, it is hoped that the present study will encourage further efforts to look further into the combination of territory and democracy as key variables in the explanation of international conflict processes. 125 CHAPTER FOUR PICKING THE FIGHT, WINNABILITY OR JUSTIFIABILITY? DEMOCRACY AND THE SELECTION OF ARMED CONFLICT Does political accountability lead democracies to choose conflicts more wisely? Further, by what metrics can we consider the choice of an armed conflict wise or unwise? By attempting to answer these questions, this chapter seeks to shed light on the linkages between conflict choice, democratic institutions, and beneficial foreign policy outcomes, often referred to as the democratic “selection effects” argument. In short, do democracies tend to win wars because they choose easier fights? Focusing on the initiation stage and factors that effect selective initiation, I directly examine democratic selectivity and identify its specific mechanisms that have been underspecified in the literature. The debate over the selection effects argument has drawn lively interest in the conflict literature. Much of this interest has focused on demonstrating whether democracy is a prime mover for the production of beneficial conflict outcomes. For instance, many studies have found that in international conflict, democracies tend to win, produce fewer battle deaths, and fight shorter conflicts, among other things (e.g., Lake 1992; Siverson 1995; Bennett and Stam 1996, 1998; Mitchell and Prins 1999; Reiter and Stam 2002). Various democratic accountability-based explanations have commonly suggested democratic selectivity as the key to explaining the beneficial foreign policy outcomes of democracies (e.g., Schultz 2001b; Reiter and Stam 2002; Bueno de Mesquita, Smith, Siverson, and Morrow 2003). However, remarkably little attention has been paid to testing whether democracies wisely choose armed conflict initiation and targeting in the first place, and if so, what factors effect such 126 wise selection. Addressing these issues is crucial because democratic selectivity, along with victorious outcomes, touches upon the very essence of the democratic selection effects argument. Empowered domestic constituencies, civic institutions, and marketplaces of ideas in democracies may make democratic leaders concerned about factors that are of little concern to autocrats in cost-and-benefit calculations of conflict initiation and targeting. In this chapter, I argue that extant research explaining, testing, or disputing the relationship between democracy and smart selection is seriously flawed in two related respects. First, it is too outcome-oriented to directly (dis)confirm democratic selectivity. The proposed democratic selectivity is only assumed but rarely directly tested mainly because existing studies select on the dependent variables. In other words, while choosing a weaker target more directly refers to initiation of conflict, the evidence has been accrued in terms of desirable outcomes from the samples of dyads already involved in wars or lower levels of armed conflict. This is also the case with critics of democratic selectivity. To cite just one example, a renowned realist, Desch (2008; see also Downes 2009), points out an analytical gap in the selection effects argument: while the selection effects argument starts from the peacetime decision making processes and goes to the initial stage of armed force, all of the supportive findings provided by the proponents come from the data at the wartime stage. However, he himself fails to meet this requirement in his critical appraisal of the selection effects argument as discussed later. Second, extant research’s unbalanced focus on the outcome stage of conflict results in underspecification of some important factors concerning what entails a smart choice of conflict. Indeed, the proponents are silent about specific selection standards that democracies use when they choose foreign enemies. If not, nothing more than relative materialistic capability is considered as an important selectivity factor that influences decision to enter into armed conflict. 127 There is little reason that democratic leaders bearing high political accountability costs only consider the prospect of victory and material capabilities. War initiation and its successful conduct require mobilization of people and their resources. Such mobilization needs justification. In need of public support for war mobilization, democratic leaders should consider whether their aggressive foreign policy against a chosen enemy is justifiable and agreeable to a broader spectrum of domestic constituencies. In this chapter, I reframe selective conflict decisions of democracies in terms of winnability and justifiability. I propose that political accountability motivates leaders to choose armed conflicts with justifiable and agreeable causes as well as a high probability of victory. I explicitly explore two kinds of factors that may affect how democracies choose their enemies in the first stage of armed conflict: (1) winnability factors such as relative capabilities, geographical proximities, and topographical features that make access difficult; (2) justifiability factors such as political differences, human rights practices, and commercial relations. My proposition is tested empirically by examining the contingent effects of these factors on armed conflict according to the regime type of initiators. In order to assess democratic selectivity properly, I look at the initiation of militarized interstate disputes (MIDs) from peacetime. This method allows us to examine not only whether but also how democracies are selective in the initial stage of interstate conflict, lessening statistical complications caused by possible selection bias (Reed 2000; Schultz 2001a). This chapter will unfold as follows. The second section outlines past research, theories and findings relating democracy and successful war performances, along with relevant critiques. The next section proposes several corrections for theorizing and testing the democratic selection 128 effects argument. The fourth section will focus on research design. The fifth section provides data analysis. The final section will discuss this chapter’s implications. DEMOCRACY AND WISE SELECTION: MOVING BEYOND VICTORIES AND CAPABILITIES Why have democracies been so successful in international conflict? While few disagree with this pattern per se, the reasons for that are less than clear. Briefly, some focus on democracies’ high power status accrued through the growth-generating effect of democratic 34 institutions (Lake 1992; see also Bueno de Mesquita et al. 2003; Schultz and Weingast 2003). Yet, democracy remains a statistically significant factor even when controlling for military capabilities and the level of economic development (Reiter and Stam 1998a, 2002; Bueno de Mesquita et al. 2004). A consensus seems to have emerged pointing to accountability-embedded selective decision making processes in democratic institutions regarding initiating use of force (Reiter and Stam 2002; Bueno de Mesquita et al. 2003; see also Siverson 1995; Schultz 2001b). Although developed ex post just as with explanations of the democratic peace, the selection effects argument has been proposed as the common explanation for the impressive successful war records of democracies. Indeed, all three major variants of the institutional democratic peace theory agree with democratic selectivity as a main cause for successful war performances. First, the public-consensus constraint model argues that democracies are 34 Some scholars emphasize other factors like the fighting effectiveness of democratic soldiers based on democratic norms (Reiter and Stam 1998b, 2002), the military effectiveness exerted by human capital, harmonious civil-military relations, and Western culture closely related to democracy (Biddle and Long 2004), and the reliability of democratic allies based on transparent democratic systems (Choi 2004). While these might be important factors for the democratic war success on their own ground, these factors are beyond the scope of this present study. 129 constrained from engaging in a risky, unpopular foreign adventure (Morgan and Campbell 1991; Bueno de Mesquita and Lalman 1992; Siverson 1995; Reiter and Stam 1998a, 2002). Regular and fair elections empower citizens and make democratic leaders susceptible to public appraisal. Leaders’ policies must represent closely the values and preferences of the people. Foreign policy is no exception. As a result, democratic leaders select conflicts that would have high chance of victory, low cost, and favorable outcomes. Second, the selectorate theory emphasizes the public goods commitment of democratic leaders who must adjust to the political environment of a large winning coalition size (Bueno de Mesquita et al. 1999, 2003). This model holds that political success and reelection depends on satisfying a winning coalition of the relevant selectorate. As the size of the winning coalition increases (decreases) relative to the size of the selectorate, the leaders’ survival becomes dependant on providing public goods (private goods). Leaders with large winning coalitions have desperate need to avoid policy failures because they cannot buy off the royalty with private goods. In relation to democracy and foreign policy, democratic leaders, usually having a large winning coalition, are required to have greater confidence of victory before initiating force and to put forth greater efforts to win once engaged in military conflict. By contrast, autocratic leaders with a smaller winning coalition can compensate for foreign policy failures by devoting resources to privately benefit supporters. In short, democracies are more careful, selective, and hard-working in starting and conducting military conflict, as compared to nondemocracies. Finally, the information perspective centers on the informational properties of democracy (Fearon 1994; Siegel 1997; Schultz 1998, 1999, 2001b; Partell and Palmer 1999; Choi and James 2007). For democratic leaders, it is hard to conceal and misrepresent true preferences, resolve, and constraints against international opponents. This is so because of distinct institutional 130 features of democracy, such as effective opposition parties (Schultz 1998, 2001b), media freedom (Siegel 1997; Choi and James 2007), and/or high audience costs imposed upon democratic leaders (Fearon 1994). These institutional features constrain leaders from bluffing, which in turn makes democratic leaders selective and allows them to send credible signals in international crises. For instance, people in democracies are assumed to dislike seeing their leaders’ retreat in international crises and have the ability to punish those who make threats and later fail to carry out due to electoral institutions (Fearon 1994; Partell and Palmer 1999). Bearing this high audience cost, democratic leaders should avoid targets that are likely to resist and fight tough. Also, democracies find it difficult to conceal weak public support for war and misrepresent their preferences due to free presses and domestic opposition parties that compete openly for electoral support (Siegel 1997; Schultz 2001b). Conversely, if there is strong public support for war, the opposition parties and the free presses confirm the veracity of the government’s resolve. As Schultz (1999: 237) states, “institutionalized competition constrains a government to be more selective about resorting to threats, while at the same time, improving the effectiveness of the threats it does make.” Other things equal, public support for war increases as the chance of a favorable war outcome increases. Thus democracies have incentives to choose targets of military threat that are unlikely to reciprocate and fight tough. In sum, democratic political institutions render political leaders accountable to a wide range of constituencies. Concerning the choice of international conflict, democratic accountability provides leaders with “incentives… to launch only short, winnable, low-cost wars, so they may avoid domestic political threats to their hold on power” (Reiter and Stam 2009: 194). This selection effects argument seems to explain the various democratic success stories. Democracies are more likely to win the wars they fight than autocracies (Lake 1992). This 131 victorious effect is strengthened especially when democracies initiate (Reiter and Stam 1998a; see also Gelpi and Griesdorf 2001). Democratic initiators suffer much fewer casualties than nondemocratic initiators (Siverson 1995). Military threats made by democracies are less likely to be resisted violently by target states than that of nondemocracies (Schultz 1999; 2001b). Even the inconvenient historical fact of democratic colonial imperialism is proposed as evidence supportive of the selection effects argument: “democracies engage in colonial wars and imperial wars against much weaker adversaries” (Bueno de Mesquita et al. 2003: 241, emphasis added). Democracies are reluctant to participate in and escalate armed conflict unless there is an overwhelming military advantage (Bueno de Mesquita et al. 2004). In addition, democracies are found to fight shorter wars in general, and wars and crises are even shorter especially when democracies are initiators or both initiators and targets are democracies (Bennett and Stam 1996, 1998; Bueno de Mesquita, Koch, and Siverson 2004). Despite substantial empirical evidence for the democratic benefits, critics have challenged the selection effects argument. According to the critics, the proponents have failed to consider the constraints imposed upon autocrats to engage in a risky war (Desch 2002, 2008; Rosato 2003; cf., Chiozza and Goemans 2004), incorporate the possibility that democratic politics can lead to poor decision makings (Kaufman 2004; Cramer 2007; Downes 2009), and test directly its causal mechanisms with a proper research design (Desch 2008; Downes 2009). These empirical and theoretical caveats stem largely from the proponents’ fascination with the impressively victorious war records of democracies. Indeed the proponents have neglected various factors that underplay the decision making processes moving a state from peacetime to conflict initiation. These weaknesses are somehow understandable in that the proponents’ theoretical efforts have proceeded to follow the seemingly interesting historical facts 132 related to democracy, which may be considered the most desirable or the least vicious form of political systems that self-interested human beings have invented so far. However, these weaknesses are significant and worth proper corrections given the scholastic interest the topic has generated in the field and the world that is increasingly democratic. Largely two corrections are needed: (1) the proponents have claimed for the wise selection regarding democracy without appropriate, direct testing; and (2) they have defined it too ambiguously and narrowly. In regard to the first point, the observation of a renowned realist critic, Desch (2008: 4, 48), rings true: if “democracy makes states better in making winnable wars before they start,” the direct evidence should come from “the peacetime decision making process in the first stage, not from the second, wartime stage.” However, he himself fails to avoid the same problem he has charged for the proponents’ findings. One common problem with these findings is that the previous research focuses on the subsets of dyads already involved in MIDs or wars. 35 By focusing on the war stage, not on the peacetime decision making processes, Desch’s process-tracing study is unable to look at “dogs that did not bark.” As he finds according to his own standards, if there were only ten cases of wars initiated by democracies since 1815, such rarity would suggest that democracies have been deliberate in initiating wars. In terms of data, more direct evidence comes from the tests that examine whether democracies are really selective in their decision to initiate armed force in the first place, especially at the lower levels that fall short of war, such as MIDs. By definition, all MIDs could potentially end up as wars, and all wars are a subset of MIDs and must escalate from some lower levels of dispute behavior. This empirical approach is also theoretically sound and more consistent with the selection effects argument proposed by the various proponents (Schultz 2001b; Bueno de Mesquita et al. 2003; 35 This is prone to selection on the dependent variable (see Geddes 1990; King, Keohane, and Verba 1994). 133 Bueno de Mesquita et al. 2004). If the survival of democratic leaders is really a function of conflict outcomes, they require a substantially higher ex ante probability of victory before making a threat and initiating any armed force, not just full scale wars. With respect to the narrow but ambiguous definition of the wise selection, the proponents have been silent about the selection criteria that selective democracies may take into account when they chooses targets of military coercion. Instead, they ambiguously use “winnable,” “weak,” “easy” or “tough” as the adjectives that describe democracies’ preferable targets of military attack. One exception exists that tests the wise selection both explicitly and directly. Anderson and Souva (2010) examine all directed dyads for the 1970-2000 period to analyze the interactive effect of political accountability and relative military power on armed conflict initiation. The relative capability measure interacted with political accountability is a clear indication that the authors define wise selection as the tendency to avoid (target) a stronger (weaker) enemy in terms of materialistic military capabilities. They find that as the size of winning coalition (W) increases, the positive effect of military advantage on MID increases. This is a direct empirical support for the selection effects argument because W is thought to be a measure for political accountability that increases conflict selectivity. In effect, Anderson and Souva’s test is an improvement of Bueno de Mesquita et al.’s (2004 World Politics article) test 36 on the same hypothesis but with the dyads already involved in militarized disputes and wars. However, the accountability-based selectivity is never tested with a democracy measure while the other two institutional explanations strongly argue for democratic institutions as the main source of accountability and selectivity. Also, some scholars question the validity of W as a 36 By selecting on militarized disputes and wars, Bueno de Mesquita et al. are unable to consider dyads that did not have armed conflicts and focus on participations or involvement rather than initiation. 