Full Text

1 Introduction Does the logic of the democratic peace extend to East Asia? Can it help explain the relative peace in the region since 1979? The democratic peace proposition is most commonly understood as an expectation that democracies are very unlikely to go to war with each other. While democratic peace theories are usually stated in universal terms, empirical tests in the Asian region have not been supportive, and authors examining the low levels of regional conflict in recent decades have not focused on increased degrees of democracy as an important factor. In this article, I provide a theoretical discussion of the institutional foundations of democratic peace that distinguishes the potential effects of different aspects of political regimes. I argue that empirical tests of this more fine-grained set of institutional expectations should yield more robust evidence of democratic peace in the region. Examination of institutional changes over time should also give insight into their relevance for the post-1979 relative East Asian peace. My focus is on institutions facilitating political competition, a relatively under-examined aspect of regime type in the context of interstate conflict. A challenge for the democratic peace proposition is to demonstrate that it applies universally, across all regions of the globe (Goldsmith, 2006; Henderson, 2009). Previous studies have not found support for democratic peace in Asia (Kivimaki, 2001; Goldsmith, 2006, 2007; Johnston, 2012; Tang, 2012).1 In this article, I present evidence for conflict initiation suggesting that focusing on political competition yields support that is robust among East Asian dyads, as well as globally, in a way that cannot be claimed for measures of other political institutions. I also ask whether a focus on political competition can help explain the relative peace experienced in the East Asian region after 1979 (Solingen, 2007; Svensson, 2011; Svensson and Lindgren, 2011; Tønnesson, 2009). Because the degree of political competition within East Asian regimes increased substantially during and after the 1980s, this variable is a plausible, systematic, and 1 Some studies have argued that democratization does not cause interstate conflict in the region (Acharya 2010;Lind 2011) but have not presented systematic evidence of a pacifying effect. 60 Benjamin E. Goldsmith at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from theoretically grounded contender for at least partially explaining the relative lack of interstate violence in East Asia since the 1979 Sino-Vietnamese war. My theoretical focus on political competition is different from other theories connecting institutional characteristics of political systems and their international conflict behavior. My empirical approach involves testing competing institutional explanations directly against each other. While prominent theories of democratic peace do focus on specific institutional characteristics of democratic regimes (Lake, 1992; Bueno de Mesquita et al., 1999), empirical analysis has generally not disaggregated regime type into its several institutional components in order to examine competing claims simultaneously (exceptions include Reiter and Tillman, 2002; Huth and Allee, 2002b; Morrowet al., 2008). Consequently, my empirical tests break down the measure of regime type into three institutional components. I focus on the importance of institutions facilitating political competition, which lead decision-makers to have expectations of either vulnerability to opposition criticism when competition is strong or dominance over political communication when competition is weak or nonexistent. These expectations, I argue, play a central role in the connection between domestic political systems and states’ international conflict behavior. Greater political vulnerability causes leaders to attempt to forestall opposition criticism, resulting in more circumspect behavior in confrontations with other states. Because democratic targets present additional diffi- culties for the justification of the use of deadly force, and for its likely success, states with higher levels of political competition will be especially reluctant to initiate militarized conflicts with them. In statistical tests, I examine whether political competition reduces the likelihood of East Asian states’ conflict initiation with more democratic regimes in the region. I show that political competition is indeed signifi- cantly related to a lower likelihood of conflict initiation against democratic regimes. I focus on the domestic institutions of potential conflict initiators in order to understand the institutional factors that condition leaders’ foreign policy choices. However, I assume that political elites and masses, although of course being affected by the particular configuration of institutions within their states, nevertheless perceive other states in more general terms, based on types of regimes rather than their specific internal mechanisms. This leads me to focus on the overall regime type of the Domestic institutions and conflict in East Asia 61 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from potential target states, because their overall degree of authoritarianism or democracy is more important for the potential initiating leaders’ choices. My argument hinges on the role of a viable opposition force in politics, which is the essence of genuine political competition. I argue that institutions of political competition drive leaders to think strategically about the political implications of foreign policy choices. They try to anticipate, and forestall, potentially damaging criticism from the opposition. Political competition compels leaders to seek policies that can be defended rhetorically against potential criticism aimed at weakening the sitting government, and to avoid policies that will lead to political vulnerability. States with institutions allowing high levels of political competition are, therefore, less likely to initiate disputes with democracies because it is harder to make a defensible moral or practical case for such disputes against the arguments of a viable opposition party or group. They may not be less likely to initiate conflict with non-democratic states, however, because these will be more readily justified in normative terms, and, other things equal, the chances of prevailing in conflict may appear higher as well. The paper proceeds as follows. The next section briefly discusses other institutional democratic peace theories in the existing literature. This is followed by an elaboration of my argument regarding the role of political competition in foreign conflict behavior. I then describe and present results of empirical tests for conflict initiation comparing the role of political competition with other aspects of regime type for global data (excluding East Asian dyads), and for East Asian dyads specifically, and I show how political competition might be an important part of the explanation for the lack of interstate war in East Asia since 1979. In the conclusion, I discuss implications for understanding the connections between regime type and interstate conflict in East Asia and in general. 