134 measure for accountability compared to the commonly employed measure for democracy from the Polity IV project (Clark and Stone 2008; Marshall, Gurr, and Jaggers 2010). Most importantly, the reference to wise selection is not limited to only materialistic military factors that increase the probability of victory. According to the logic of the selection effects argument, a wise selection in a democracy means that rational leaders make foreign policy in the way that enhances the chance of reelection. The prospect of achieving goals in military operations matters for the public support for leaders and their foreign policy to the extent that the public’s tolerance on the human costs of wars rises with a high likelihood of success (Gelpi, Feaver, and Reifler 2005/6, 2009). Victory is a necessary step for the success in war so balance of military power should concern the costs-and-benefits calculations of democratic leaders in the decision to initiate armed force. But victory is not sufficient to achieving goals that accrue the public support. The war aims must make sense to the public and it depends on the rightness or wrongness of war as well as the benefits that victory would bring about (see Gelpi, Reifler, and Feaver 2007; Gelpi et al. 2009). This is the very aspect of selective conflict decision making that existing theoretical and empirical studies of the democratic selectivity are mostly silent about. In summary, both the proponents and the critics have relied upon war outcomes with the subsample of actual war participants. This outcome-oriented approach has led to indirect statistical tests of the democratic selection argument, unreliable selection of cases, complications of measurements regarding war participants and outcomes, and theoretical inattention to factors going underway the selective decision making of war. In the next sections, I propose and test several refinements of the accountability-based selection effects model, focusing on conflict initiation and justifiability factors that shape conflict selection. 135 SELECTION MECHANISMS: WINNABILITY VS JUSTIFIABILITY What matters to the decision of democracies to initiate armed conflict? Previous research on democratic selectivity has only considered the winnability dimension associated with the probability of quick and easy victory. This exclusive focus misses another important dimension of careful conflict decision made by democratically elected leaders: democratic leaders should be able to justify the causes of military conflict and they can make up a justification for attacking a certain kind of enemy that would not be easy to defeat. I argue that the logic underlying the selection effects suggests that deliberate democracies choose not just easy conflicts but also justifiable conflicts. I reframe democratic selection of armed conflict initiation in terms of justifiability as well as winnability, building upon insights from the accountability-based explanations of the democratic peace (Fearon 1994; Schultz 2001b; Huth and Allee 2002; Reiter and Stam 2002; Bueno de Mesquita et al. 2003). I proceed with the common assumption about democratic accountability that, although both democrats and autocrats are office-seeking, the former hold more accountability than the latter for policy choices and outcomes. Unlike autocrats, democrats must earn 51% of the public’s votes while interplaying with empowered political rivals, actively monitoring free media, and vociferous independent experts. These all work together to impose the requirement on democratic leaders that an enacted policy in a democracy satisfy a broad spectrum of domestic constituencies. With respect to foreign policy, the democratic accountability is thought to allow leaders to be very careful in initiating armed conflict compared to autocratic leaders, because armed conflict and security issues are probably the most salient foreign policy matters. One underspecified implication of democratic accountability in the conflict literature is that decision makers need to persuade the public and earn its consent for 136 their military operations if international relations really matter for the public in evaluating the incumbents as commonly assumed in the literature (Schultz 2001a). Hence, in a democracy what count for office-seeking democratic leaders before initiating military operations that can always turn out to be risky and costly are the public’s mindset and costs-and-benefits calculation. Public consent takes on effects both at the time of election and at the time of policy enactment. First, in an election, other things being equal, voters elect candidates who appear to be most competent to deal with the current external challenges at hand or impending ones in the foreseeable future. Regarding incumbents, voters’ expectations of future competence are likely formed on the basis of incumbents’ past foreign policy performances. In election, voters’ ex post evaluations punish or reward incumbents’ past foreign policy performances. Therefore the public’s electoral control allows selecting out incompetent and reckless leaders and offers incumbents incentives to be deliberate and selective in making of foreign policy so that their chosen courses of foreign policy actions bring about beneficial outcomes. Second, the electoral control also allows public opinion to concurrently affect foreign policy at the time of enactment. As Reiter and Stam (2002: 6) correctly state, “voters and leaders alike tend to focus on the matter at hand. Leaders in liberal democracies seek out contemporaneous approval for political choices.” Other democratic institutions also provide ample sources of the concurrent consent effects such as effective political opposition parties (Schultz 2001b), free media (Choi and James 2007), legislative constraints (Clark and Nordstrom 2005; Choi 2010), active political participations and civic cultures (Putnam 1993; Reiter and Tillman 2002). These institutional sources require leaders’ policy choices to be in line with the public sentiments. Therefore, democratic leaders are constrained from initiating unpopular international conflicts. 137 Figure 4.1. Democracy, Accountability, and Choice of Armed Conflict Democracy Accountability Public Evaluation & Public Consent Popular Foreign Policy Winnable Conflict Justifiable Conflict The most important problem to this present study is: what kinds of international conflicts are unpopular to initiate? Previous research has only taken into account the winnability dimension. Democracies are presumed to be reluctant to initiate armed force unless they have a high ex ante chance of quick and easy victory with minimal human costs. Initiating armed force against a difficult opponent should be unpopular to voters. Yet, the required high chance of easy and quick victory makes a far less complete story for the accountability-based selective conflict decision mechanism. The most famous examples include the US decision to participate in World War I and World War II. At this point in the argument, popular sentiment comes to the fore. I argue that voters do not just consider materialistic benefits that victory will bring about but also the rightness or wrongness of use of force against a particular target state to their eyes. Conversely, leaders need to justify their military operation, and such justification should touch on the public’s sentiments. Otherwise, even if a government’s warlike foreign policy had been victorious, if it had caused the government to lose the hearts and minds of people, voters would punish incumbents’ inattentiveness to their sentiments. Public sentiment is also central to the discussion of this study’s main contributions to the literature. Ironically, while probably most directly related to the public consent models of 138 accountability (Fearon 1994; Schultz 2001b; Reiter and Stam 2002), justifiability is the very aspect of selective conflict decision making about which relevant existing research is mostly silent. Especially, as quoted below, the Reiter and Stam model explicitly suggests that the democratic institutional requirement for concurrent consent can reduce the intensity of the requirement that proposed or enacted policy bring about successful outcomes. However, Reiter and Stam make an inadequate theoretical leap from the justifiability dimension toward the winnability dimension due to their commitment to explain the successful war performances of democracies based on the concurrent consent model. While correctly identifying the importance of popular sentiment, Reiter and Stam (2002) take it only as an underlying assumption in explaining the contingent effect of initiation of war and democracy on victory rather than directly develop testable hypotheses. According to the Reiter and Stam (2002: 6) consent model, the key to the public support is the justifiability and agreeability of a particular policy rather than its success and failure: “Voters … punish leaders not so much for particular failures or successes, but instead for failing to heed the more popular sentiments at the time the leaders settle on a particular policy.” If leaders’ policy choices are supposed to be consistent with public opinion, it is reasonable to argue that their decision to initiate armed force against a target must appeal to public sentiments. Democratic leaders must choose to initiate conflict for the reasons that they can justify or the public can agree with, and must be unwilling to attack even much weaker targets without agreeable and justifiable causes. For instance, the US should have an overwhelming chance to win a military invasion against its neighbors like Canada and Mexico. But the US is not likely to do so because military invasion of Canada or Mexico is seen as hardly just to the US constituencies. The justification requirement affects both retrospective and prospective voting at the time of election. Leaders who have 139 initiated unjustifiable and thus unpopular international conflicts will not appear sincere and competent, and voters will be likely to punish them. Even victory in those conflicts may not help the leaders’ reelection bids much. Conversely, leaders who had to initiate drawn or losing conflicts with just and agreeable causes may not appear seriously incompetent, and voters may consider the justifiable causes of unsuccessful conflicts when casting votes. Therefore, a more directly appropriate proposition from the consent based accountability model is that democratic leaders pick armed conflicts that are highly justifiable as well as winnable. That is, democracies are unwilling to initiate armed conflict without just and agreeable causes as well as high confidence of victory. The logic of justifiability helps explain an empirical anomaly against the outcomeoriented version of the selection effects argument. 37 One difficulty with the selection effects argument comes from the historical fact that autocratic leaders were removed at least as frequently as democratic leaders for fighting a losing or costly war, and in fact the former suffered much harsher postwar punishments than the latter for the same reasons (Rosato 2003). Also, some statistical findings are inconsistent with the accountability-based arguments of the democratic peace. Chiozza and Goemans (2004) find that defeat in a crisis or war only significantly reduce the tenure of mixed regime and autocratic leaders while war or crisis 37 Realists challenge the accountability mechanism that allows democratic leaders to be wise and selective in decision making of war and conflict. For instance, Desch (2002, 2008) and Rosato (2003) argue that starting a losing war incurs significant costs to autocratic leaders as well as democratic leaders, and even democratic leaders can be less cautious in that the postwar punishment for losing war is much more severe for autocratic leaders than democratic leaders (see also Goemans 2000). However, this may not be regarded only as counter-evidence for the accountability-selection mechanism. It is not implausible that democratic leaders’ fewer frequencies of losing wars, as a result, removal, exile, and death reflect their selective decision making they made before going to costly wars; few democracies go to losing war, and sometimes they do so only when they expect no severe punishments (Slantchev, Alexandrova, and Gartzke 2005; see also Schultz 2001a). 140 outcomes, no matter victory or loss, little affect the political survival of democratic leaders. However, these findings may not be as necessarily against the selection effects argument as they seem. These seemingly incongruent findings may fit within democratic selectivity. Democratic leaders sometimes cannot help but going to war with inevitable or justifiable causes despite uncertainty of quick and easy victory. When war is inevitable or justifiable, unfavorable outcomes will be less relevant to the fate of democratic leaders. Sometimes, the public push leaders to engage in military adventurism, and sometimes, leaders find it relatively easy to justify their military adventurism. This dimension of selection (i.e., justification) along with the other (i.e., quick and easy victory) can explain why democracies have infrequently involved 38 themselves in losing and costly wars. But if they have done so, their leaders are no more likely to be removed or punished compared to autocrats; the former must have chosen to go to those risky wars because they felt war would be unavoidable or justifiable. Let us identify factors that can make armed conflict more or less unpopular or unjustifiable for democratic leaders. HYPOTHESES I identify two kinds of selection factors—one affecting the justifiability of conflict initiation and the other affecting its winnability—and develop related hypotheses to test my main argument. Justifiability Factors What kinds of conflict initiation are against public sentiments in democracies? I argue that military aggression against three kinds of countries—those that are seen as morally just, 38 See the relevant numbers in Table 4 in Rosato (2003: 594). 141 share political and ideological values, and/or have substantial economic interdependence—will be exceptionally unpopular in democratic societies. Human Rights Practices I propose a possible interaction of potential challenger’s regime type and potential target’s human rights abuse in affecting conflict initiation. For a democracy, when its opponent has infamous human rights abuse problems, it can relatively easily generate public support for its military operation against this enemy. Although extant research on democratic selectivity, in general, has assigned little role for democratic norms and principles, they can qualify as a prime force that affects deliberate foreign policy making. For instance, several recent studies have shown that the US public opinion on war and vote choice are shaped by beliefs about the rightness or wrongness of the war and those about its likely success to achieve agreeable purposes (e.g., Gelpi et al. 2009). Attacking unjust foreign governments rather than just ones will be more likely to be perceived as a right thing by the public. Which kind of a country is seen as unjust, especially to a democratic society? I argue that democratic citizens consider unjust and immoral a foreign government that badly treats its domestic constituents by oppressing and abusing their freedom and basic rights. Violent oppression of the people by government force is directly against the principles of democracy that assure the importance of human dignity, free exchange of ideas on various issues, inalienable physical rights, and the like. According to these principles, all human beings are believed to deserve to pursue lives of dignity and integrity free from arbitrary arrest, detention, and torture, regardless of ones’ religious, political, ethnic, and socioeconomic origins. Democratic citizens socialized in this normative context may have greater abhorrence against human rights abuse abroad than their autocratic counterparts. In turn, leaders in democracies may find it relatively 142 easy to persuade the public about the justness of military force against possible targets fraught with human rights abuses. Conversely, it will be extremely difficult for democratic leaders to justify military attack against foreign countries with clean human rights records. Therefore, if people in democracies are highly concerned with human rights issues and influence foreign policy decisions on matters at hand, then I expect: Justification Hypothesis 1: Unlike its nondemocratic counterpart, a democratic potential initiator will be less likely to attack a potential foreign target as the latter’s human rights respect increases compared to that of the former. It should be noted that the above hypothesis contains two qualifications, one about the potential challenger’s perception of the potential target’s human rights performance and the other about the regime type of that potential challenger. The first qualification concerns how abusive and unjust a foreign country is with respect to human rights. Though presumed to be correlated, recent empirical studies find that democracy per se is not a robust causal force that boosts human rights performance (Bueno de Mesquita, Cherif, Downs, and Smith 2005; Cingranelli and Filippov 2010). Some even find that certain democratic regimes have inferior human rights practices than certain nondemocracies (Fein 1995; Regan and Henderson 2002; Butler, Gluch and Mitchell 2007). I believe that the perception of whether a foreign country is human rights abusing or respecting is comparatively shifted. A democracy with moderate human rights practices may not see a nondemocracy with fair human rights records as an attractive target of military incursion, because the former’s leaders would find it difficult to describe the latter as unjust to the public. Therefore, as the gap between the human rights practices of a potential 143 target and those of a potential democratic challenger (i.e., target scores - challenger scores) becomes widened negatively (positively), the former becomes a more (less) justifiable target of military incursion to the latter. The second qualification is germane to potential nondemocratic challengers. Citizens in nondemocracies would also, to some degree, abhor anti-humanitarian acts and consider abusing foreign countries as inhumane and unjust. But these people have little say on rulers and their policies, unlike people in democratic societies. Therefore, public sentiments on human rights issues may not be translated into actual foreign policy acts in nondemocracies. Rather, the comparative perception of potential targets’ human rights abuses and respect as compared to those of potential challengers may have an opposite effect among nondemocracies. That is, other things being equal, armed conflict may be most likely between a potential nondemocratic initiator with the worst human rights practices and a potential target with the best human rights practices. Nondemocracies with bad human rights performances would be criticized frequently abroad by well-respecting countries with good human rights records. These criticisms may harm the legitimacy of ill-respecting nondemocracies domestically. Their leaders may have incentives to convert the criticisms into a security threat and inflate it in order to divert people’s attention from domestic ill-treatment and rally domestic support. Sometimes, leaders would find use of force the most effective to this purpose, increasing the likelihood of armed force initiation against the criticizing countries. Target Democracy I propose that a democracy is unwilling to attack another democracy and that this unwillingness increases as the latter’s level of democracy increases compared to that of the 144 former. Democracies, though they can exist in various forms (e.g., parliamentary or presidential, and proportional or majoritarian), share political and ideological values that emphasize individual rights, freedom, fairness, and the public’s leverage over rulers. Attacking a fellow democracy would be rarely justifiable to a democratic public. As Owen (1994: 95) states, citizens in a democracy tend to believe other democracies, unlike nondemocracies, are “reasonable, predictable, and trustworthy, because they are governed by their citizen’s true interests….” Furthermore, according to Doyle (1983: 230), regimes that rest on the public consent presume each other to be “just and therefore deserving of accommodation.” In sum, the consensual and representing aspects of democracy are the key to democratic citizens’ perception of justice regarding foreign countries and their regime types. A foreign policy corollary is that a democracy’s military incursion into another fellow democracy rather than a nondemocracy tends to be seen as more unjust to the domestic constituencies. There is one important qualification in the argument above. That is, how democratic is a potential target state? Skeptics would argue that political elites, as opinion leaders (Rosenau 1961; see also Almond 1950), could manipulate the general public’s perception according to their interests. When elites need peace, they define a foreign country as friendly and democratic, but when they need war, they define it as hostile and undemocratic (Layne 1994; c.f., Owen 1994). A less informed and less attentive public would find it simply convenient to trust and follow the leaders’ opinion. This suggests that the regime type of a potential target may have little independent causal impact on a potential democratic challenger’s decision to initiate armed force. However, defining democracy is not wholly subjective, but rather inter-subjective. The public may perceive the democraticness of foreign countries comparatively vis-à-vis that of their own government. Although two countries share democratic constitutions, a more democratic 145 country would regard a less democratic country as undemocratic. Leaders may find it comparatively easy to fabricate a foreign country as undemocratic when its regime is regarded as less democratic than their own regime according to some objective standards. Conversely, leaders would have difficulty fabricating as undemocratic a country that is regarded as objectively more democratic than their own regime. Therefore, if the public in a democracy considers fellow democracies more just than nondemocracies, and if it perceives foreign countries comparatively, then I expect: Justification Hypothesis 2: Unlike its nondemocratic counterpart, a democratic potential initiator will be less likely to attack a potential target as the latter’s democraticness increases compared to the former. With respect to potential nondemocratic initiators, I measure targets’ democracy differently. Comparative democraticness may be inappropriate to measure target’s level of democracy when initiators are nondemocracies. It should be noted that people in nondemocracies would have few (if any) pure reasons to believe democratic regimes are more unjust and illegitimate than other nondemocratic regimes. In fact, citizens in many nondemocratic societies have aspired to the democratization of their own governments. However, even if they thought highly of foreign democracies as political systems, people in nondemocracies have little leverage on rulers and their foreign policies whatsoever. This suggests that the just cause requirement for armed force initiation is barely at work to constrain nondemocratic leaders. Rather, nondemocracies may find democracies more attractive targets of armed force initiation. 146 Interestingly, proponents of the accountability-based conflict theories have displayed contrasting expectations about the general attractiveness of democracies as targets of interstate conflict. For instance, Schultz (2001b) argues that democracies make attractive targets because the additional information provided by the democratic competitions creates opportunities for a potential challenger to probe the true intension of the former without increasing the chance of unwanted war. By contrast, Bueno de Mesquita et al. (2003) hold that the democratic accountability motivates democracies not only to selectively choose targets but also to fight hard once involved in war, so they make unattractive targets for war. I believe that the validity of these two contrasting statements is conditional upon the regime types of potential initiators especially at the initiation stage, allowing a possible theoretical reconciliation. It is plausible that nondemocracies bearing little accountability on foreign policy matters are little constrained from both bluffing behaviors and bringing unsuccessful conflict outcomes, unlike democracies. This provides nondemocracies with leverage in armed force initiation against democracies to conditionally bluff. A bluffing-capable nondemocracy can issue a threat against a democracy “that it will carry out in the event it observes domestic dissent but which it will not carry out if it observes unified support” from the targeted democracy (Schultz 2001b: 99). Although tough fighters once involved in war, democracies are careful and selective before engaging in armed conflict because they are constrained by the public sentiments. This may be also true when democracies confront and manage foreign military threat. Opinion-gathering and persuading the public takes time in democracies, allowing time and opportunities for nondemocratic initiators to lift threat when the situations evolve unfavorably. Of course, backing down can cost nondemocracies, but it should be better than stumbling into an unwanted war. 147 Therefore, a potential nondemocratic initiator will be more likely to initiate conflict against a potential target as the latter’s democraticness increases vis-à-vis that of the former. Economic Interdependence I propose an interactive effect of democracy and economic interdependence on armed conflict initiation. Liberals have long argued that trade promotes international peace. Although some question this proposition empirically and theoretically (e.g., Barbieri 2002), evidence has been accrued for it (e.g., Gartzke, Li, and Boehmer 2001; Oneal, Russett, and Berbaum 2003; Hegre, Oneal, and Russett 2010). I argue that use of force against trade partners will be exceptionally unpopular in democracies for two reasons. The first reason is that military conflict reduces trade ex ante when expected (Morrow 1999) and ex post when unexpected (Li and Sacko 2002), which in turn will be detrimental to the political survival of democratic leaders. In the field of economics, trade is thought to promote economic growth (e.g., Edwards 1993, 1998; Krueger 1997; Lawrence and Weinstein 1999). Also in mainstream economics theory, trade increases the overall welfare of a society (see Oatley 2005: Ch. 3). Hostile actions may ruin economic benefits that many domestic constituents have enjoyed from good economic relations with the targeted state. Thus, maintaining and promoting trade is of greater importance for democratic leaders than autocratic leaders because the former’s political survival is a function of providing broader policy successes (Bueno de Mesquita et al. 2003; Gelpi and Grieco 2008). The second reason is related to transnational ties and friendship between two trading partners that their economic exchanges have brought about. Economic interdependence through bilateral trade increase contacts, communications, and transnational ties all of which facilitate 148 mutual understanding and respect between two societies. The increased familiarity and understanding help distinct societal preferences converge on a broader cosmopolitan identity and friendship (Deutsch, Burrell, and Kann 1957). Attacking a country that an initiator’s citizens consider familiar and friendly will be hardly acceptable. The public will be unlikely to endorse its government’s hostile action against economically friendly countries. Therefore, democratic leaders, whose political life depends upon the median voter’s approval and who have reasons to heed the general people’s hearts and minds, are highly motivated to maintain peaceful relations with the nation’s important economic friends. Taken together, when a democratic government issues a military threat against an important trading partner, the government will find it difficult to justify its military agenda to and attain public support from the opposition parties, media, and people. Therefore, I hypothesize: Justification Hypothesis 3: Unlike its nondemocratic counterpart, a democracy will be less likely to choose a country with thick economic ties as the target of its military attack than a country without. Regarding nondemocracies, I expect that the pacifying effect of economic interdependence would not exist for potential nondemocratic initiators. Some nondemocratic leaders would respond to rising trade integration in a similar way that democratic leaders are determined to avoid initiating armed force against important trading partners. But nondemocratic leaders, in general, bear little political accountability and public constraints, the workhorses for interdependence to promote peace (see Gelpi and Grieco 2008). Barbieri (1996, 2002) has suggested that economic interdependence heightens the likelihood of interstate conflict. Her 149 argument is in line with the realist argument that due to the anarchic imperatives of international politics the reliance on foreign economies increases vulnerability that can be exploited by enemies in time of crisis or war (Waltz 1970; Mearsheimer 1990a, 1990b). When increasingly dependent on foreign markets, nations become desperate to extend the scope of their control to the source of vulnerability. Therefore, for realists, “interdependence... lead[s] to greater [security] competition, not to cooperation” (Mearsheimer 1990a: 45). This vulnerability scenario of trade integration would be a workhorse for nondemocracies that may have many reasons to inflate the threat while trading off the possible benefits of integration the public has enjoyed. Winnability Factors According to the accountability-embedded selection effects argument, careful and selective democracies try to pick easy targets and avoid difficult ones in initiating armed force. Yet, the proponents have not explicitly identified factors that make particular targets or conflicts easy or difficult other than military capabilities. Also, few have directly tested whether democracies really pick easy fights. It is implied in the literature that an easy target is a country against which a democracy has a high chance of quick victory with few expected battle deaths. I attempt to identify easy or difficult targets in terms of factors expected to affect the probability of quick and less costly victory. Relative Military Capabilities Other things being equal, armed conflict against militarily weaker foes should have a higher chance of quick victory and cost less military deaths as compared to conflict against equal or stronger targets. Many studies show that relative power ratio is an essential explanation for 150 conflict outcome (e.g., Reiter and Stam 1998a; Schultz 1999; Desch 2002). The democratic “selection effect” argument suggests an interactive effect of initiator regime types and power ratio between potential initiators and potential targets. 39 If democracies are concerned about conflict outcomes more than autocracies, the former should be less likely to choose militarily 40 stronger opponents than do the latter. Winnability Hypothesis 1: The presumed negative effect of military disadvantage on the likelihood of conflict initiation is stronger among democratic initiators than nondemocratic initiators. Terrain and Distance Certain topographic features make countries more difficult to attack. Areas of difficult access offer safe haven to potential target countries. For instance, the mountainous borders of Switzerland have provided strong defense against the European powers like Italy, Germany, 39 For instance, Reiter and Stam (2003: 169, emphasis added) explicitly note, “Democracies might simply pick on weaker states, thereby arriving on the battlefield with more of the material determinants of victory.” Also, Bueno de Mesquita and his colleagues have made similar statements: “Weak opponents are attractive targets for democracies” (Bueno de Mesquita et al. n.d.: 6, emphasis added); Large-coalition systems… provide incentives for leaders to pick on much smaller rival states…” (Bueno de Mesquita et al. 2003: 219, emphasis added); “Relative to autocracies, proportionately fewer democracies are expected to be observed waging war when their ex ante military advantage over the rival is not overwhelming” (Bueno de Mesquita, et al. 2004: 365, emphasis added). 40 It should be noted that, as commonly found in the bargaining literature, regardless of regime types, leaders as agents decide to initiate conflict only when the benefits of doing so, especially in terms of retaining office, exceeds what they expect from a negotiated settlement or compromise rather than only when they are near certain of victory. However, as Bueno de Mesquita et al. (2003) formally show, the more likely a state is to win, the more likely this condition is met. 151 Austria, and France. The battlefields in jungle terrain seriously deterred effective full functioning of US military force in the Vietnam War. Rough terrain (e.g., jungle, swamps and mountains) hinders the mobility and deployment of troops for both foreign invaders and defenders but favors the defenders by providing them with the opportunities to hide (Stam 1996). Also, fighting and survival in rough terrain requires special tactics and techniques that are not easily acquirable, especially for foreign forces. Therefore, rough terrain makes it difficult for attacking states to achieve quick victory. Indeed, the war history tells us about those who ignored these geographic factors paid the price (e.g., Napoleon and the Oberkommando des Heers of Nazi Germany in their attempts to conquer Russia and the Soviet Union respectively). Likewise, Bennett and Stam (1998) find that terrain has a significant impact on war outcome and duration. In relation to democracy, careful democracies are unlikely to risk initiating forces against potential targets of difficult access. Therefore, I hypothesize: Winnability Hypothesis 2.1: The presumed negative effect of rough terrain on conflict initiation is stronger among potential democratic initiators than nondemocratic ones. Physical distance between states is also an important geographical factor that affects the initiation and operation of military forces. Distance constrains states in exercising military forces effectively. As the distance between a fighting force and its headquarters increases, the power projection capacity significantly decreases. This is even true for a great power with the land, sea, and air capability to carry substantial forces far away. As Bueno de Mesquita states (1981: 83), the greater the distance for a state, “the smaller the proportion of its total capabilities that it can expect to use.” Deliberate and selective democracies, then, must consider geographic constraints 152 more seriously than nondemocracies. Therefore, I expect that democracies are less likely than nondemocracies to initiate military conflict against distant targets. Winnability Hypothesis 2.2: The negative effect of geographic distance on conflict initiation is stronger among democracies than nondemocracies. RESEARCH DESIGN I examine all the directed dyad years from 1981-1999 in the international system for which all the measures are jointly available. 