2 Political systems and militarized interstate conflict 2.1 Some existing institutional approaches While the democratic peace literature is vast, a fairly limited number of studies focus on testing competing institutional explanations of why democracies might be very unlikely to fight each other. Some of these studies use aggregate regime-type measures to test specific institutional arguments by implication, or they examine only one institutional feature of regime 62 Benjamin E. Goldsmith at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from type, without adequately controlling for other institutions that may be strongly correlated with it.2 In this study, I disaggregate regime type into three components and include each in my models, providing more robust controls for competing institutional expectations. In this sub-section, I discuss existing theories about institutions of participation and constraint on the executive. In the next sub-section, I discuss institutions of political competition, and the underlying logic of my expectations and hypotheses. Immanuel Kant’s seminal explication of the causal mechanisms of a ‘republican’ peace focused strongly on leaders’ accountability to a consentgiving public who, in choosing war, ‘would necessarily be resolving to bring upon themselves all the horrors of War’ (Kant, [1795] 1917, p. 122). Mass participation in leader selection is a key part of this argument. More recently, the differential costs of war for leaders and subjects have been a central aspect of widely known ‘structural’ arguments of democratic peace (Russett and Oneal, 2001, pp. 53–58). Meaningful mass participation in selecting and removing national leaders also plays a constraining role on conflict initiation according to this approach (perhaps monadically, i.e. regardless of the regime type of the target state). As Russett and Oneal (2001, pp. 273–74) write, ‘in a democracy, those who would bear the costs of war are the ones who decide whether it shall be fought…. The regular occurrence of elections is obviously important in this process. It is the mechanism that forces government to consider the will of the people’. 3 It is therefore important to consider political participation in my models. Another institutional argument focuses on the ability of political executives to extract rents from society, and the degree to which they are 2 See the discussion of these issues in a series of related papers: Bueno de Mesquita et al. (1999);Bueno de Mesquita et al. (2004); Clarke and Stone (2008); andMorrow et al. (2008). 3 Another prominent institutional theory of the connection between regime type and international conflict behavior is selectorate theory, first presented by Bueno de Mesquita et al. (1999). Their focus is on the size of the ‘winning coalition’ – the sub-set of the population necessary to place or maintain a leader in power. Wider winning coalitions provide leaders with incentives both to avoid costly wars and to reliably provide the public good of security. This is clearly related to the concept of political participation. However, Bueno de Mesquita et al. are actually agnostic about whether states with wide winning coalitions will be less likely to initiate conflict with democracies (Bueno de Mesquita et al., 2003, pp. 243–45). Specifically, they state that‘war…between democracies is unlikely, though disputes are not.’ Democracies might initiate disputes with each other as frequently as they do with other states, but those between democracies will be less likely to escalate to full-scale war. Since I do not separately examine the escalation process in this article, I do not focus on controlling for selectorate theory, although I believe my participation indicator is a reasonable proxy for winning coalition size. Domestic institutions and conflict in East Asia 63 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from constrained from doing so. Leaders who are more constrained from extracting rents have less incentive for territorial expansion and will also be less threatening to each other. As Lake (1992, pp. 26–27) writes ‘[i]n practice, there are always positive costs of monitoring state behavior and exercising voice and exit. As a result, all states possess some ability to earn rents, although the ability will be larger in autocracies than in democracies…. To the extent that a state can earn rents, state and societal interests will diverge and the state will be biased toward an expansionary foreign policy’. This occurs for several reasons including ‘most important, the larger the state’s rent-seeking ability, the higher the total revenue earned by the state. The more revenue (ceteris paribus), the larger the optimal size of the political unit’. Lake (1992, pp. 29–30) continues: ‘democracy, in this approach, constrains the ability of the state to extract monopoly rents at society’s expense’, with the result that ‘democracies are less likely to fight each other, for only in this area is the absence of an imperialist bias manifest’. Schultz (1999; 2001) presents a theory involving constraints, which has monadic implications for democracies’ conflict behavior. Because of the greater transparency of democratic regimes, they will be more credible in interstate crisis bargaining. He ties this transparency to both political competition and constraints on leaders. I take both Lake’s and Schultz’s approaches to focus on institutions that constrain executive decisionmakers. This might include judicial review by the courts of the legality of executive decisions, a parliament with budgetary powers, or parliamentary oversight committees that scrutinize and investigate executive action. Thus, it is important to account for constraints on executives in my models. There may be a dyadic or monadic effect. However, I argue that these institutional arguments overlook an important aspect of political systems: the strategic implications of internal political competition. Competition plays a role in Schultz’s monadic theory of democratic coercive diplomacy. However, (i) the implications are monadic rather than dyadic, and (ii) it is the information that external adversaries are able to receive due to the necessary transparency of genuinely democratic political processes, which is key for his approach. For my approach, it is not external transparency, but strategic political calculations by the incumbent that are of interest and have dyadic implications. I also point out that empirically there is considerable, politically meaningful, variation across institutional components of regime type in East 64 Benjamin E. Goldsmith at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Asia (and globally). This makes it possible to analyze the association with conflict of one institutional component, while controlling for the others. For example, Singapore since 1965 has had wide participation, a moderately constrained executive, but severely stunted competition, whereas Cambodia under King Sihanouk (mid-1950s to late 1960s) had a wellconstrained executive but quite low participation and competition; postindependence Burma (now Myanmar; late 1940s to early 1960s) and the Philippines with a new constitution during the Aquino presidency (late 1980s to early 1990s) had high executive constraints, only moderate levels of competition, and lower levels of participation; and South Korea during Park’s Third Republic (mid-1960s to early 1970s) and Thailand (mid-1970s to 1990, intermittently) had high competition, but stunted participation and executive constraints.4 Of course, this should not obscure the fact that these indicators are in general correlated – I discuss checks for problems due to multicollinearity later in the article. In the next subsection, I turn to an elaboration of my argument for the centrality of institutions of political competition. 2.2 Institutions of political competition Competition is a core aspect of democracy and has often been the focus of democratic theorists: ‘The democratic method is that institutional arrangement for arriving at political decisions in which individuals acquire power to decide by means of competitive struggle for the people’s vote’ (Schumpter, 1950, p. 269). Political competition, I suggest, should be related to more cautious decisions at the initiation stage of interstate con- flict. This is so because competition implies a viable opposition, and the opposition is more likely to point out the mistakes of the decision-makers in power and, crucially, to call the attention of voters or other key power brokers, retrospectively, to apparently poor government decisions. Under effective institutions of political competition, foreign policy decisionmakers will fear policy failure because of the existence of a viable 4 For these examples, ‘high’ indicates values of 16 or 17 on my 20 point scale, whereas‘moderate’ indicates values 1–4 points lower, and ‘low’ indicates values 5 or more points lower. More generally, for East Asia during the period of my study, when Competition is at least 17, the minimum for Participation is 12.5 and the minimum for Constraints is 11.7. When Participation is 17 or higher, the minimum for Competition is 5.3 and the minimum for Constraints is 5. When Constraints are 17 or higher, the minimum for Competition is 16 and the minimum for Participation is 12.5. Domestic institutions and conflict in East Asia 65 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from opposition advocating alternative policies (Achen and Bartels, 2004; Colaresi, 2004). Accordingly, leaders in competitive political environments seek to make decisions that are defensible at the initiation stage of conflict, and after the fact, on both practical and moral legitimacy grounds, which are common and effective bases for criticizing foreign policy (Jentleson, 1992; Gelpi et al., 2007). Leaders will face incentives to avoid criticism that their choices are not in the national interest, for example, in terms of values, security, costs, and the chance of success. Political competition therefore makes leaders risk averse, which should help states avoid the costs of war, and conversely encourage them only to be willing to initiate conflict when the chances of victory (and thus of the potential target’s choice for capitulation rather than escalation) are relatively high, and when the issue at stake itself will gain and sustain wide support among the population (Jentleson, 1992). Conversely, I also suggest that very low political competition will allow foreign policy decision-makers to engage in conflict behaviors such as bluffing and probing, because the consequences of backing down from belligerent threats made during the initiation of a militarized dispute will be relatively mild. Without a critical opposition, such contradictory or risky behavior might simply be downplayed or unremarked by national leaders. News media, even if relatively free, tend to take cues from elite opinion leaders. With no strong opposition voices, there will be few competing opinion leaders to counter government frames (Chong and Druckman, 2007; Entman, 2004). Questions will not be asked in public debate about the wisdom of such behavior, for example, regarding costs in damaged trade relations or international reputation. There will be minimal criticism of the moral justification for conflict. Facing a potential adversary of uncertain capability or will, foreign policy decision-makers with little political competition at home will have material and informational incentives to ‘test the waters’ with an initial belligerent act. The target state may concede some or all issues at stake, or it may respond with a convincing show of force itself. In either case, decision-makers will have gained greater understanding of the potential target, while suffering little in terms of domestic interests. This means that states with little competition should be more likely to initiate interstate disputes, whether or not they are resolved to escalate them further to war. 66 Benjamin E. Goldsmith at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from 2.3 Dyadic expectations: initiators and targets My expectations about higher and lower political competition should not apply to target states of all regime types in the same way, because regimes vary in their perceived legitimacy, in their expected military effectiveness, and in their ability to negotiate. Polities with high levels of political competition – seeking to forestall opposition criticism – will be less likely to target democracies when initiating disputes, because the moral case for militarized conflict would be harder to make, given that democracies are perceived as more legitimate in international norms of governance. Aspects of this claimed by theorists are that it is‘very hard for democratic leaders to dehumanize people living in another democracy’ in order to justify war and that dictatorial analogies such as invoked by George H.W. Bush comparing Saddam Hussein to Hitler ‘would be unimaginable between the leaders of two democracies’ (Russett and Oneal, 2001, p. 65). There is also considerable evidence that democracies are stronger adversaries during war in terms of the capabilities they can devote to the conflict (Lake, 1992) and of their forces’ ability to fight effectively on the battlefield (Reiter and Stam, 2002). Thus, risk-averse leaders in a politically competitive domestic environment will avoid them as targets, other things equal. Leaders facing serious domestic political competition will also be reluctant to initiate a dispute if the opposition is likely to make a strong case that the use of force is unnecessary to reach an acceptable agreement with the target state. Since democracies are known to be more able to reach negotiated settlements in many instances (Dixon and Senese, 2002; Huth and Allee, 2002b; Regan and Leng, 2003; although perhaps not over territorial issues, Miller and Gibler, 2011), this will also mean that highcompetition regimes will be less likely to initiate conflict with democracies. However, polities with high competition may more frequently target authoritarian regimes, because such states violate political and civil rights, and their leaders hold power without the electoral consent of the population, making (at least) low-level conflict more justifiable against any potential opposition criticism on moral grounds, while inaction might be criticized. US leaders, for example, were often criticized by the party in opposition as being ‘soft on communism’ during the cold war in their relations with the Soviet Union and China. South Korean criticism of Kim Dae Jung’s‘Sunshine Policy’toward the North has often been framed in a similar way (Choi, 2010). In addition, because authoritarian targets are Domestic institutions and conflict in East Asia 67 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from not transparent and are potentially less able to mobilize resources or motivate personnel for war, successful conflict outcomes may seem more likely against them. There are, therefore, several reasons why regimes with higher levels of political competition will be less likely than regimes with low competition to initiate conflict against democracies, leading to dyadic expectations. These arguments lead to my first hypothesis: H1: The chance of a state initiating militarized interstate conflict will decrease as its level of political competition and the adversary’s level of democracy increase. I further expect that this effect is geographically universal, rather than limited to any region. As noted, East Asia is a region that has shown little evidence of a ‘democratic peace’ effect. If my expectations that political competition is a driving force behind the observed democratic peace phenomenon are correct, then my specification of this should lead to more robust results, even in regions where evidence of a democratic peace reliant on indicators of overall regime type has not emerged. I therefore propose that political competition will be associated with a lower chance of interstate conflict initiation against more democratic states in East Asia, similar to its effect globally. For both global and regional expectations, it is important to emphasize that these patterns should emerge while controlling for other aspects of regime type. Without such controls, political competition might serve as a proxy for regime type, defeating my intent to test a more specific theory. This expectation of relevance for East Asia leads to my second hypothesis: H2: For East Asian dyads, the chance of a state initiating militarized interstate conflict will decrease as its level of political competition and the adversary’s level of democracy increase. 