41 The directed dyad analysis differentiates two states in a dyad into initiator side and target side (i.e., Sides A and B). I use EUGene software to generate the basic template of my dataset (Bennett and Stam 2000a). The dependent variable is initiation of MID, coded 1 if there was a new initiation of a MID by Side A state in a given dyad year (otherwise 0) according to the MID dataset (version 3.1). Since our theoretical and empirical focus is on armed force initiation, ongoing dispute-dyad years are excluded unless a new dispute is initiated in a given dyad year. The MID dataset provides two kinds of initiation indicators. One is coded as an initiation when a Side A state first crosses the MID threshold (threat or use of military force). The other is coded as an inititation when Side A state is the revisionist. I choose the second option because “revisionist” would better fit deliberate and willing conflict selection. Some initiators that used force first were not the states that provoked their target first. One oft-cited extreme example is the Poland-Germany 41 Use of Thompson’s (2001) rivalry list defines the ending point of this study’s time frame (1981-1999) while use of CIRI data defines the starting point as discussed below. 153 dyad in a dispute that developed into World War II, where Poland is coded as using force first while Germany was the actual intentional provoker. My hypotheses focus on how the effects of presumed selectivity factors differ between democratic and nondemocratic states. The initiator’s regime type is identified using the Polity IV dataset (Marshall, Gurr and Jaggers 2010). Polity2 score is used to facilitate the time series analyses of this chapter. This score ranges from 10 (a fully democratic dyad) to –10 (a fully autocratic dyad). Then, following the convention (e.g., Dixon 1994), if Side A in a dyad has a score of six or more it will be coded as a democracy, otherwise a nondemocracy. Then, in order to assess the proposed interactive effects, I divide the data into two subsets according to initiators’ regime type (i.e., democratic initiators vs. nondemocratic initiators). This scheme of splitting data allows more tractable and succinct analyses than the usual multiplicative interaction method (here in this chapter, multiplying the proposed selection-related interactive 42 factors with the initiators’ regime type). The justifiability hypotheses propose contingent effects of targets’ comparative human rights abuse, their comparative democraticness, and mutual interdependence on conflict initiation according to initiators’ regime type. The human rights practices of target states are measured using the physical integrity rights index of the CIRI human rights dataset available from 1981 through 2007 (Cingranelli and Richards 2010). This variable defines the starting point of this study’s time frame (1981-1999). Physical integrity rights refer to freedom from torture, political imprisonment, extrajudicial killing, and disappearance. The physical integrity rights index ranges 42 These two methods are technically equivalent, especially when most of the independent variables are hypothesized to interact with a dichotomous moderator variable (here in this study, initiators’ regime type). Achen (2002) strongly recommends the splitting method over the multiplicative one for the sake of sample homogeneity and model efficiency. Examples of studies that use the splitting method include Gowa (1999) and Miller (1999). 154 from 0 (no respect) to 8 (complete respect). I focus only on this sub-class of human rights among others such as empowerment rights and women’s political rights, because inflicting arbitrary physical harm on citizens will be seen as the most offensive to the universal ideas about human dignity (Keck and Sikkink 1998; Walsh and Piazza 2010). To measure this concept comparatively as hypothesized, I subtract Side A’s score from Side B’s score, resulting in a 17 point comparative index raging from -8 (comparatively worst human right performances of a target vis-à-vis an initiator’s performances) to 8 (comparatively best human right performances). Comparative democraticness of targets vis-à-vis democratic challengers is measured using the Polity2 score of the Polity IV project. I subtract Side A’s score from Side B’s score, producing a 41 point comparative index. This variable ranges from -20 (comparatively most autocratic) to 20 (comparatively most democratic). However, as for nondemocratic challengers, I measure target democracy differently based only on Side B’s Polity2 score, since there is little reason for nondemocratic challengers to think of targets’ level of democracy comparatively. One of the justifiability hypotheses suggests that democracies are unwilling to use force against economically important and friendly countries. Who is economically important and friendly? I regard each state’s trade dependence on the other as the dyadic trade level as a proportion of GDP. I use Gleditsch’s (2002b) expanded trade and GDP dataset 4.1 (1950-2000). Then, I measure economic interdependence as the minimum of trade dependence levels of two states in a given dyad year, following the weak link assumption (Dixon 1994). Although often used in non-directed dyadic analyses to operationalize the extent of transnational ties and the relative importance of an economy to its partner state (McDonald 2004), this measure is also consistent with my directed theory that transnational ties and friendliness as well as material benefits commercial trades generate are underlying forces to reduce conflict. “Not only how you 155 are important to my economy but also how I am important to your economy” is important in forming and shaping interstate friendship and interactions. Therefore, I should define economic interdependence in terms of mutual recognition. I believe that a less dependent state in a dyad is a stronger determinant of the mutual recognition of economic importance and friendliness that make violent interactions between two states more and more unthinkable (c.f., Russett and Oneal 2001). The winnability/difficulty hypotheses suggest that capability ratio, distance, geographical proximities between states A and B, and topographical features of state B are conditional on state A’s regime type. Capability disadvantage is measured as a natural logarithm of the state B/state A’s military capability ratio, based on the COW Composite Index of National Capabilities (CINC, v.3.02) (Singer, Bremer, and Stuckey 1972). This variable is expected to have a negative sign on MID initiation in both democratic and nondemocratic initiator samples, because regardless of whether democratic or nondemocratic, few states are willing to attack stronger enemies. However, the unwillingness of conflict initiation caused by capability disadvantages is expected to be more intense for potential democratic initiators than nondemocratic initiators. Geographic proximity matters to MID initiation. Following Oneal and Russett’s practices, I consider two related but different factors, contiguity and distance between two states. The first term is “NONCONTIG,” a measure that equals 1 unless two states do not share a land boundary or are separated by more than 150 miles of water. This dummy measure is based on the COW direct contiguity data (v.3.1) (Stinnett et al. 2002). I also employ the natural logarithm of the great circle distance in miles between the capitals of the two. I label this variable as “LOGDIST.” These two measures are not highly correlated in my sample (correlation 0.46). It is a well-established finding in the literature that the greater the distance between two countries, the 156 less chance state A initiates a MID against state B because of geographical constraints in exercising military force over long distance. These constraining effects of geographical nonproximities are expected to be larger for democratic initiators than nondemocratic counterparts. Also, certain geographical characteristics of targets affect conflict initiation by challengers. Rough or inaccessible topographical features like mountainous terrain, tropical swamps, or jungles make countries difficult targets. These terrains are poorly served by roads, favor defenders, and thus deter foreign attacks. I employ two measures for rough topography, mountainous terrain and tropical area. As for mountainous terrain, I use Fearon and Laitin’s (2003: 81) data for “the proportion of the country that is mountainous,” which are based on the codings of A.J. Gerard. I measure tropical topography as the proportion of the country area in tropical weather, based on Gallup, Mellinger, and Sachs’ (2001) geographical datasets. In addition to the main variables, I also include three control variables. First, other studies have found that a dyad involving at least one major power increases the probability of MID involvement and onset (Russett and Oneal 2001; Senese and Vasquez 2003). This has been taken into account in other directed dyad analyses of conflict initiation (Bennett and Stam 2000b; Gelpi and Grieco 2008). The measure for major power dyads is coded 1 if at least one state in a dyad is a major power and 0 otherwise. Second, recent scholarship on international conflicts has revealed that dyads of certain characteristics known as rivalries have produced a disproportionate number of the most striking interstate events including disputes, crises, and wars (Diehl and Goertz 2000; Colaresi and Thompson 2002). Rasler and Thompson (2001) argue for inclusion of rivalry information in predicting armed conflict in terms of political institutions. Controlling for rivalries will be likely to prove important in this study. Rivalry status of a target state may not only provide a reason for 157 an initiator to use force but also affect the latter’s conception of the former’s democraticness, justness, friendliness, etc. The presence of a rivalry relationship between states A and B is identified using Thompson’s (2001) list of interstate strategic rivalries. This list was created using historical research to identify whether or not two states perceived each other as main competitors and enemies. Alternative armed conflict-density measures like enduring rivalries (Diehl and Goertz 2000) are available, but they lack the perceptual element that is crucial for this study. In addition, the density-based measure causes a tautological problem to this present study, where the dependent variable is the initiation of armed conflict. Third and lastly, past conflict engenders future conflict. To capture the effect of conflict history, I include a measure for peace years. This variable is a count of the number of years since the last armed conflict in a dyad (Beck, Katz, and Tucker 1998). I use the adjusted peace years developed by Suzanne Werner (2000) to account for the timing of pre-1816 armed conflict. I also 2 3 include cubic polynomials of peace years (i.e., peace years and peace years in addition to peace years) as suggested by Carter and Signorino (2010). Carter and Signorino’s Monte Carlo study shows this method can perform as well as Beck et al.’s cubic splines and even better when knots are misspecified. RESULTS AND ANALYSIS What factors are taken into account for democracies’ selective decision to initiate armed force? Two kinds of factors, winnability and justifiability, are proposed as such factors for empirical investigation in this chapter. Existing studies seem to emphasize the winnability dimension by focusing on conflict outcomes. However, no studies explicitly theorize and test the 158 justifiability dimension of the democratic selectivity. A set of probit regression models analyze both dimensions simultaneously using all directed dyads for the 1981-1999 period. Table 4.1. Contingent Effects of Winnability Factors on MID Initiation, 1981-1999. Test for Democratic Nondemocratic Coefficient Diff. Challengers Challengers Dem—Nondem Winnability Capability Ratio -.0590628**** -.0534764**** -.0055864 (Target/Challenger) (.0144857) (.0113786) -.996805**** -.843575**** -.15323 Noncontiguity (.1188121) (.0931127) -.0906076** -.1774893**** .0868817^ Log-Distance (.0479192) (.0361045) Mountainous Terrain .0015296 -.0007696 .0022992 (% of Target) (.0014195) (.0012067) Tropical Area -.1080897 -.1478928** .0398031 (% of Target) .0948892 (.0665816) Controls .4909616**** .5738733**** Major powers (.0751031) (.0817518) .9443501**** .8260765**** Rivalries (.151818) (.0973257) -.0571158**** -.0410722**** Peace years (.0114415) (.0048222) -1.267241**** -.5207946**** Constant (.2738692) (.226043) N 161305 235634 2 735.48 1100.10 Wald Chi 2 0.3231 0.3666 Pseudo R Log-likelihood -1045.9498 -1893.5651 *p < 0.1, **p < 0.05, ***p < 0.01, ****p < 0.001, one-tailed. ^p <0.1, wrong direction, one tailed. Robust standard errors, adjusted for clustering on dyads, are in parentheses. The results 2 3 for t and t are significant in the expected directions but not reported in the table. I begin with a multivariate analysis of the winnability hypotheses. Table 4.1 reports the results on the winnability factors for democratic initiators and nondemocratic initiators. None of the winnability hypotheses are supported in these models without the justifiability factors. 159 Capability disadvantage has a significant, negative sign in both democratic and nondemocratic models. Though its estimated coefficient is slightly higher in the democratic sample, where democracies are in Side A, than in the nondemocratic sample, where nondemocracies are in Side A, such difference is not statistically significant. A Wald test performed by Seemingly Unrelated Estimation (SUE) fails to reject the null hypothesis of the same coefficients of capability ratio across the two populations, democracies and nondemocracies. Similar patterns are revealed for other winnability factors. While having negative signs as expected both for democracies and nondemocracies, two measures for geographical proximities do not appear to have a larger effect for democracies more than nondemocracies. Rather, the estimated coefficient of distance is larger for nondemocratic initiators than democratic ones, and this coefficient difference is marginally significant at the 0.1 level (one-tailed). Mountainous terrain is not a significant factor for either democratic or nondemocratic initiators. Additionally, though statistically insignificant, its coefficient sign is in the wrong direction for democracies. As expected, tropical area has a negative sign for both democracies and nondemocracies, but it is only statistically significant in the nondemocratic sample. In sum, the results suggest that democracies are no more reluctant to attack difficult targets that are strong, distant, and inaccessible than nondemocracies. According to this chapter’s main arguments, the models in Table 4.1 that only contain the winnability factors are underspecified. The null findings in a misspecified model may be subject to Type II error. Democratic leaders, bearing higher accountability and audience costs than nondemocratic leaders, need to persuade the public before going to military fight. Some conflicts are more justifiable than others, although not more winnable. Table 4.2 reports the results including justifiability factors such as comparative human rights respect by targets, their comparative democraticness, and economic interdependence. The inclusion of these 160 underspecified factors are not only important on their own ground but also helpful to identify the presumed democratic selectivity associated with the winnability factors. Table 4.2. Contingent Effects of Winnability and Justifiability Factors on MIDS, 1981-1999. Democratic Nondemocratic Test for Challengers Challenger Coefficient Diff. Winnability Dem−Nondem -.0724403**** -.0354361*** -.0370042** Capability Ratio (Target/Challenger) (.0174155) (.0125754) -1.198776**** -.8755654**** -.3232106** Noncontiguity (.1264775) (.0970504) -.1333774*** -.1866027**** .0532253 Log-Distance (.0514148) (.036487) -.000061 .0008466 -.0009076 Mountainous Terrain (% of Target) (.0016529) (.0012183) -.2150019** -.0328151 -.1821868* Tropical Area (% of Target) (.1075228) (.0768515) Justifiability Target Human Rights -.0666004**** .0613492**** -.1279496**** Practices (.016115) (.0113332) -.008009* .0091697** -.0171787*** Target Democracy (.0053863) (.0046677) -22.54575**** 2.599374** -25.145124**** Interdependence (6.750439) (1.383443) Controls .472962**** .5257052**** Major powers (.0759004) (.0790453) .7950973**** .8476192**** Rivalries (.1781728) (.1033805) -.0622043**** -.0435432**** Peace years (.0125924) (.0051822) -.8532641*** -.6559911*** Constant (.3065394) (.2364929) N 147836 197409 2 602.21 1056.77 Wald Chi 2 0.3543 0.3810 Pseudo R Log-likelihood -934.96521 -1635.9027 *p < 0.1, **p < 0.05, ***p < 0.01, ****p < 0.001, one-tailed. Robust standard errors, adjusted 2 3 for clustering on dyads, are in parentheses. The results for t and t are significant in the expected directions but not reported in the table. 