3 Research design I conduct statistical tests of the hypotheses in annual time-series crosssectional data including measures of political institutions and international conflict initiation. I use ‘directed dyads’, which code for conflict initiators and targets (Reiter and Stam, 2003), allowing me to assess the effects of political competition specifically in the initiating state. I also include all three components of regime type: participation, constraint, and political competition. This provides a more specific test of theoretical 68 Benjamin E. Goldsmith at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from expectations by guarding against the possibility that any one institutional component acts as a proxy for democracy itself in the models. The rarity of conflict occurrence among all dyad-years and other data issues may lead to unstable models or results dependent on particular variable specifications. To address such concerns, I conduct a number of robustness checks. These include: dropping all control variables, using four different indicators for the dependent variable, assessing whether multicollinearity distorts the results, and adding a control for territorial claims. My main focus is on conflict dynamics within East Asia. However, I also use new variable specifications that I claim are valid universally, not just for international relations within East Asia. I therefore conduct two sets of analysis, one for dyads for which both states are in East Asia, and another for all other dyads globally, excluding these East Asian dyads. I define East Asia as Southeast and Northeast Asia. Specifically, it includes: Cambodia, China, Indonesia, Japan, Laos, Malaysia, Mongolia, Myanmar, North Korea, Philippines, Republic of Vietnam, Singapore, South Korea, Taiwan, Thailand, and Vietnam (Socialist Republic of Vietnam).5 The time period of my study is 1948–2006 for most analyses. These particular dates are determined by data availability for specific variables (US military aid data determine the early bound, and joint international organization membership data determine the later bound). But, the post-World War II period I believe is appropriate due to potential concerns about data quality and the number and type of states that could be included in the analysis. For example, prior to World War II, there were relatively few independent states in the region, and some nominally independent ones (Thailand and China) were heavily influenced by colonial powers. Also, variables including the measures of regime type and its components that I use might not reflect the same political dynamics as found in more modern states and might not be based on the same quality or quantity of information about each society in the region. I use all dyads in the international system or East Asian region, yielding 1,229,664 observations globally excluding East Asia, 0.04% of which saw crisis initiation by State A against State B, and 12,346 observations in East Asia, 0.35% of which experienced crisis initiation by State A against State B. 5 Brunei is not included because it is categorized as a microstate with a population of <500,000 and therefore is not included in several of the datasets I use, including Polity IV, which provides the political institution indicators. Domestic institutions and conflict in East Asia 69 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from 3.1 Measures of political institutions My key independent variables are indicators of regime type and its institutional components. For overall regime-type measures, I use the Polity IV dataset’s‘polity2’ variable (Marshall et al., 2010), rescaled to range from 0 to 20 for each state in a dyad (State A, potential initiator, and State B, potential target). This variable usually ranges from −10 (fully authoritarian) through +10 (fully democratic), so rescaling is accomplished simply by adding 10 to the indicator for each state-year. I prefer to use non-negative values, on the same scale, for all political institution indicators, in order to simplify interpretation of interaction terms among them. Although the Polity data set does provide three regime-type component indices (Executive Recruitment, Independence of Executive Authority, and Political Competition and Opposition), inspection of the specific indicators used to create them led me to conclude that they do not closely enough reflect the theoretical concepts of Political Participation, Executive Constraints, and Political Competition that I am focusing on. In particular, Polity’s Executive Recruitment index includes an indicator of competitiveness, whereas its Political Competition and Opposition index excludes that, but includes an indicator of the regulation of participation. I therefore use Polity’s six distinct indicators of political institutions to create my three key variables, in the following way. Competition. I identify two of the six regime indicators that specifically measure competition: ‘the competitiveness of executive selection’ (xrcomp) and ‘the extent of government restriction on political competition’ (parcomp). ‘Competitiveness of executive selection’ codes the ‘extent that prevailing modes of advancement give subordinates equal opportunities to become superordinates’. Non-competitive regimes use ‘selection’ through, for example, hereditary lineage as in a monarchy, by the military, or the use of rigged elections. Moderately competitive regimes use ‘dual executives’ or ‘transitional’ arrangements in which real power is divided between an executive chosen through a competitive process and another chosen via methods of ‘selection’ as mentioned earlier (for example, post-1979 Iran with a clerical Supreme Leader as well as an elected president), or in which elections have only limited competition (as in Russia under Vladimir Putin). The most competitive regimes use ‘election’ involving relatively free and fair elections in which voting for the executive is 70 Benjamin E. Goldsmith at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from either ‘popular or by an elected assembly’ (Marshall et al., 2010, pp. 21–22). ‘Extent of government restriction on political competition’ codes ‘extent to which alternative preferences for policy and leadership can be pursued in the political arena’. This is assessed on a scale including ‘repressed’, ‘suppressed’, ‘factional’, ‘transitional’, or ‘competitive’ systems. ‘Repressed’ implies ‘[n]o significant oppositional activity is permitted outside the ranks of the regime and ruling party’ while ‘competitive’ implies ‘relatively stable and enduring, secular political groups which regularly compete for political influence at the national level’ (Marshall et al., 2010, pp. 26–27). I create a Political Competition variable for State A, which is an equally weighted combined measure of these, ranging from 0 to 20. I also created variables for participation and constraints, but I do not describe the indicators used in as much detail, because these are included in the models as controls, rather than being the theoretical focus of my analysis. In order to achieve equal weighting, attention must be given to the scale of each indicator. For example, my Competition variable divides each indicator by its maximum value, then adds the quotients, for a maximum value of 2, and multiplies the sum by 10 to achieve a measure that can range from 0 to 20 (Competition = ((xrcomp/3) + (parcomp/5))*10). Participation. I identify two variables that specifically measure political participation: ‘the openness of executive recruitment’ (xropen) and ‘the degree of organization and institutionalization of participation’ (parreg). My Political Participation variable is an equally weighted combined measure of these two, ranging from 0 to 20. Constraints. I identify two variables that specifically measure Executive Constraints: ‘constraint on chief executive’ (xconst) and ‘the extent of institutionalization of executive transfers’ (xrreg). My executive constraint indicator is rescaled to range from 0 to 20. I use multiplicative interactions of the Regime Type or components of State Awith the Regime Type of State B, to test my hypotheses and control for the other dyadic effects of regime type on conflict (Braumoeller, 2004). Domestic institutions and conflict in East Asia 71 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from 3.2 Measures of interstate conflict initiation I use four different indicators of interstate conflict initiation. For most of my analyses, I use the International Crisis Behavior (ICB; Wilkenfeld and Brecher, 2009) data set to code the initiation of a substantial international crisis by State A (using the ‘triggering entity’ or TRIGENT variable) against State B (‘target’). I believe this is the most appropriate indicator of conflict initiation in the available data sets, because it most directly measures any serious security threat conveyed by one state toward another. It is based on whether a state ‘triggered a foreign policy crisis, i.e. initiated the act which was perceived by a state as creating a threat’. In order to further assure that my results are relevant for initiations of a genuinely threatening nature, I also use an indicator that is restricted to crises coded by ICB as ‘clashes’, which can include ‘minor clash’, ‘serious clash’, or ‘full-scale war’, but exclude those crises that involved ‘no violence’ although they did involve perceived threats. As a further robustness test, I also use the Correlates of War (COW) Militarized Interstate Dispute (MID) data set (Ghosn et al., 2004) to code the initiation of any MID, and the initiation of any MID which incurs at least one battle death. A MID involves any threat, show, or use of force by one state explicitly directed against another. I flag the potential drawbacks of the MID indicators for identifying conflict initiators in the robustness checks section. 3.3 Control variables I include control variables, which might plausibly be related to both internal political institutions, and the initiation of interstate conflict by one state (State A or the initiator) against another (State B or the target). These include the natural log of the trade dependence of State A on dyadic trade between A and B. This is calculated by dividing A’s trade with B by A’s total trade with all trading partners. The variable is called ‘ln (TradeShareA)’. This measure of A’s trade dependence is based on Barbieri (2002), and the data source is the COW trade data (Barbieri et al., 2008). I also include a measure of the trade volume between A and B (i.e. the natural log of the numerator of the dependence indicator, in constant inflation-adjusted dollars (variable name ‘ln(TradeVolume)’), because trade volume might have a positive effect on the chances of 72 Benjamin E. Goldsmith at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from conflict onset (Goldsmith, 2013), while dependence is expected to have a negative effect (Russett and Oneal, 2001). I include indicators of ongoing civil conflict within State A or State B, based on UCDP/PRIO Armed Conflict Dataset indicators (Gleditsch et al., 2002), which code two levels of annual intensity for internal armed conflicts: 1 indicates 25–999 battle deaths, whereas 2 indicates 1000 or more. A value of 0 indicates either conflict below this threshold or no con- flict. The variable names are ‘CivilConflictA’ and ‘CivilConflictB’. It is also important to control for the proximity of States A and B, because both regime type and the likelihood of conflict cluster spatially. I therefore include an indicator of Contiguity that is coded 1 if the states share a land border or are separated by <25 miles of water. I also include a measure of the distance between their capital cities, since noncontiguous states might actually be quite close to each other (e.g. Thailand and Vietnam) whereas some contiguous states might have great distances between their population centers (e.g. China and Laos). These variables are generated from EUGene software (Bennett and Stam, 2000). Democracies tend to be wealthier than non-democracies, and to muster greater military resources during wartime. I therefore include controls for the military capabilities of each state, measured by COW’s index of national capabilities including military spending, military personnel, population, industry, and other factors. The variables, also accessed via EUGene, are named ‘PowerA’ and ‘PowerB’. Especially, when considering intra-regional conflict dynamics, it is important to account for the roles of extra-regional powers. Both extraregional alliance ties and military aid might be related to a country’s regime type, and to its conflict behavior. I therefore include indicators for military aid received from the most powerful state during the period of the study, the United States. The variables ‘ln(USMilAid RatioA)’ and ‘ln (USMilAid RatioB)’ are the ratio of each state’s annual US military aid, in constant dollars, to its power variable, in natural log. Military aid data are from the US government ‘Green Book’ (USAID, 2012). Because states besides the United States have been important players in the region, and regional alliance patterns have had their own distinct patterns, I also include indicators for ‘Regional Alliance Similarity’ and ‘Global Alliance Similarity’, which calculate Signorino and Ritter’s (1999) ‘S’ indicator for alliance portfolio (weighted by states’ regional or global power), as provided in EUGene software. Domestic institutions and conflict in East Asia 73 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from As an indicator of political rather than just security ties, I also include the number of ‘Joint IGO Memberships’ – the number of international governmental organizations in which both states share membership, as coded by the COW IGO project (Pevehouse et al., 2004). In order to correct for temporal dependence, I use a cubic polynomial for the years without conflict initiation by A against B in a dyad (‘Peace Years’, ‘Peace Years2 ’, and ‘Peace Years3 ’), as suggested by Carter and Signorino (2010). These are calculated for each dependent variable. 3.4 Methods of analysis The dependent variables are all dichotomous indicators of conflict, coded 1 for initiation of conflict by State A against State B, 0 otherwise (ongoing conflicts are dropped). A common and appropriate choice for such a limited dependent variable is probit regression. I use probit, with Huber– White standard errors corrected for clustering on the directed dyads. To strengthen causal inference, the model is identified with a one-year lag between the dependent and independent variables, so the initiation of con- flict is measured in yeart (e.g. 1948–2006), but all of the independent variables are measured in yeart-1 (e.g. 1947–2005). Interpreting the coefficients in limited dependent variable models is complicated by the fact that each depends on the particular values taken by other variables in the model. Interaction terms, which are central to my analysis of both hypotheses, are especially difficult to interpret (Brambor et al., 2010). I therefore follow common practice and use simulations of marginal effects to provide an additional assessment of the substantive and statistical significance of my key variables (Brambor et al., 2010; King et al., 2000). These are Monte Carlo simulations using 10,000 sample draws from the multivariate normal distribution and they show the estimated effect of a movement from one standard deviation below the mean to one standard deviation above the mean of State A’s Regime Type, Competition, Participation, or Constraints, as State B’s Regime Type becomes more democratic (moves from 0 to 20). To do this, I hold other variables in the model at their mean values for East Asia during the period, with the exception of contiguity, which I set to ‘1’ to provide a plausible scenario for conflict initiation among otherwise typical dyads. Inclusion of all elements of regime type (i.e. competition, constraint, and participation) in each model is an important part of my approach. 