161 Now, the findings for the interactive effects of several winnability factors turn into being in the hypothesized directions, with the presence of the justifiability factors in the equations. Capability disadvantage ratio appears to have a larger negative effect on MID initiation for democratic initiators (-0.072) than nondemocratic ones (-0.035), suggesting democracies tend to avoid stronger targets to a greater degree than nondemocracies. The coefficient difference for capability disadvantage is 0.037 in absolute value, which is statistically significant at the 0.05 level (one-tailed). The conflict-reducing effect of noncontiguity also appears to be larger for democracies (-1.20) than nondemocracies (-0.88). This coefficient difference is 0.32 in absolute value and statistically significant at the 0.05 level. This result suggests that democracies tend to initiate military operations in proximate places for strategic reasons to a greater degree than nondemocracies. However, the result regarding another measure for geographical proximities, logged distance, does not support this claim. There is little difference between democracies and nondemocracies in the tendency of a country to avoid initiating conflict against a far away enemy. The estimated coefficient of logged distance is statistically significant for both democratic and nondemocratic initiators. Though larger for nondemocracies, the difference in coefficients for distance is not statistically significant. Mountainous terrain still appears to have little impact on conflict initiation for both democracies and nondemocracies. Though its sign turns out to be in the right direction, the estimated coefficient of mountainous terrain is far short of statistical significance for democracies. In contrast to mountainous terrain, the winnability hypothesis regarding rough terrain and easy access finds weak support in terms of tropical area. Democracies appear to avoid attacking targets with large tropical areas fraught with swamps and jungles whereas this is not the case with nondemocracies. The negative coefficient for tropical 162 area is statistically significant only for democracies at the 0.05 level. Yet, this contingent effect is only marginally significant at the 0.1 level (one-tailed). Table 4.2 also contains the results for the justifiability factors. All of the justifiability hypotheses are supported in this table. The respect for human rights by targets vis-à-vis initiators has a significant, negative effect on MID initiation among democracies, suggesting that democracies are unwilling to attack countries that respect their citizens’ physical integrity well or better than themselves. Contrastingly, comparatively better human rights practices of target states than initiator states appear to increase the probability of MID initiation in the nondemocratic sample. This contingent effect is statistically significant at the 0.001 level. Clearly, as suggested by Justifiability Hypothesis 1, comparative respect for physical integrity by targets has sharply different effects between democratic and nondemocratic challengers. Challengers’ conception of the level of targets’ democracy also matters differently for democracies and nondemocracies. Comparative democraticness has a marginally significant, negative effect on MID initiation in the democratic sample. This suggests that democracies become increasingly unwilling to use force as their target of military incursion becomes increasingly democratic compared to their own democraticness. By contrast, in the nondemocratic sample, targets’ unitary level of democracy has a statistically significant, positive effect on MID initiation, suggesting that nondemocracies increasingly initiate conflict when their targets are increasingly democratic. Economic interdependence also appears to have different effects on MID initiation among democracies and nondemocracies. In the sample of democratic challengers, the impact of interdependence appears negative and statistically significant at the 0.001 level, suggesting interdependence decreases the likelihood of MID initiation among democracies. Contrastingly, in 163 the sample of nondemocratic challengers, interdependence appears to have a statistically significant, positive impact to increase the likelihood of MID initiation at the 0.05 level (onetailed). As theorized and hypothesized, democratic accountability makes leaders unwilling to risk unpopular conflicts against important economic partners, while economic interdependence makes nondemocratic leaders more concerned about being vulnerable to foreign markets. In general, the above results indicate that democratic leaders are selective in choosing foreign foes in international conflict. The winnability factors related to the probability of easy victory influence conflict initiation with different intensity between democratic challengers and nondemocratic ones. Also, democracies have different conflict selection criteria than their nondemocratic counterparts. The factors related to public justification have different kinds of effects between the two samples. With the need to be in line with the popular sentiments, democracies tend to choose justifiable as well as winnable conflicts. Therefore, any model of selective conflict initiation that neglects the justifiability dimension is susceptible to underspecification. In fact, in this chapter, the under-specification is found to produce incorrect nullfindings against the winnability dimension of democratic selectivity. With respect to the control variables, briefly, all findings are in the expected direction. Major power dyads and international rivalries experience more international conflict than other types of dyads. The estimated coefficients for major power dyads and rivalries are positive and statistically significant in all models. The interactive effects of the justifiability and winnability factors have been so far examined and discussed in terms of beta coefficients and their differences according to challengers’ regime type. To deliver more substantive meanings of my findings, I conduct the analyses of probability changes—i.e., changes in the predicted probability of the dependent 164 variable according to a change in an independent variable of interest—for the significant interactive factors found in Table 4.2. For this purpose, the CLARIFY program (King, Tomz, and Wittenberg 2000) is utilized. I set a baseline probability of initiation against which to make comparison by holding all continuous variables at their mean and setting all dummy variables at their modal category. Table 4.3. Contingent Probability Changes in Risk for Annual MID initiation by Statistically Significant Variables according to Challengers’ Regime Type. Democratic Challengers Nondemocratic Challengers Variables % Δ of Prob. Variables % Δ of Prob. Capability Ratio -51.60% Capability Ratio -26.51% Noncontiguity -98.81% Noncontiguity -95.12% Tropical Area -38.26% Tropical Area insig Human Rights -59.61% Human Rights +111.05% or -52.62% Target Democracy -18.61% Target Democracy +29.48% Interdependence -27.96% Interdependence +2.1% Note: insig. denotes statistically insignificant factors. Table 4.3 reports the results for the probability changes of MID initiation caused by moving the values of continuous independent variables by one standard deviation (ranging from -1/2 to +1/2 standard deviation) around their mean. As for dichotomous variables, the probability changes caused by moving from zero to one value are reported. The results are separately reported in Panel A (democratic challengers) and Panel B (nondemocratic challengers) according to challengers’ regime types. For democratic challengers, one standard deviation increase of target capability ratio decreases the probability of MID initiation by 51.60%. For nondemocratic challengers, the equivalent standard deviation change amounts to a 26.51% reduction. The effects of noncontiguity differ slightly between democracies and nondemocracies. Noncontiguity reduces MID initiation by 98.81% for democracies and 95.12% for nondemocracies. Tropical 165 area of target states reduces the probability of a military attack from democratic challengers by 38.26% as its proportion increases by one standard deviation. However, tropical area was not a significant factor to affect conflict initiation in the sample of nondemocratic challengers as reported in Table 4.2. For democracies, the comparative human rights conduct decreases MID initiation by 59.61% as it increases by one standard deviation around its mean. By contrast, for nondemocracies, the equivalent one standard deviation change in the difference of the physical integrity rights index leads to a 111.05% increase of armed conflict. Conversely, as the comparative difference in human rights conduct decreases by one standard deviation, the probability of MID decreases by 52.62%. The level of target democracy has a similar contingent effect, although it is measured differently for democratic challengers and nondemocratic ones. For democracies, the comparative democracy, measured as the difference in polity scores (target score minus challenger score) decreases MID initiation by 18.61% as its value moves by one standard deviation around the mean. However, this discrete probability change is fragile, since the conventional 95% confidence interval of the first difference does include zero, while the 80% confidence interval solely includes negative values. By contrast, for nondemocracies, one standard deviation change of target democracy around its mean (from about 7 to about 15) leads to a 29.48% increase of MID initiation. Lastly, the probability of MID initiation decreases by 27.96% as the level of economic interdependence increases by one standard deviation for democratic challengers. By contrast, the equivalent value change in the level of interdependence for nondemocratic challengers leads to 2.1% increase of MID initiation. This probability change is not only trivial in magnitude but also statistically insignificant. 166 CONCLUSION The various institutional theories of the democratic peace argue that bluffing and backing down in international crises, engaging in unnecessary, long, and costly wars, and losing wars are especially detrimental to political survival of democratic leaders. These foreign policy failures are thought to reveal incompetence of leaders, reduce national reputations, and imperil national security. Therefore, democracies have been assumed to be careful and selective in choosing international conflict. Yet, extant research has been too much outcome-oriented, emphasizing materialistic factors that would affect conflict outcomes. Even the over-emphasized materialistic factors have received scant empirical examination in order to test the presumed democratic selectivity more directly and explicitly. The key to democratic selectivity is unique democratic accountability. That is, democratic leaders must answer to the broad spectrum of constituencies regarding the outcomes of their foreign policies. My modified consent model adds the justifiability dimension to it, producing a more nuanced proposition. Concerned about the matters at hand, voters and leaders also interact at the time of foreign policy enactment. Voters consider the rightness or wrongness of impending armed conflict, so leaders need to justify causes for proposed foreign policy actions and voters need to agree with those causes. The institutional constraints exerted by vigorous media, powerful opposition parties, and well-established legislative bodies help citizens to be attentive to foreign policy issues and monitor government’s inappropriate foreign policy actions. These will enforce democratic leaders’ selectivity in the sense they will pick the fights they can justify. Democratic leaders who ignore these unique institutional constraints and engage in unpopular and disagreeable as well as unsuccessful international conflicts will hardly survive the appraisal of the domestic constituencies. 167 Based on these arguments, I test explicitly and directly whether democracies are selective in choosing foes in interstate conflict based upon the justifiability factors as well as winnability factors. The empirical findings I report in this chapter suggest that democracies are not only concerned about whether they can defeat easily their enemies but also whether they can persuade their constituencies about the justification of military attack against certain enemies. Unlike nondemocracies, democracies appear to be unwilling to start military fights with a country that treats its own citizens well, maintains a solid democracy, and is economically friendly and important. These factors significantly reduce the probability of MID initiation by democratic challengers. The inclusion of the justifiability factors also helps reveal the theoretically expected results for the winnability factors. Controlling for the justifiability factors, it is revealed that democracies are more likely than nondemocracies to avoid attacking targets of strong power and difficult access. These findings have important implications for future research. Although there are many findings supportive of the democratic war success hypothesis, critiques have continues to challenge the democracy-war success relationship by replicating Reiter and Stam’s (1998a, 2002) seminal studies (e.g., Desch 2002, 2008; Downes 2009). These critiques show that Reiter and Stam’s main finding—democratic initiators tend to win wars—is not robust to research design, measurement, and model specification, while Reiter and Stam (2009) dispute the critiques’ approach. One common flaw found in this debate is the failure to consider the justifiability dimension of conflict initiation. Democratic countries can go to war for just, grand causes even when they do not have overwhelming power advantages. Additionally, as justifiability of war increases the probability of victory may increases as well, because justifiability helps facilitate war mobilization that is necessary for successful war conduct. 168 Therefore, inclusion of justifiability factors may help reveal the true relationship between democracy and victory at war. The justifiability dimension of conflict selection can also help answer a question like which kinds of countries are susceptible to foreign attacks in general? The role of justice in international conflict has been neglected in the statistical studies. By considering justifiability, we can attempt to link the traditional just war thesis to the scientific study of war and conflict. In particular, it is reasonable to argue that governments that treat their subjects unfairly and unjustly (indicated by human rights abuse and political/economic inequalities) are vulnerable to invasions by hostile foreign forces. From the perspective of potential attackers, countries that are considered unjust domestically and internationally are more susceptible to invasion. First, it would be easier for an attacker to justify its military adventure to both domestic and international societies. Second, unjust countries’ armed forces are likely structured to oppress internal threat from the dissatisfied at the expense of efficiency in coping with external threat. 169 CHAPTER FIVE CONCLUSION: DEBATES, FINDINGS, LIMITATIONS, AND IMPLICATIONS At its origin, this dissertation touches upon two of the most essential concepts, democracy and peace (or war), in the study of politics and international relations. Each has drawn extensive scholarly interest on its own. The research program that examines the linkage between peace and democracy has been thriving to the extent that almost every single issue of each major political science journal contains at least one article that has a connection or reference to this subject. Generally accepted scientific findings are that democracies tend to interact peacefully with each other and that once involved in war, democracies tend to be victorious. However, the research program of democratic peace has not been without criticisms and problems. The purpose of this dissertation is to explore and help resolve the problems found in the top-notch debates between proponents and critics. In the previous chapters, I have examined and addressed the challenges posed by three recent important critiques against the democratic peace. The arguments, hypotheses, and findings from these critiques would nullify the well-confirmed evidence for the unique foreign policy benefits of democracy that has been accumulated over the past five decades. These criticisms have been unfolded in terms of commonality, territoriality, and winnability. Throughout the separate empirical and theoretical analyses, I have shown that these concepts are important on their own grounds but do not negatively qualify the well-established characteristics related to the democratic peace. Rather, addressing these concepts helps discern and clarify the important roles of democracy in impacting upon interstate relations. Indeed, democracy does matter in 170 international conflict processes, based on the results obtained in this dissertation. In what follows, I restate the motivations of this dissertation, summarize its findings, and acknowledge its limitations, discuss what I have learned from it, and provide its practical and academic implications. ISSUES AND PROBLEMS IN THE DEBATE In the introduction (Chapter 1), I review the research program of democratic peace by tracking its evolution, discussing its theoretical and empirical importance, and identifying the unresolved problems that would seriously impair our confidence in the democratic peace. The review finds that three problems remain unresolved: (1) the Cold War effect has never been directly and correctly assessed by either realist critics or liberal proponents to test whether common security interests induced by the Cold War had shaped international politics under the guise of the democratic peace; (2) the democracy-peace relationship has been confirmed and challenged without appropriate treatment of issues that are at stake in interstate relations; and (3) both proponents and critics have treated the concept of democratic selectivity in war as if it were only related to materialistic capabilities, physical difficulties, and outcomes of war. These three problems motivate the three general research questions. I ask the first question to examine the post-Cold War democratic peace: does the democratic peace persist or even get strengthened after the winding down of the Cold War? The second question is asked about the spuriousness or weakening of the democratic peace vis-à-vis territorial issues: do territorial issues limit the scope of the democratic peace or do they work as the real driving force behind the scenes? Lastly, the linkage among democratic accountability, selectivity, and winnability is questioned so as to refine and qualify: do political accountability and resulting 171 deliberation lead to smart selection of conflict, and if so, what defines a deliberate and smart choice for democracies at war? The three substantive chapters attempt to answer these general questions with the independent analytical frameworks. SUMMARY OF ANALYSES AND FINDINGS The first substantive essay (Chapter 2) tackles one prong of the realist-liberal debate on the true cause of the democratic peace. I compare and contrast realists’ and liberals’ arguments, expectations, and empirics to develop the hypotheses and key tests to appraise one’s explanatory power against the other. While realists hold that the democratic peace is a byproduct of the Cold War that generated common interests among democracies, liberals emphasize a macrohistorical progress toward a perpetual peace among liberal democracies. Both arguments are plausible but empirical evidence for each perspective is limited. For instance, realists never provide any direct measure of common security interests while liberals’ empirical domain does not exceed the end of the Cold War. My study goes beyond the Cold War period and directly measures Cold War common interests. I find that during the Cold War (1950-1989), while the measures for common interests have strong statistical effects on the probability of military conflict, the pacifying effect of democracy is not washed out, even after controlling for the common interests. In fact, against the realist expectations, the pacifying effect still exists for the post-Cold War period (1990-2001). By contrast, the difference in conflict propensities between democratic dyads and nondemocratic dyads is estimated to be progressively stronger over time as liberals argue. However, this difference itself is not statistically significant in the main analysis, although all other alternative analyses produce significant interactive results in favor of the liberals. 