74 Benjamin E. Goldsmith at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from While Clarke and Stone (2008)suggest a control for the aggregate regimetype measure, this would not allow me to concurrently test several competing institutional expectations, and it would introduce more collinearity into the model. Since I use all six indicators that comprise the polity index, a control for the polity index itself would be redundant. 4 Results In this section, I present the empirical results, beginning with what I believe are the best specified, most appropriate tests of the hypotheses. I then discuss a number of robustness checks. Finally, I address the issue of the relative peace of East Asia since 1979 and increases in levels of political competition. I first present probit models for all dyads globally, excluding East Asian dyads. The dependent variable for these models is the initiation of an international crisis (as defined in the ICB data set) by State A (the ‘triggering entity’) against State B. The democratic peace effect overall is supported in Model 1 in Table 1. The interaction of the Regime Type of states A and B is negatively and significantly associated with State A’s initiation of a crisis with State B as a target. That is, the more democratic A, the less likely it is to initiate a crisis against B if B is also relatively democratic. My three regime component variables are substituted for A’s overall regime indicator in Model 2. This tests H1 that political competition in A will have a negative (pacifying) effect on the chances of crisis initiation as B’s level of democracy increases. As expected, the coefficient for CompetitionA interacted with Regime TypeB is negative, and significant, supporting the first hypothesis. Neither the interaction for ParticipationA nor ConstraintA is significant, however, casting doubt on the importance of some other approaches. It is preferable to present marginal effect plots for interaction terms in limited dependent variable models like probit. To save space, I do so only for the analysis of central concern in this paper, that of East Asia.6 I now turn to the results for East Asian dyads. These are presented in Models 3–5, and in the marginal effect plots in Figure 1. There is some evidence of a democratic peace effect when the interaction of A and B’s 6 I did produce such plots based on simulations for all analyses, and the results shown in Models 1 and 2 are supported in terms of sign and significance. Domestic institutions and conflict in East Asia 75 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Table 1 Crisis initiations by State A against State B, 1948–2006 Model 1 Model 2 Model 3 Model 4 Model 5 Global, excluding East Asia Global, excluding East Asia East Asia East Asia East Asia coef. s.e. sig. coef. s.e. sig. coef. s.e. sig. coef. s.e. sig. coef. s.e. sig. Regime TypeA 0.02 (0.01) ** 0.04 (0.02) ** Regime TypeA × Regime TypeB −0.00 (0.00) *** −0.01 (0.00) ** CompetitionA 0.01 (0.02) 0.06 (0.09) 0.00 (0.05) CompetitionA × Regime TypeB −0.00 (0.00) *** −0.02 (0.01) *** −0.02 (0.01) *** ParticipationA −0.01 (0.01) −0.09 (0.04) ** −0.05 (0.03) * ParticipationA × Regime TypeB 0.00 (0.00) 0.01 (0.00) * 0.01 (0.00) *** ConstraintsA 0.01 (0.02) −0.03 (0.10) −0.00 (0.05) ConstraintsA × Regime TypeB 0.00 (0.00) 0.01 (0.01) * 0.01 (0.00) * Regime TypeB 0.02 (0.01) *** 0.02 (0.01) 0.05 (0.01) *** −0.04 (0.05) −0.08 (0.05) * ln(TradeShareA) 0.01 (0.01) 0.00 (0.01) 0.09 (0.05) 0.10 (0.06) * ln(TradeVolume) −0.00 (0.01) −0.00 (0.01) −0.11 (0.05) ** −0.12 (0.06) ** CivilConflictA 0.24 (0.03) *** 0.25 (0.03) *** −0.62 (0.2) *** −0.74 (0.17) *** CivilConflictB 0.15 (0.04) *** 0.17 (0.04) *** −0.35 (0.15) *** −0.41 (0.13) *** Contiguity 0.92 (0.09) *** 0.91 (0.08) *** 1.26 (0.3) *** 1.46 (0.31) *** 1.24 (0.22) *** ln(Distance) −0.20 (0.04) *** −0.21 (0.04) *** −0.14 (0.11) −0.11 (0.12) −0.10 (0.09) PowerA 5.39 (0.54) *** 5.53 (0.54) *** 5.96 (2.18) *** 6.10 (2.60) *** PowerB 5.00 (0.60) *** 5.01 (0.63) *** 1.21 (2.95) 2.02 (3.26) ln(USMilAid RatioA) −0.00 (0.00) −0.00 (0.00) −0.00 (0.01) −0.00 (0.00) ln(USMilAid RatioB) −0.00 (0.00) −0.00 (0.00) 0.00 (0.01) 0.01 (0.01) 76 Benjamin E. Goldsmith at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Regional Alliance Similarity −0.13 (0.18) −0.14 (0.18) −6.22 (4.50) −8.36 (4.57) * Global Alliance Similarity −0.08 (0.25) −0.07 (0.27) 8.92 (6.08) 11.72 (6.31) * Joint IGO Memberships 0.01 (0.00) *** 0.01 (0.00) *** 0.02 (0.01) ** 0.02 (0.01) ** Peace Years −0.05 (0.01) *** −0.05 (0.01) *** −0.09 (0.03) *** −0.09 (0.02) *** −0.12 (0.03) *** Peace Years 2 0.00 (0.00) *** 0.00 (0.00) *** 0.00 (0.00) *** 0.00 (0.00) *** 0.00 (0.00) *** Peace Years 3 −0.00 (0.00) *** −0.00 (0.00) *** −0.00 (0.00) *** −0.00 (0.00) *** −0.00 (0.00) ** Constant −1.83 (0.29) *** −1.65 (0.31) *** −4.52 (1.64) *** −3.93 (2.02) * −1.15 (0.75) N 743,999 721,286 8013 7699 10,729 Wald chi2(19,23,19,23,12) 1568.8 *** 1723.7 *** 446.4 *** 1036.7 *** 292.8 *** Pseudo R 2 0.287 0.289 0.356 0.384 0.302 coef., Coefficient; sig., Significance. *P< 0.10, **P< 0.05, ***P< 0.01; conflict onset measured at year t, independent variables at t-1; directed dyads; probit models with standard errors corrected for clustering on directed dyads. Domestic institutions and conflict in East Asia 77 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Regime Type is considered in Model 3, contrary to some previous findings in the literature (perhaps due to the directed-dyad design focused on crisis initiation, rather than simply the occurrence of a conflict). However, the upper left panel in Figure 1, showing the marginal effect derived from Model 3, reveals this to be relatively small. To test H2, regarding political competition in East Asia, the interaction term in Model 4 and the lower left panel in Figure 1 are relevant. As was the case in the global data, for East Asian dyads, political competition appears to inhibit crisis initiation against target states that are relatively democratic. This is evident in the negative and significant coefficient in Model 4, as well as in the marginal effect plot, which shows a statistically significant, relatively large substantive effect. Examining the two other component interaction terms in Model 4, they may have a small positive relationship with conflict initiation, but this is only significant at the 90% level. Any substantive impact appears to be small, judging by the right panels in Figure 1. The findings therefore seem clear. Both hypotheses find considerable support. My expectations regarding political competition and the initiation of crises against relatively democratic regimes are upheld in this Figure 1 East Asia: contiguous states: marginal effects (95% CI, mean ±1 SD). 78 Benjamin E. Goldsmith at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from initial and, I believe, most appropriate set of tests. However, I now turn to a discussion of the robustness of these results in a range of other tests. 4.1 Robustness tests One concern that might be raised is that my findings are sensitive to model specification. Model 5 in Table 1 addresses this by dropping all of the control variables, with the exception of those for geographic proximity and temporal dependence which I see as essential to properly modeling the outcome. While there are minor differences in the coefficients or signifi- cance for other aspects of regime type, the result for the interaction of CompetitionA and Regime TypeB is unchanged. Another concern might be that the results depend on some unusual features of the dependent variable, the initiation of an international crisis. To address this, I use the alternative measures of conflict initiation by state A against State B. First, in Models 6–7 in Table 2, I restrict crises to those categorized as‘clashes’ in the ICB data. These are more severe crises, and