172 The second essay investigates the linkage between democracy and territory in impacting upon the onset of armed conflict. The research program of democratic peace has been relatively silent about the roles of issues that states have fought over. Foreign policy enactment and outcomes may vary by the kinds of issues at stake. Recent research on territorial conflict expands its empirical and theoretical canvas to identify the limitation of the democratic peace and even subsume its explanation. However, I point out the possibility that the effects of these factors are contingent upon each other rather than one muting out the other’s effect or one’s effect operating unconditionally regardless of the other’s moderation. While it is theoretically plausible, no study looks directly at the interaction between these two variables in affecting armed conflict. I conduct two separate statistical analyses. One is replication analysis of Gibler’s (2007) work on the stable border peace, where Gibler argues that democracy and peace are both symptoms, not causes, of the removal of unstable borders from the agenda of issues between neighbors. Questioning both its theory and empirics, I show that the results of Gibler’s analysis are not replicable. Instead, I find that even controlling for border variables, joint democracy remains an important factor in reducing armed conflict while Gibler’s border variables hardly have robust, consistent independent effects. The other is based on Militarized Interstate Dispute (MID) data and the Huth and Allee territorial claims list from 1919 to 1995. The results from interactive models are consistent with the conclusion that the pacifying effect of democracy stands up for both territorial dyads and nonterritorial ones, and the effect is reinforced in the territorial context in spite of the imperatives toward militarization created by territorial conflict. The third essay studies how democracies select their foes within interstate military conflict. Focusing on war outcomes, existing studies provide only mixed evidence that democracies carefully choose easy foes. However, these tests are at best indirect, failing to 173 analyze any specific causal mechanisms. I directly look at factors that affect how democracies choose their enemies in the first stage of military conflict. I reframe the selective conflict decisions of democracies in terms of winnability and justifiability. I theorize that political accountability motivates leaders to choose armed conflict with just or justifiable and agreeable causes as well as a high probability of victory. This argument is tested empirically by considering target countries’ comparative military capabilities, geographical constraints, human rights conduct, and commercial relations. The statistical results show that democracies tend to attack weaker foes to a greater degree than do autocracies, seek out countries with bad human rights practices as foreign targets, and are unwilling to fight with economically important trading partners. LIMITATIONS Although all of the three substantive chapters contribute to filling important theoretical and empirical gaps in the democratic peace literature, each has its own limitations. The first analysis is still limited in terms of time frame and measurement for a thorough investigation of the time varying effect of democracy. The post-Cold War period in this analysis spans just over a decade (i.e., 1990-2001). This 12-year time frame may be too short to let the data reveal the nature of the Kantian process characterized as a macrohistorical progression. In addition, the static, dyadic measurement of democracy does not allow an investigation of maturity effects or neighborhood effects of democracy that may condition or enhance the peace-inducing effect of democracy. The second analysis is unable to discern whether the enhanced effect of democracy in the dispute setting is unique to territorial issues. Because of the data limitation, territorial dyads are 174 contrasted against counterparts that are not keenly sorted out so that dyads without any conflict propensities and those with other contentious issues are lumped together to constitute the counter category, “nonterritorial dyads.” Therefore, a more adequate test for the presumed issues-varying effects of democracy must be assessed with the data that include refined and variegated types of issues. Paul Hensel’s (2001) ICOW data project is promising in this respect once its collection over the entire globe is completed. The third analysis directly addresses the problems associated with democratic selectivity and reveals the importance of non-materialistic factors in choosing and targeting international conflicts. However, it is incapable of proving whether the justifiability dimension of democratic selectivity is positively and significantly related to victory in war and conflict. Also, there is a problem with data limitation. The use of the CIRI human rights data limits the temporal domain substantially by not allowing data analysis for the pre-1981 period. Tracking old human right records cross-nationally does not seem quite feasible. Fortunately, the CIRI data is updated to include the year of 2009. When the MID data and other COW data for the independent variables are updated to cover more recent years, the data limitation problem will be lessened significantly. WHAT I HAVE LEARNED The debates between realists and liberals have long been focused on whether domestic politics and political institutions matter to foreign policy. Realists assume and prescribe states to be functionally alike regardless of regime types. Describing sates as unitary self-interested actors, realists argue that in order to survive or dominate, all states do and should behave in accordance with the imperatives of the international anarchy, often characterized by self-help and the security dilemma. In realism, only relative military capability is the key variable that 175 differentiates states in pursuit of national interests, often defined in terms of power as means for survival (Waltz 1979) or dominance (Mearsheimer 2001). In sum, realists picture world politics in a somewhat grim manner. By contrast, liberals and their democratic peace theory allow us to view the world in a brighter way. While taking self-interests and international anarchy seriously, liberals suggest that preferences are variegated within a state and the configuration of the variegated preferences within and across states shape world politics (Moravcsik 1997). They argue that these variegations in preferences and their formation impose binding constraints on state behavior, often resulting in positive foreign policy outcomes. In this exploration of the democratic peace debates, I have identified three key caveats in the literature and found supportive evidence for liberals rather than realists. My theoretical and empirical analyses show that unlike realist arguments, democracy does matter for international relations. Indeed, realists’ conception of self-interests is unrealistic and misleading. Individual self-interests do not necessarily coincide with national interests. Realists ignore the dynamics of domestically variegated preferences and its roles in influencing international politics in a rather positive way. I have learned that the institutions-embedded accountability is the key that makes democratic foreign policy different from nondemocratic foreign policy. I also find that libertarian norms and humanitarian values have influence on foreign policy decision making in democracies. This does not mean that in contrast to autocrats, democrats are inherently good, ethical, and just. Norms and institutions go hand in hand. Norms consolidate institutions and institutions nurture norms. Especially, democratic institutions provide milieus where norms can be effective forces in making foreign policy. I believe that leaders in democracies and those in nondemocracies are equally self-interested, politically ambitious, and strategic enough to achieve and maintain top 176 political positions. Rather, democratic institutions hold office-seeking leaders answerable to a broad spectrum of self-interested constituents with polarized preferences. Accountability imposed upon leaders will make them unwilling to make controversial foreign policy choices and deliberate to make agreeable policy possibly with desirable outcomes. The configuration of diverse preferences through democratic institutions tend to produce beneficial foreign policy outcomes such as peace with other democracies, victorious and inexpensive war outcomes, reliable international commitments and cooperation (Gaubatz 1996; Leeds 1999; McGillivray and Smith 2000; Broz 2002; Mansfield, Milner, and Rosendorff 2002; Leventoğlu and Tarar 2005), and even increased foreign direct investment inflow (Jensen 2006). Indeed, liberalism and the democratic peace theory show that how politics among self-interested agents can produce favorable outcomes in international relations, through accountability-enhancing institutions. I should acknowledge an important inability of my dissertation. Although this dissertation research has identified diverse sources that relate democracy to various positive aspects of world politics, it was unable to identify which institutional component among others is the prime mover. Of the variants of the democratic peace theory, I find that the revised concurrent consent model, which considers the nonmaterialistic dimension of public sentiments as well as the winnability dimension, touches upon the most overarching and fundamental feature of democratic politics in explaining a variety of democratic peace-related phenomena. Some other variant explanations do not collide with the public consent model in their basic internal logic. For instance, does neither the information models (Fearon 1994; Schultz 1998, 2001b) nor the selectorate theory (Bueno de Mesquita et al. 1999, 2003) oppose that public support is of key influence contemporaneously at the time of decision making more to democratic politics than autocratic politics. Especially, Fearon’s audience cost model and Schultz’s theory of political opposition take for granted the 177 concurrent roles of public consent. The public is assumed to dislike leaders who decide to back down in international crises (Fearon 1994), or opposition parties are thought to reveal whether the public concurrently supports or opposes the incumbent party’s foreign policy (Schultz 1998, 2001b). I have shown that because of the requirement of the concurrent public consent, democratic leaders only choose international conflicts that the public finds supportable. However, what is unclear is which specific characteristic of democratic institutions is of the most importance in enhancing accountability and empowering public opinion so as to make democracies selective, producing the democratic peace-related phenomena. Different models even in the same vein emphasize different institutional features of democracy. Some scholars focus on electoral institutions (Fearon 1994; Reiter and Stam 2002), while others focus on political participations (Reiter and Tillman 2002), legislative constraints (Clark and Nordstrom 2005; Choi 2010), effective opposition parties (Schultz 1998, 2001b), free media (Siegel 1997; Choi and James 2007), winning coalition size (Bueno de Mesquita et al. 1999, 2003), liberal values (Owen 1994), or institutional characteristics. Future studies should delve more directly into these factors to examine which is the most responsible to the democratic peace or whether they should coexist to be effective. IMPLICATIONS Both the findings and limitations carry important implications for the real world and future research. First, can we expect a more democratic world to be a more peaceful world? This proposition has mistakenly been regarded as a corollary of the lower dispute proclivities of democratic dyads compared to other types of dyads. However, this is not necessarily the case. The lower propensities refer to the gap between the democratic dyads and the nondemocratic 178 ones in the dispute propensities. Note that the gap in the armed conflict rates between democratic dyads and nondemocratic dyads can be wider when both of the rates increase if the latter’s rate goes up faster than the former’s. It is also possible that a reduced number of nondemocracies, each of which is having more chances to meet democracies as its pairs, fight more with an increasing number of democracies. Besides, the small number of nondemocracies may feel threatened by other democracies and their pushes for democratization, and thus form an antidemocratic alliance. For these reasons, this system-level implication of the dyadic peace must be assessed with an appropriate system-level analysis. Despite this difficulty, there is some favorable evidence toward that system-level direction. Recently, two scholars offer some strong evidence of the dynamics in the democracypeace relationship (Cederman 2001; Mitchell 2002; see also Mitchell, Gates, and Hegre 1999). For instance, Cederman finds that the gap in the conflict propensities between democratic dyads and autocratic dyads gets wider over time despite the downward conflict trend found in both democratic and autocratic dyads. Cederman infers that this unexpected finding may be due to the role of democracies as inspirers of peaceful norms of conflict resolutions in the system. This inference is in line with Kant, who presumed a spillover effect of democracy. What is unsatisfactory with Cederman’s study is that his dyadic analysis is unable to consider such spillover effect, despite his macro-level perspective. If the diffusion, deepening, and transformation of democratic learning process are macrohistorical, the systemic units of analysis will be desirable. In relation to the spatial dynamics, Mitchell (2002) explores the idea of democratic pacifying neighborhoods. She highlights Kant’s belief that “democracies will extend their domestic rule of law to an international rule of law, such that the norms characteristic of 179 democratic interaction will extend to the international arena” (p. 751). In other words, as Cederman (2001) also emphasizes, democracies are “inspirers” or sources of international norms that mitigate the strains of power politics imposed by the anarchic world system. Mitchell finds that democratic norms become more pervasive as the fraction of democracies in the system increases, to the point where even the behavior of non-democracies begins to change in a more democratic way, adopting the democratic norm of third-party dispute resolution. She also finds that for democratic dyads, the probability of third-party settlement increases when the proportion of democracies increases. Mitchell’s earlier study with Gates and Hegre (1999) focuses on the systemic level analysis of the democracy-peace relationship. Exploring the evolutionary and endogenous relationship between democracy and war, their Kalman filter analysis shows that democratization tends to follow war, that democratization decreases the systemic amount of wars, and the pacifying effect of democracy on war increases over time. A related interesting question is whether these findings hold up for a less violent level of conflict such as MIDs at the systemic level. The answer is probably yes, given that the democratic peace proposition receives a lot of evidence at the MID level as well as at the war level of conflict. Also, the findings of this dissertation suggest a similar expectation. The finding in Table 2.3, Chapter 2 suggests that the baseline probability of MID onset is reduced for both democratic dyads and nondemocratic dyads in the post-Cold War period while the gap in the rates of armed conflict becomes wider in the same period. Future research based on the system level and MID data will help reveal the true pattern. With respect to the interaction of democracy and issue types, the analysis in Chapter 3 should be expanded to consider other important nonterritorial issues when the ICOW data project is completed. Doing so can reveal whether the interactive finding in this dissertation is a function 180 of the dispute management capacity of democracies or their special unwillingness to fight over a certain issue like territory. Issues can make a difference in world politics—perhaps on par with type of government. Although more states become democracies and the world more peaceful, conflict between human groups may not decrease to a great degree. They may continue to struggle for control over the contentious issues that take on important values. Surely, more issueoriented or issue–informed studies will help us to find clues about the vast and complex characteristics and causes of conflict. The world early in its 21st century is fraught with contentious issues and problems such as global warming, terrorism, energy shortage, the proliferation of nuclear weapons, drug trafficking, ethnic and religious conflicts, open seas, open skies, shortage of fresh water, the North-South debate, various infectious diseases, and others. The regime type-focused studies must be expanded to incorporate the nuances that these various issues can make against peaceful managements of conflict. My third analysis on democratic selectivity could go one step further to investigate the roles of justifiability factors in producing victorious outcome at interstate disputes and wars. Once included in the analyses like those in Schultz (1999, 2001b) or Reiter and Stam (1998a; 2002), justifiability factors may help find the important roles of democratic selectivity in producing favorable outcomes of wars and crises. One of the justifiability findings related to human rights conduct carries also an important suggestion for the future research. It is revealed that democracies tend to attack countries that mistreat their own citizens. In terms of military effectiveness as well as justifiability, this may be reasonable. It is hard to expect that sons and daughters of the suffering and ill-treated people are willing to fight to death to defend their government against foreign invasions. Also, the ill-treating government’s armed forces may be too inefficiently organized to cope with external threats because of their need to oppress the 181 internal threat posed by the disgruntled. Autocracies also may find this kind of government more vulnerable, other things being equal. With an appropriate operationalization of ill-treating government based on indicators of human rights abuse and political and economic inequalities, we can examine an interesting question, how countries that are considered unjust and unfair domestically are internationally vulnerable in terms of foreign attacks and defeats. In that way, we can scientifically assess the roles of normative factors in influencing the occurrence and outcome of war and conflict. 182 REFERENCES 183 REFERENCES Achen, Christopher. H. 2002. “Toward a New Political Methodology: Microfoundations and ART.” Annual Review of Political Science 5(1):423-450. Ai, Chunrong, and Edward C. Norton. 2003. “Interaction Terms in Logit and Probit Models.” Economics Letters 80(1):123-129. Allee, Todd, and Paul Huth, 2006. “Legitimizing Dispute Settlement as Domestic Political Cover.” American Political Science Review 100(2):219-34. Almond, Gabriel. 1950. The American People and Foreign Policy. New York: Praeger. Alvarez, Michael, Jose Antonio Cheibub, Fernando Limongi, and Adam Przeworski. 1996. “Classifying Political Regimes.” Studies in Comparative International Development 31(2):336. Anderson, Sally and Mark Souva. 2010. “The Accountability Effects of Political Institutions and Capitalism on Interstate Conflict.” Journal of Conflict Resolution 54(4):543-565. Babst, Dean. 1964. “Elective Sociologist 3(1):9-14. Government: A Force for Peace.” The Wisconsin Barbieri, Katherine. 1996. “Economic Interdependence: A Path to Peace or Source of Interstate Conflict?” Journal of Peace Research 33(1):29-49. Barbieri, Katherine. 2002. The Liberal Illusion: Does Trade Promote Peace? Ann Arbor: University of Michigan Press. Barro, Robert. 1997. Determinants of Economic Growth: A Cross-Country Empirical Study. Cambridge, MA: MIT Press. Beck, Nathaniel, Jonathan Katz and Richard Tucker. 1998. “Taking Time Seriously: TimeSeries-Cross-Section Analysis with a Binary Dependent Variable.” American Journal of Political Science 42(4):1260-1288. Belsley, David A., Edwin Kuh, and Roy E. Welsch. 1980. Regression Diagnostics: Identifying Influential Data and Sources of Collinearity. New York: Wiley. Berry, William. D., Jacqueline H.R. DeMeritt, and Justin Esarey, 2010. “Testing for Interaction in Binary Logit and Probit Models: Is a Product Term Essential?” American Journal of Political Science 54(1):248-266. 184 Bennett, D. Scott. 2006. “Towards a Continuous Specification of the Democracy-Autocracy Connection.” International Studies Quarterly 50(2):513-537. Bennett, D. Scott, and Allan C. Stam. 1996. “The Duration of Interstate Wars, 1816-1985.” American Political Science Review 90(2):239-257. Bennett, D. Scott, and Allan C. Stam. 1998. “The Declining Advantages of Democracy: A Combined Model of War Outcomes and Duration.” Journal of Conflict Resolution 42(3):344366. Bennett, D. Scott, and Allan C. Stam. 2000a. “EUGene: A Conceptual Manual.” International Interactions 26(2):179-204. Bennett, D. Scott, and Allan C. Stam. 2000b. “Research Design and Estimator Choices in the Analysis of Interstate Dyads: When Decisions Matter.” The Journal of Conflict Resolution 44(5):653-685. Bennett, D. Scott and Allan C. Stam. 2007. “Documentation for EUGENE.” Version 3.2. Biddle, Stephen, and Stephen Long. 2004. “Democracy and Military Effectiveness: A Deeper Look.” Journal of Conflict Resolution 48(4):525-546. Box-Steffensmeier, Janet M., Dan Reiter, and Christopher J.W. Zorn. 2003. “Temporal Dynamics and Heterogeneity in the Quantitative Study of International Conflict.” In Economic Interdependence and International Conflict: New Perspectives on an Enduring Debate, eds. Edward D. Mansfield and Brian M. Pollins. Ann Arbor: University of Michigan Press, 273–89. Brambor, Thomas, William. R. Clark, and Matt Golder. 2006. “Understanding Interaction Models: Improving Empirical Analyses.” Political Analysis 14(1):63-82. Braumoeller, B. F. 2004. “Hypothesis Testing and Multiplicative Interaction Terms.” International Organization 58(4):807-820. Bremer, Stuart. 1980. “National Capabilities and War Proneness.” In The Correlates of War II: Testing Some Realpolitik Models, ed. J. David Singer. New York: Free Press, 57-82. Bremer, Stuart. 1993. “Democracy and Militarized Interstate Conflict, 1816-1965.” International Interaction 18(3):231-249. Broz, Lawrence J. 2002. “Political System Transparency and Monetary Commitment Regimes.” International Organization 56(4):861-887. Bueno de Mesquita, Bruce. 1981. The War Trap. New Haven, CT: Yale University Press. Bueno de Mesquita, Bruce, and David Lalman. 1992. War and Reason: Domestic and International Imperatives. New Heaven, CT: Yale University Press. 185 Bueno de Mesquita, Bruce, Michael T. Koch and Randolph M. Siverson. 2004. “Testing Competing Institutional Explanations of the Democratic Peace: The Case of Dispute Duration.” Conflict Management and Peace Science 21(4):255-267. Bueno de Mesquita, Bruce, Feryal Marie Cherif, George W. Downs, and Alastair Smith. 2005. “Thinking Inside the Box: A Closer Look at Democracy and Human Rights.” International Studies Quarterly 49(3):439-458. Bueno de Mesquita, Bruce, James D. Morrow, Randolph M. Siverson, and Alastair Smith. 1999. “An Institutional Explanation of the Democratic Peace.” American Political Science Review 93(4):791-807. Bueno de Mesquita, Bruce, James D. Morrow, Randolph M. Siverson, and Alastair Smith. 2004. “Testing Novel Implications from the Selectorate Theory of War.” World Politics 56(3):363388. Bueno de Mesquita, Bruce, James D. Morrow, Randolph M. Siverson, and Alastair Smith. n.d. “Testing the Selectorate Explanation of the Democratic Peace.” Typescript. Bueno de Mesquita, Bruce, Alastair Smith, Randolph M. Siverson and James D. Morrow. 2003. The Logic of Political Survival. Cambridge, MA: MIT Press. Butler, Christopher K., Tali Gluch, and Neil J. Mitchell. 2007. “Security Forces and Sexual Violence: A Cross-National Analysis of a Principal-Agent Argument.” Journal of Peace Research 44(6):669-687. Carter, David B., and Curtis S. Signorino. 2010. “Back to the Future: Modeling Time Dependence in Binary Data.” Political Analysis 18(3):271-292. Cederman, Lars-Erik. 2001. “Back to Kant: Reinterpreting the Democratic Peace as a Macrohistorical Learning Process.” American Political Science Review 95(1):15-31. Chiozza, Giacomo and Hein E. Goemans. 2004. “International Conflict and the Tenure of Leaders: Is War Still Ex Post Inefficient.” American Journal of Political Science 48(3):604619. Choi, Ajin. 2004. “Democratic Synergy and Victory in War, 1816-1992.” International Studies Quarterly 49(3):663-682. Choi, Seung-Whan. 2010. “Legislative Constraints: A Path to Peace?” Journal of Conflict Resolution 54(3):438-470. Choi, Seung-Whan, and Patrick James. 2007. “Media Openness, Democracy and Militarized Interstate Disputes.” British Journal of Political Science 37(1):23-46. 186 Cingranelli, David and Mikhail Filippov. 2010. “Electoral Rules and Incentives to Protect Human Rights.” Journal of Politics 72(1):243-257. Cingranelli, David L. and David L. Richards. 2010. “The Cingranelli-Richards (CIRI) Human Rights Dataset.” http://www.humanrightsdata.org. Clark, David, and Timothy Nordstrom. 2005. “Democratic Variants and Democratic Variance: Examining Domestic Processes and Interstate Conflict.” Journal of Politics 67(1):250-270. Clark, Kevin, and Jonathan P. Olmsted. 2009. “Selection Bias and Other Misspecifications: When the Cure is Worse than the Disease.” Paper presented at the 2009 Midwest Political Science Association Conference. Clark, Kevin, and Randall Stone. 2008. “Democracy and the Logic of Political Survival.” American Political Science Review 102(3):387-392. Colaresi, Michael, and William R. Thompson. 2002. “Strategic Rivalries, Protracted Conflict, and Crisis Escalation.” Journal of Peace Research 39(3):263-287. Colaresi, Michael, and William R. Thompson. 2003. “The Economic DevelopmentDemocratization Relationship: Does the Outside World Matter?” Comparative Political Studies 36(4):381-403. Cramer, Jane Kellett. 2007. “Militarized Patriotism: Why the U.S. Marketplace of Ideas Failed before the Iraq War.” Security Studies 16(3):489–524. Dafoe, Allan. 2011. “Statistical Critiques of the Democratic Peace: Caveat Emptor.” American Journal of Political Science 55(2):247-262. Desch, Michael. C. 2002. “Democracy and Victory: Why Regime Type Hardly Matters.” International Security 27(2):5-47. Desch, Michael. C. 2008. Power and Military Effectiveness: The Fallacy of Democratic Triumphalism. Baltimore, Maryland: Johns Hopkins University Press. Deutsch, Karl W., Sidney A. Burrell, and Robert A. Kann. 1957. Political Community and the North Atlantic Area: International Organization in the Light of Historical Experience. Princeton, NJ: Princeton University Press. Diehl, Paul F. 1992. “What Are They Fighting for?: The Importance of Issues in International Conflict Research.” Journal of Peace Research 29(3):333-344. Diehl, Paul F., and Gary Goertz. 2000. War and Peace in International Rivalry. Ann Arbor, MI: University of Michigan Press. 187 Dixon, William J. 1993. “Democracy and the Management of International Conflict.” Journal of Conflict Resolution 37(1):42-68. Dixon, William J. 1994. “Democracy and the Peaceful Settlement of International Conflict.” American Political Science Review 88(1):14-32. Doucouliagos, Hristos, and Mehmet Ali Ulubaşoğlu. 2008. “Democracy and Economic Growth: A Meta-Analysis.” American Journal of Political Science 52(1):61-83. Downes, Alexander B. 2009. “How Smart and Tough are Democracies? Reassessing Theories of Democratic Victory in War.” International Security 33(4):9-51. Doyle, Michael. 1983. “Kant, Liberal Legacies, and Foreign Affairs.” Philosophy and Public Affairs 12(3):205-235. Doyle, Michael. 1986. “Liberalism and World Politics.” American Political Science Review 80(4):1151-69. Eckstein, Harry. 1975. “Case Study and Theory in Political Science.” in The Handbook of Political Science, eds. Fred I. Greenstein and Nelson W. Polsby. Reading, MA: AddisonWesley, 79-138. Edwards, Sebastian. 1993. “Openness, Trade Liberalization, and Growth in Developing Countries.” Journal of Economic Literature 31(3):1358-93. Edwards, Sebastian. 1998. “Openness, Productivity and Growth: What Do We Really Know?” Economic Journal 108(March):383- 98. Enterline, Andrew. 1996. “Driving while Democratizing.” International Security 20(4):183-196. Enterline, Andrew. 1998. “Regime Changes, Neighborhoods, and Interstate Conflict, 18161992.” Journal of Conflict Resolution 42(6):804-829. Farber, Henry S., and Joanne, Gowa. 1995. “Polities and Peace.” International Security 20(2):123-146. Farber, Henry S., and Joanne Gowa. 1997. “Common Interests or Common Polities? Reinterpreting the Democratic Peace.” Journal of Politics 59(2):393-417. Fearon, James D. 1994. “Domestic Political Audiences and the Escalation of International Disputes.” American Political Science Review 88(3):577-92. Fearon, James D. 1995. 49(3):379-414. “Rationalist Explanations for War.” 188 International Organization Fearon, James D., and David D. Laitin. 2003. “Ethnicity, Insurgency, and Civil War.” American Political Science Review 97(1):75-90. Fein, Helen. 1995. “More Murder in the Middle: Life-Integrity Violations and Democracy in the World, 1987.” Human Rights Quarterly 17(1):170-191. Gallup, John L., Andrew D. Mellinger, and Jeffrey D. Sachs. 2001. “Geography Datasets: General Measures of Geography: Physical Geography and Population.” http://www.cid.harvard.edu/ciddata/geographydata.htm. Gartzke, Erik. 1998. “Kant We All Just Get Along?: Motive, Opportunity, and the Origins of the Democratic Peace.” American Journal of Political Science 42(1):1-27. Garzke Erik. 2000. “Preferences and the Democratic Peace.” International Studies Quarterly 44(2):191-210. Gartzke, Erik. 2007. “The Capitalist Peace.” American Journal of Political Science 51(1):166191. Gartzke, Erik, Quan Li, and Charles Boehmer. 2001. “Investing in the Peace: Economic Interdependence and International Conflict.” International Organization 55(2):391-438. Gaubatz, Kurt Taylor. 1996. “Democratic States and Commitment in International Relations.” International Organization 50(1):109-139. Geddes, Barbara. 1990. “How the Cases You Choose Affect the Answers You Get: Selection Bias in Comparative Politics.” Political Analysis 2(1):131-150. Gelpi, Christopher F., and Joseph Grieco. 2008. “Democracy, Trade and the Nature of the Liberal Peace.” Journal of Peace Research 45(1):17-36. Gelpi, Christopher F. and Michael Griesdorf. 2001. “Winners or Losers? Democracies in International Crisis, 1918-1994.” American Political Science Review 95(3):633-647. Gelpi, Christopher, Peter Feaver, and Jason Reifler. 2005/06. “Success Matters: Casualty Sensitivity and the War in Iraq.” International Security 30(3):7-46. Gelpi, Christopher, Peter Feaver, and Jason Reifler. 2009. Paying the Human Costs of War. Princeton, NJ: Princeton University Press. Gelpi, Christopher, Jason Reifler, and Peter Feaver. 2007. “Iraq the Vote: Retrospective and Prospective Foreign Policy Judgments on Candidate Choice and Casualty Tolerance.” Political Behavior 29(2):151-174. 189 Gibler, Douglas M. 1997. “Control the Issues, Control the Conflict: The Effects of Alliances That Settle Territorial Issues on Interstate Rivalries.” International Interactions 22(4):341368. Gibler, Douglas M. 2007. “Bordering on Peace: Democracy, Territorial Issues, and Conflict.” International Studies Quarterly 51(3): 509-532. Gibler, Douglas M. 2010. “Outside-In: The Effects of Territorial Threat on State Centralization.” Journal of Conflict Resolution 54(4):519-542. Gibler, Douglas M., and Meredith Reid Sarkees. 2004. “Measuring Alliances: The Correlates of War Formal Interstate Alliance Dataset, 1816-2000.” Journal of Peace Research 41(2):21122. Gleditsch, Kristian Skrede. 2002a. All International Politics Is Local: The Diffusion of Conflict, Integration, and Democratization. Ann Arbor: University of Michigan Press. Gleditsch, Kristian Skrede. 2002b. “Expanded Trade and GDP Data.” Journal of Conflict Resolution 46(5):712-24. Goemans, Hein. 2000. War and Punishment: The Causes of War Termination & the First World War. Princeton, NJ: Princeton University Press. Gowa, Joanne. 1995. “Democratic States and International Disputes.” International Organization 49(3):511-522. Gowa, Joanne. 1999. Ballots and Bullets: the Elusive Democratic Peace. Princeton, NJ: Princeton University Press. Gowa, Joanne. 2009. “The Democratic Peace after the Cold War.” Typescript. Princeton University. Gujarati, Damodar N. 2004. Basic Econometrics, 4th ed. Boston, MS: McGraw-Hill. Harrison, Ewan. 2010. “The Democratic Peace Research Program and System-Level Analysis.” Journal of Peace Research 47(2):155-165. Hegre, Håvard; John Oneal & Bruce M. Russett. 2010. “Trade Does Promote Peace: New Simultaneous Estimates of the Reciprocal Effects of Trade and Conflict.” Journal of Peace Research 47(6):763-774. Hensel, Paul. 1999. “Charting a Course to Conflict: Territorial Issues and Interstate Conflict, 1816-1992.” In A Road Map to War: Territorial Dimensions of International Conflict, ed. Paul F. Diehl. Nashville, TN: Vanderbilt University Press, 115-146. 190 Hensel, Paul. 2001. “Contentious Issues and World Politics: The Management of Territorial Claims in the Americas, 1816-1992.” International Studies Quarterly 45(1):81-109. Hewitt, Joseph., and Garry Young. 2001. “Assessing the Statistical Rarity of Wars between Democracies.” International Interactions 27(3):327-351. Hutchison, Marc, and Douglas M. Gibler. 2007. “Political Tolerance and Territorial Threat: A Cross-National Study.” Journal of Politics 69(1):128-142. Huth, Paul. K. 1996. Standing Your Ground: Territorial Disputes and International Conflict. Ann Arbor: University of Michigan Press. Huth, Paul K. 2000. “Territory: Why Are Territorial Disputes between States a Central Cause of International Conflict?” In What Do We Know about War?, ed. John A. Vasquez. Boston, MD: Rowman & Littlefield, 85-110. Huth, Paul K., and Todd L. Allee. 2002. The Democratic Peace and Territorial Conflict in the Twentieth Century. Cambridge, UK: Cambridge University Press. Jaggers, Keith and Ted Robert Gurr. 1995. “Tracking Democracy's Third Wave with the Polity III Data.” Journal of Peace Research 32(4):469-82. James, Patrick, and Glenn E. Mitchell II. 1995. “Targets of Covert Pressure: The Hidden Victims of the Democratic Peace.” International Interactions 21(1):85-107 James, Patrick, Johann Park, and Seung-Whan Choi. 2006. “Democracy and Conflict Management: Territorial Claims in the Western Hemisphere Revisited.” International Studies Quarterly 50(4):803-817. James, Patrick, Eric Solberg, and Murray Wolfson. 1999. “An Identified Systemic Model of the Democracy-Peace Nexus.” Defence and Peace Economics 10(1):1-37. Jensen, Nathan. 2006. Nation-State and the Multinational Corporation: A Political Economy of Foreign Direct Investment. Princeton, N.J.: Princeton University Press. Jones, Daniel M., Stuart A. Bremer, and J. David Singer. 1996. “Militarized Interstate Disputes, 1816-1992: Rationale, Coding Rules, and Empirical Patterns.” Conflict Management and Peace Science 15(2):163-213. Kacowicz, Arie. M. 1995. “Explaining Zones of Peace: Democracies as Satisfied Powers?” Journal of Peace Research 32(3):265-276. Kant, Immanuel. 1957 [1795]. Perpetual Peace. ed. Lewis White Beck. New York: The BobbsMerrill Co. 191 Kant, Immanuel. 