20% of the crises do not meet this threshold in East Asia over the period of study (9 of 43). Next, in Models 8–9, I use another conflict data set that allows me to code initiators and targets, although not with as much confidence as the ICB data. The MID data set codes the first state in a dyad to show, threaten, or use force in a way that appears to be directed at the target. Ghosn et al. (2004, pp. 138–139) suggest‘caution against misinterpretations of identification of the ‘initiator’. The state or states on Side A on the first day of the dispute are simply the first states to take codeable military action. They should not be interpreted to be the states that ‘started’ the conflict, or that are responsible for the conflict’. While it is reasonable to expect that the first state to use or threaten force directly is likely to be the initiator of the militarization of the dispute, this does not seem as good a measure as the ICB coding provides. I also use a more restrictive set of MIDs, those which incur at least one battle death, to ensure that what is being coded is serious con- flict, not just an example of belligerent bluster. Across all of these alternative indicators of conflict initiation, of my three component interaction terms, only that for political competition produces any negatively signed coefficient in Models 6–9. Indeed, all of its coefficients are negative, and all save that for all MIDs are statistically significant (P= 0.18). I include models with and without the control variables for the ICB clashes (Models 6 and 7, respectively), as a further check. Domestic institutions and conflict in East Asia 79 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Table 2 East Asia, alternative conflict indicators: initiations by State A against State B Model 6 Model 7 Model 8 Model 9 ICB clash initiation (1948–2006) ICB clash initiation (1948–2006) MID initiation (1948–2001) Fatal MID initiation (1948–2001) coef. s.e. sig. coef. s.e. sig. coef. s.e. sig. coef. s.e. sig. CompetitionA 0.08 (0.09) 0.01 (0.05) 0.02 (0.05) 0.06 (0.05) CompetitionA × Regime TypeB −0.03 (0.01) *** −0.02 (0.00) *** −0.01 (0.00) −0.02 (0.01) *** ParticipationA −0.10 (0.04) ** −0.07 (0.03) *** 0.01 (0.03) −0.01 (0.06) ParticipationA × Regime TypeB 0.01 (0.00) *** 0.02 (0.00) *** 0.00 (0.00) 0.00 (0.00) ConstraintsA −0.05 (0.09) 0.00 (0.05) −0.03 (0.05) −0.06 (0.05) ConstraintsA × Regime TypeB 0.01 (0.01) 0.01 (0.00) * 0.01 (0.00) 0.02 (0.00) *** Regime TypeB −0.12 (0.05) ** −0.21 (0.06) *** 0.02 (0.05) 0.04 (0.07) ln(TradeShareA) 0.10 (0.06) 0.01 (0.03) −0.03 (0.06) ln(TradeVolume) −0.13 (0.06) ** 0.00 (0.03) 0.01 (0.06) CivilConflictA −0.68 (0.21) *** 0.00 (0.09) −0.17 (0.13) CivilConflictB −0.55 (0.21) *** 0.00 (0.06) −0.18 (0.15) Contiguity 1.32 (0.38) *** 1.22 (0.26) *** 0.46 (0.15) *** 1.09 (0.25) *** ln(Distance) −0.14 (0.12) −0.09 (0.10) −0.50 (0.09) *** −0.45 (0.11) *** PowerA 7.15 (3.34) ** 4.86 (1.59) *** 0.79 (2.70) PowerB 3.64 (2.92) 6.02 (1.69) *** 5.03 (3.24) ln(USMilAid RatioA) 0.00 (0.01) 0.01 (0.00) * 0.00 (0.01) ln(USMilAid RatioB) 0.00 (0.01) 0.01 (0.00) * 0.01 (0.01) Regional Alliance Similarity −12.88 (6.41) ** −6.52 (1.93) *** 0.34 (3.48) 80 Benjamin E. Goldsmith at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Global Alliance Similarity 19.26 (9.34) ** 8.35 (2.64) *** −1.47 (4.62) Joint IGO Memberships 0.02 (0.01) ** −0.01 (0.00) * 0.00 (0.01) Peace Years −0.14 (0.04) *** −0.14 (0.03) *** −0.12 (0.02) *** −0.14 (0.03) *** Peace Years2 0.01 (0.00) *** 0.01 (0.00) *** 0.00 (0.00) *** 0.00 (0.00) *** Peace Years3 0.00 (0.00) ** 0.00 (0.00) *** 0.00 (0.00) *** 0.00 (0.00) *** Constant −6.02 (2.95) ** −0.64 (0.8) −0.33 (1.04) 1.21 (1.52) N 7689 10,715 6496 6419 Wald chi2(19,12,23,23) 599.2 *** 220.7 *** 378.1 *** 734.8 *** Pseudo R 2 0.416 0.313 0.304 0.443 coef., Coefficient; sig., Significance. *P< 0.10, **P< 0.05, ***P< 0.01; conflict onset measured at yeart, independent variables att-1; directed dyads; probit models with standard errors corrected for clustering on directed dyads. Domestic institutions and conflict in East Asia 81 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from In addition to the robustness tests presented in the tables, I conducted others which I describe, but for which I do not show results to save space. Although the issue of multicollinearity between the regime-component variables might be a concern, the results are not affected by it. The variance inflation factors (VIFs) for regime components in East Asia do not exceed the threshold of 10. They do exceed 5, which is sometimes considered an indicator of possible multicollinearity. The danger here would be that variables with similar effects included in the same model would exhibit distortions. When models otherwise equivalent to Model 4 are run, but using only each regime component and its interaction with State B’s regime type, dropping the other two components and their interactions, the results are nearly unchanged in terms of sign and significance of the component terms and their interactions. Crucially, the interaction for ConstraintsA and Regime TypeB is positive and insignificant, that for ParticipationA and Regime TypeB is positive and significant (P= 0.012), and that for CompetitionA and Regime TypeB remains negative and significant (P= 0.002). Thus, H2 retains strong support, and no other regime component takes on an analogous effect. Recently, a ‘territorial peace’ argument (Miller and Gibler, 2011; Gibler, 2012) has been advanced positing that democracies are unlikely to be involved in territorial disputes with each other, because settled borders make the development of a democratic system itself more likely. Thus, there could be a degree of endogeneity to the observed democratic peace. However, in East Asia, this proposition might seem less plausible. Some jointly democratic dyads have outstanding territorial issues and have experienced serious confrontations, including South Korea and Japan, Taiwan and Japan, and the Philippines and Malaysia (when Malaysia was coded as democratic and the Philippines nearly so in the 1960s). Nevertheless, I have run my main models 4 and 5 while including a control for territorial claims developed by Huth and Allee (2002a). The sign and significance for the interaction of CompetitionA and Regime TypeB are not meaningfully changed in either model, although the territorial claim indicator is positive and significant as well. 4.2 Timing of the East Asian peace In this section, I briefly present evidence that supports a role for political competition as a partial explanation for the observed East Asian peace 82 Benjamin E. Goldsmith at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from after 1979. My study is focused only on the initiation of international crises, rather than the escalation of ongoing crises to war. This limits its implications for understanding the absence of war itself.7 However, other things equal, if there are fewer crises initiated, then there should be fewer opportunities for escalation to full-scale war. In addition, looking to the future, the implication of my findings for East Asia if political competition continues to increase in the region is that this will further reduce the frequency of crises and help maintain the peace, especially if overall democracy levels also continue to increase. Empirically, it is the case that levels of political competition in East Asia began to increase around the late 1970s, as evident in Figures 2 and 3, and this trend has continued, perhaps reaching a plateau at a much higher level toward the start of the 21st century. Overall democracy levels also increased along a similar trajectory. This increase in competition corresponded with a fall-off in the annual frequency of crisis initiations in the region (shown on the left axis of Figure 2 as a portion of all regional directed dyads per year). Combined with the statistical results in Tables 1 and 2, this lends support to the contention that a rise in political competition (and relative democracy of potential target states) led to a considerable Figure 2 Crises and regimes in East Asia, 1948–2006. 