1970 [1784]. “Idea for a Universal History with a Cosmopolitan Purpose.” In Kant’s Political Writings, ed. Hans Reiss and trans. H. B. Nisbet. Cambridge, UK: Cambridge University Press, 93-130. Kaufmann, Chaim. 2004. “Threat Inflation and the Failure of the Marketplace of Ideas: The Selling of the Iraq War.” International Security 29(1):5-48. Keck, Margaret, & Kathryn Sikkink. 1998. Activists beyond Borders: Advocacy Networks in International Politics. Ithaca, NY: Cornell University Press. Keohane, Robert O. 1984. After Hegemony: Cooperation and Discord in the World Political Economy. Princeton, NJ: Princeton University Press. King, Gary, Robert. O. Keohane, and Sidney Verba. 1994. Designing Social Inquiry: Scientific Inference in Qualitative Research. Princeton, NJ: University Press. King, Gary, Michael Tomz, and Jason Wittenberg. 2000. “Making the Most of Statistical Analyses: Improving Interpretation and Presentation.” American Journal of Political Science 44(2):347-361. Kinsella, David, and Bruce Russett. 2002. “Conflict Emergence and Escalation in Interactive International Dyads.” Journal of Politics 64(4): 1045-1068. Krasner, Stephen D. 1983. “Structural Causes and Regime Consequence: Regimes as Intervening Variables.” In International Regimes, ed. Stephen D. Krasner. Ithaca, NY: Cornell University Press, 1-22. Krueger, Anne O. 1997. “Trade Policy and Economic Development: How We Learn.” American Economic Review 87(1):1-22 Kugler, Jacek and Douglas Lemke. 1996. Parity and War: Evaluations and Extensions of the War Ledger. Ann Arbor: University of Michigan Press. Lake, David A. 1992. “Powerful Pacifists: Democratic States and War.” American Political Science Review 86(1):24-37. Lawrence, Robert Z., and David E. Weinstein. 1999. “Trade and Growth: Import-Led or ExportLed? Evidence from Japan and Korea.” National Bureau of Economic Research Working Paper 7264, available on the World-Wide Web at www.nber.org/papers/w7264. Layne, Christopher. 1994. “Kant or Cant: The Myth of the Democratic Peace.” International Security 19(2):5-49. Leeds, Brett Ashley. 1999. “Domestic Political Institutions, Credible Commitments, and International Cooperation.” American Journal of Political Science 43(4):979-1002. 192 Legro, Jeffrey W., and Andrew Moravcsik. 1999. “Is Anybody Still a Realist?” International Security 2(2):5-55. Lektzian, David, Brandon C. Prins and Mark Souva. 2010. “Territory, River, and Maritime Claims in the Western Hemisphere: Regime Type, Rivalry, and MIDs from 1901 to 2000.” International Studies Quarterly 54(4):1073-1098. Lemke, Douglas, and William Reed. 2001. “The Relevance of Politically Relevant Dyads.” Journal of Conflict Resolution 45(1):126-144. Leventoğlu, Bahar, and Ahmer Tarar. 2005. “Pre-negotiation Public Commitment in Domestic and International Bargaining.” American Political Science Review 99(3):419-433. Levy, Jack. S. 1988. “Domestic Politics and War.” Journal of Interdisciplinary History 18(4):653-673. Levy, Jack S. 2002. “Qualitative Methods in International Relations.” In Millennial Reflections on International Studies, eds. Michael Brecher and Frank P. Harvey. Ann Arbor: University of Michigan Press, 432-54. Li, Quan, and David Sacko. 2002. “The (IR)Relevance of Interstate Militarized Disputes to International Trade.” International Studies Quarterly 46(1):11-44. Lipson, Charles. 2003. Reliable Partners: How Democracies Have Made a Separate Peace. Princeton, NJ: Princeton University Press. Lynn-Jones, Sean M. 1996. “Preface.” In Debating the Democratic Peace, eds. Michael E. Brown, Sean M. Lynn-Jones, and Steven E. Miller. Cambridge, MA: MIT Press, ix-xxxiii. Mansbach, Richard W. and John A. Vasquez. 1981. In Search of Theory: A New Paradigm for Global Politics. New York: Columbia University Press. Mansfield, Edward, and Jack Snyder. 1995. “Democratization and War.” Foreign Affairs 74(3):79-97. Mansfield, Edward, and Jack Snyder. 2005. Electing to Fight: Why Emerging Democracies Go to War. Cambridge, MA: MIT Press. Mansfield, Edward, Helen Milner, and B. Peter Rosendorff. 2002. “Why Democracies Cooperate More: Electoral Control and International Trade Agreements.” International Organization 56(3):477-514. Maoz, Zeev. 1998. “Realist and Cultural Critiques of the Democratic Peace: A Theoretical and Empirical Reassessment.” International Interactions 24(1):3-89. 193 Maoz, Zeev, and Nasrin Abdolali. 1989. “Regime Type and International Conflict: 1816-1976.” Journal of Conflict Resolution 33(1):3-35. Maoz, Zeev and Bruce M. Russett. 1993. “Normative and Structural Causes of the Democratic Peace.” American Political Science Review 87(3):624-638. Marshall, Monty. G., and K. Jaggers. 2004. “Polity IV project: Dataset Users Manual.” College Park: University of Maryland. Marshall, Monty G., Ted Robert Gurr, and Keith Jaggers. 2010. “Political Regime Characteristics and Transitions, 1800-2009: Dataset Users’ Manual.” http://www.systemicpeace.org/inscr/p4manualv2009.pdf. McDonald, Patrick J. 2004. “Peace through Trade or Free Trade?” Journal of Conflict Resolution 48(4):547-572. McGillivray, Fiona, and Alastair Smith. 2000. “Trust and Cooperation through Agent-specific Punishments.” International Organization 54(4):809-824. Mearsheimer, John J. 1990a. “Back to the Future: Instability in Europe after the Cold War.” International Security 15(1):5-56. Mearsheimer, John J. 1990b. “Why We Will Soon Miss the Cold War.” The Atlantic Monthly 266(2):35-50. Mearsheimer, John J. 2001. The Tragedy of Great Power Politics. New York: W.W. Norton & Company, Inc. Menard, Scott. 2002. Applied Logistic Regression Analysis, 2nd ed. Thousand Oaks, California: Sage Publications. Miller, Warren E. 1999. “Temporal Order and Causal Inference.” Political Analysis 8(2):119-42. Mitchell, Sara McLaughlin. 2002. “A Kantian System: Democracy and Third-Party Conflict Resolution.” American Journal of Political Science 46(4):749-759. Mitchell, Sara McLaughlin, and Brandon Prins. 1999. “Beyond Territorial Contiguity: Issues at Stake in Democratic Militarized Disputes.” International Studies Quarterly 43(1):169-183. Mitchell, Sara McLaughlin, Scott Gates, and Håvard Hegre. 1999. “Evolution in DemocracyWar Dynamics.” Journal of Conflict Resolution 43(6):771-792. Mitchell, Sara McLaughlin, Kelly M. Kadera, and Mark J. C. Crescenzi. 2008. “Practicing Democratic Community Norms: Third Party Conflict Management and Successful Settlements.” In International Conflict Mediation: New Approaches and Findings, eds. Jacob Bercovitch and Scott Sigmund Gartner. New York: Routledge, 243-264. 194 Moravcsik, Andrew. 1997. “Taking Preferences Seriously: A Liberal Theory of International Politics.” International Organization 51(4):513-553. Morgan, T. Clifton, and Sally H. Campbell. 1991. “Domestic Structure, Decisional Constraints, and War: So Why Kant Democracies Fight?” Journal of Conflict Resolution 35(2):187-211. Morrow, James. 1999. “How Could Trade Affect Conflict?” Journal of Peace Research 36(4):481-489. Mousseau, Michael. 1998. “Democracy and Compromise in Militarized Interstate Conflicts, 1816-1992.” Journal of Conflict Resolution 42(2):210-230. Mousseau, Michael. 2009. “The Social Market Roots of Democratic Peace.” International Security 33(4):52-86. Mousseau, Michael, and Yuhang Shi. 1999. “A Test for Reverse Causality in the Democratic Peace Relationship.” Journal of Peace Research 36(6):639-663. Murphy, Alexander. 1990. “Historical Justifications for Territorial Claims.” Annals of the Association of American Geographers 80(4):531-548. Narang, Vipin, and Rebecca M. Nelson. 2009. “Who Are These Belligerent Democratizers? Reassessing the Impact of Democratization on War.” International Organization 63(2):357379. North, Douglass C. 1981. Structure and Change in Economic History. New York: Norton. Oatley, Thomas. 2005. International Political Economy: Interests and Institutions in the Global Economy, 2nd ed. New York: Longman. Olson, Mancur 1993. “Dictatorship, Democracy, and Development.” American Political Science Review 87(3):567-576. Oneal, John R. and Bruce Russett. 1999a. “The Kantian Peace: The Pacific Benefits of Democracy, Interdependence, and International Organizations.” World Politics 52(1):1-37. Oneal, John R. and Bruce Russett. 1999b. “Is the Liberal Peace Just an Artifact of Cold War Interests? Assessing Recent Critiques.” International Interactions 25(3):213-241. Oneal, John R. and Bruce Russett. 2005. “Rule of Three, Let It Be? When More is Really Better.” Conflict Management and Peace Science 22(4):293-310. Oneal, John. R., Bruce Russett, and Michael L. Berbaum. 2003. “Causes of peace: Democracy, Interdependence, and International Rrganizations, 1985-1992.” International Studies Quarterly 47(3):371-394. 195 Owen, John M. 1994. “How Liberalism Produces Democratic Peace.” International Security 19(2):87-125. Partell, Peter J., and Glenn Palmer. 1999. “Audience Costs and Interstate Crises: An Empirical Assessment of Fearon's Model of Dispute Outcomes.” International Studies Quarterly 43(2):389-405. Przeworski, Adam, Michael E. Alvarez, Jose Antonio Cheibub, and Fernando Limongi. 2000. Democracy and Development: Political Institutions and Well-Being in the World, 1950-1990. Cambridge, UK: Cambridge University Press. Putnam, Robert D. 1993. Making Democracy at Work: Civic Traditions in Modern Italy. Princeton, NJ: Princeton University Press. Rasler, Karen, and Thompson, William R. 1989. War and State Making: The Shaping of the Global Powers. Boston: Unwin Hyman. Rasler, Karen, and William R. Thompson. 2001. “Rivalries and the Democratic Peace in the Major Power Subsystem.” Journal of Peace Research 38(6):659-683. Rasler, Karen and William R. Thompson. 2004. “The Democratic Peace and the Sequential, Reciprocal, Causal Arrow Hypothesis.” Comparative Political Studies 37(8):879-908. Ray, James Lee. 2005. “Constructing Multivariate Analyses (of Dangerous Dyads).” Conflict Management and Peace Science 22(4):277–92. Raymond, Gregory A. 1994. “Democracies, Disputes and Third Party Intermediaries.” Journal of Conflict Resolution 38(1):24-42. Reed, William. 2000. “A Unified Statistical Model of Conflict Onset and Escalation.” American Journal of Political Science 44(1):84-93. Reed, William, and Daina Chiba. 2010. “Decomposing the Relationship between Contiguity and Militarized Conflict.” American Journal of Political Science 54(1): 61-73. Regan, Patrick, and Errol A. Henderson. 2002. “Democracy, Threats, and Political Repression in Developing Countries: Are Democracies Internally Less Violent?” The World Quarterly 23(1):119-136. Reiter, Dan. 2001. “Does Peace Nature Democracy?” Journal of Politics 63(3):935-948. Reiter, Dan, and Allan C. Stam. 2002. Democracies at War. Princeton, NJ: Princeton University Press. Reiter, Dan, and Allan C. Stam. 1998a. “Democracy, War Initiation, and Victory.” American Political Science Review 92(2):377-389. 196 Reiter, Dan, and Allan C. Stam. 1998b. “Democracy and Battlefield Military Effectiveness.” Journal of Conflict Resolution 42(3):259-277. Reiter, Dan, and Allan C. Stam. 2003. “Understanding Victory: Why Political Institutions Matter.” International Security 28(1):168-179. Reiter, Dan, and Allan C. Stam. 2009. “Correspondence: Another Skirmish in the Battle over Democracies and War.” International Security 34(2):194-200. Reiter, Dan, and Erik Tillman. 2002. “Public, Legislative, and Executive Constraints on the Democratic Initiation of Conflict.” Journal of Politics 64(3):810-826. Rosato, Sebastian. 2003. “The Flawed Logic of Democratic Peace Theory.” American Political Science Review 97(4):585-602. Rosenau, James N. 1961. Public Opinion and Foreign Policy. New York: Random House. Rummel, Rudolph. 1983. “Libertarianism and International Violence.” Journal of Conflict Resolution 27(1):27–72. Russett, Bruce M. 1993. Grasping the Democratic Peace: Principles for a Post-Cold War World. Princeton, NJ: Princeton University Press. Russett, Bruce M. 1995. “The Democratic Peace: And Yet It Moves.” International Security 19(4):164-75. Russett, Bruce M. and John R. Oneal. 2001 Triangulating Peace: Democracy, Interdependence, and International Organizations. New York: Norton. Sartori, Anne E. 2003. “An Estimator for Some Binary-Outcome Selection Models without Exclusion Restrictions.” Political Analysis 11(2):111-138. Schuessler, John M. 2010. “The Deception Dividend: FDR’s Undeclared War.” International Security 34(4):133-165. Schultz, Kenneth 1998. “Domestic Opposition and Signaling in International Crises.” American Political Science Review 92(4):829-844. Schultz, Kenneth. 1999. “Do Democratic Institutions Constrain or inform?: Contrasting Two Institutional Perspectives on Democracy and War.” International Organization 53(2):233266. Schultz, Kenneth. 2001a. Resolution 45(1):32-60. “Looking for 197 Audience Costs.” Journal of Conflict Schultz, Kenneth. 2001b. Democracy and Coercive Diplomacy. Cambridge, UK: Cambridge University Press. Schultz, Kenneth and Barry R. Weingast 2003. “The Democratic Advantage: Institutional Foundations of Financial Power in International Competition.” International Organization 57(1):3-42. Senese, Paul. D. 1997. “Between Dispute and War: The Effect of Joint Democracy on Interstate Conflict Escalation.” The Journal of Politics 59(1):1-27. Senese, Paul. D., and John. A. Vasquez. 2003. “A Unified Explanation of Territorial Conflict: Testing the Impact of Sampling Bias, 1919-1992.” International Studies Quarterly 47(2):275-298. Senese, Paul D., and John A. Vasquez. 2005. “Assessing the Steps to War.” British Journal of Political Science 35(4):607-633. Senese, Paul D. and John A. Vasquez. 2008. The Steps to War: An Empirical Study. Princeton, NJ: Princeton University Press. Siegel, Eric. 1997. “I Know that You Know and You Know that I Know: An Information Theory of the Democratic Peace.” Typescript. Signorino, Curtis S. and Jeffrey M. Ritter. 1999. “Tau-b or Not Tau-b: Measuring the Similarity of Foreign Policy Positions.” International Studies Quarterly 43(1):115-144. Simmons, Beth. 1999. “See You in “Court?: The Appeal to Quasi-Judicial Legal Processes in the Settlement of Territorial Disputes.” In A Road Map to War: Territorial Dimensions of International Conflict, ed. Paul F. Diehl. Nashville, TN: Vanderbilt University Press, 205237. Simmons, Beth. 2005. “Rules over Real Estate: Trade, Territorial Conflict and International Borders as Institutions.” Journal of Conflict Resolution 49(6):823-848. Singer, J. David, Stuart Bremer, and John Stuckey. 1972. “Capability Distribution, Uncertainty, and Major Power War, 1820-1965.” In Peace, War and Numbers, ed. Bruce M. Russett. Beverly Hills, CA: Sage Publications, 19-48. Siverson, Randolph. 1995. “Democracies and War Participation: In Defense of the Institutional Constraints Argument.” European Journal of International Relations 1(4):481-89. Slantchev, Branslav L., Anna Alexandrova, and Erik Gartzke. 2005. “Probabilistic Causality, Selection Bias, and the Logic of the Democratic Peace.” American Political Science Review 99(3):459-462. 198 Small, Melvin, and J. David Singer. 1976. “The War Proneness of Democratic Regimes.” Jerusalem Journal of International Relations 1(1):50-69. Spiro, David. 1994. “The Insignificance of the Liberal Peace.” International Security 19(2):5086. Stam, Allan C. 1996. Win, Lose, or Draw: Domestic Politics and the Crucible of War. Ann Arbor: University of Michigan Press. STATA, Release 9, 2005. “User’s Manual.” College Station, Texas: Stata Press. Stinnett, Douglas M., Jaroslav Tir, Philip Schafer, Paul F. Diehl, and Charles Gochman. 2002. “The Correlates of War Project Direct Contiguity Data, Version 3.” Conflict Management and Peace Science 19(2):58-66. Thompson, William R. 1996. “Democracy and Peace: Putting the Cart before the Horse?” International Organization 50(1):141-174. Thompson, William R. 2001. “Identifying Rivals and Rivalries in World Politics.” International Studies Quarterly 45(4):557-586. Tilly, Charles. 1975. “Reflections on the History of European State Making.” In The Formation of National States in Western Europe, ed. Charles Tilly. Princeton, NJ: Princeton University Press. Walsh, James I., and James. A. Piazza. 2010. “Why Respecting Physical Integrity Rights Reduces Terrorism.” Comparative Political Studies 43(5):551-577. Waltz, Kenneth N. 1970. “The Myth of National Interdependence.” In The International Corporation, ed. Charles P. Kindleberger. Cambridge, MA: MIT Press, 205-223. Waltz, Kenneth N. 1979. Theory of International Politics. New York: McGraw-Hill. Ward, Michael D., and Kristian S. Gleditsch. 1998. “Democratizing for Peace.” American Political Science Review 92(1):51-61. Wendt, Alexander. 1992. “Anarchy Is What States Make of It: The Social Construction of Power Politics.” International Organization 46(2):391-425. Werner, Suzanne. 2000. “The Effects of Political Similarity on the Onset of Militarized Disputes, 1816-1985.” Political Science Quarterly 53(2):343-374. Valentino, Benjamin A., Paul K. Huth, and Sarah E. Croco. 2010. “Bear Any Burden? How Democracies Minimize the Costs of War.” Journal of Politics 72(2):528-544. Vasquez, John A. 1993. The War Puzzle. Cambridge, UK: Cambridge University Press. 199 Vasquez, John A. 1995. “Why Do Neighbors Fight? Proximity, Interaction, or Territoriality.” Journal of Peace Research 32(3):277-293. Vasquez, John A. 1996. “Distinguishing Rivals That Go to War from Those That Do Not: A Quantitative Comparative Case Study of the Two Paths to War.” International Studies Quarterly 40(4):531-558. Vasquez, John A. 2001. “Mapping the Probability of War and Analyzing the Possibility of Peace: The Role of Territorial Disputes.” Conflict Management and Peace Science 18(2):145-173. Vasquez, John A. 2004. “The Probability of War, 1816-1992.” International Studies Quarterly 48(1):1-27. Vasquez, John A. 2009. The War Puzzle Revisited. Cambridge, UK: Cambridge University Press. Vasquez, John A., and Douglas M. Gibler. 2001. “The Steps to War in Asia, 1931-1945.” Security Studies 10(3):1-45. Vasquez, John A. and Brandon Valeriano. 2008. “Territory as a Source of Conflict and a Road to Peace” in Sage Handbook on Conflict Resolution, eds. Jacob Bercovitch, Victor Kremenyuk, and I. William Zartman. Los Angeles: Sage, 193-209. Weesie, Jeroen. 1999. “Seemingly Unrrelated Estimation and the Cluster-adjusted Sandwich Estimator.” STATA Technical Bulletin (November):34-47. 200