7 Elsewhere I propose that trade volume in East Asia strongly inhibits conflict escalation to war and shows a steep increase in the region after 1978 (Goldsmith, 2013). Domestic institutions and conflict in East Asia 83 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from decrease in crisis initiations, which seems especially apparent from the late 1980s, as average governance levels approached those of functioning democracies (16 or 17 [6 or 7 on the Polity scale] are common thresholds for democracy). This possibility is illustrated by examination of two important sets of directed dyads with different values for the key variables. For the North– South Korea dyad, the South’s increase in political competition appears to have little effect on its propensity to initiate disputes with the North, nor does the South’s rise in overall democracy seem to affect the likelihood of the authoritarian North’s initiation. By way of contrast, once Cambodia’s level of competition increased after Vietnam deposed the Khmer Rouge, Cambodia did not initiate any disputes with Thailand, which had Regime Type levels close to functioning democracy. Similarly, as Thailand’s political competition rose in the 1980s during its democratization process, Thailand did not initiate crises with Cambodia. This lack of crisis initiations in spite of Thailand’s giving refuge to the Khmer Rouge, and Vietnamese military forays into Thai territory from Cambodia, seems telling of the degree of restraint each side exercised to not directly threaten Figure 3 Directed dyads. 84 Benjamin E. Goldsmith at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from each other in more serious ways (but more recent events might contradict my expectations8 ). Although a discussion in any detail of the possible causes of the regional increases in political competition and democracy levels from the late 1970s onwards is beyond the scope of this article, one plausible explanation would seem to be the relaxation of regional tensions, and the de-ideologization of regional international relations, that occurred after the US-PRC rapprochement in the 1970s and the PRC’s reforms under Deng Xiaoping begun in 1978. The relaxation of regional security tensions allowed regional governments greater security (Yahuda, 2004, pp. 174– 183). One aspect of this was the withdrawal of Chinese support for rebel groups. For example, in 1978 Deng committed to withdraw support for Thai communists, and the following year the PRC ceased revolutionary propaganda broadcasts to the country (Vogel, 2011, p. 284, 786n50). Similar moves were made in relations with Indonesia, Malaysia, and Singapore (Vogel, 2011, pp. 283–287). It also may have enhanced internal stability of regional regimes by reducing the perception that the (nowabandoned) Maoist model was a viable alternative to the status quo. This increased security should have given freer rein to forces behind the global diffusion of democracy (Simmons et al., 2008). In addition, Sino-US rapprochement over the period 1972–1979 (from Nixon’s visit to Carter’s normalization of relations) probably increased US influence in the region (relative to the USSR), allowing it to apply pressure for democratization when or where it chose (although the United States of course also cooperated closely with illiberal but allied regimes such as Suharto’s Indonesia). 5 Conclusion This article’s central contribution is to provide what I believe is compelling evidence for a democratic peace effect in East Asia, driven by institutions providing for genuine political competition within states in the region, even if they fall short of democracy along other dimensions. I have presented evidence supporting a primary role for institutions of political 8 In 2008, a militarized conflict began on the Thai-Cambodian border over the status of a Temple. This is beyond the temporal scope of my study, or either the ICB or MID data, but the dynamics of this conflict should provide important case evidence for further assessing my theory in East Asia. Especially, if Thailand is considered the initiator, this might provide some evidence contrary to my expectations, which are of course probabilistic, not deterministic. Domestic institutions and conflict in East Asia 85 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from competition in the democratic peace effect in East Asia. States with higher levels of political competition are less likely to initiate international crises against targets with higher levels of democracy in the region, mirroring the global pattern. Neither constraints on the executive nor political participation exhibits a similar dyadic pattern. While there is some indication that overall democracy levels for the initiating state are also associated with a lower likelihood of initiation against democracies in East Asia, the substantive impact of this broader measure is considerably smaller than the relatively large apparent effect of political competition. These results are obtained in models controlling for a range of possibly confounding variables relating to trade, capabilities, intra- and extra-regional military and political ties, civil war, and other domestic political institutions. They are also robust to concerns about multicollinearity and model and variable specification. I have also shown that the rise in regional levels of political competition corresponds well to the observed ‘East Asian peace’, although at most my analysis provides a part of the explanation for that phenomenon because I do not assess the escalation stage, in which crises might escalate into wars. But political competition provides a compelling general explanation for reduced conflict initiation in the region from the 1980s onward. The implication of my analysis is that the deepening of political competition and democratic governance in East Asia will also help maintain, but of course not guarantee, this regional era of interstate peace. Funding Support from the East Asia Peace Programme, Uppsala University, and the Australian Research Council (DP1093625) is gratefully acknowledged. Previous versions of this paper were presented at the East Asian Security Workshop, Kobe University, 18–19 April 2013, the Eighth International Convention of Asia Scholars (ICAS 8), Macao, 24–27 June 2013, and at the Australian National University, 27 September 2013. I thank all participants in those sessions for helpful feedback. References Acharya, A. (2010) ‘Democracy or death? Will democratisation bring greater regional instability to East Asia?’,Pacific Review, 23(3), 335–358. 86 Benjamin E. Goldsmith at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Achen, C.H. and Bartels, L.M. (2004) Blind Retrospection: Electoral Responses to Drought, Flu, and Shark Attacks. (typescript, Princeton University). Barbieri, K. (2002) The Liberal Illusion: Does Trade Promote Peace. Ann Arbor: University of Michigan Press. Barbieri, K., Keshk, O.M.G. and Pollins, B.M. (2008) Correlates of war project trade data set codebook. Version 2.01. http://www.correlatesofwar.com (3 October 2008, date last accessed). Bennett, D.S. and Stam, A. (2000)‘EUGene: A conceptual manual’,International Interactions, 26, 179–204. Brambor, T., Clark, W. R. and Golder, M. (2010) ‘Understanding interaction models: improving empirical analyses’,Political Analysis, 14, 63–82. Braumoeller, B.F. (2004) ‘Hypothesis testing and multiplicative interaction terms’, International Organization, 58, 807–820. Bueno de Mesquita, B., Morrow, J.D., Siverson, R.M. and Smith, A. (1999) ‘An institutional explanation of the democratic peace’, American Political Science Review, 93(4), 791–807. Bueno de Mesquita, B., Smith, A., Siverson, R. and Morrow, J. (2003) The Logic of Political Survival. Cambridge: MIT. Bueno de Mesquita, B., Koch, M.T. and Siverson, R.M. (2004) ‘Testing competing institutional explanations of the democratic peace: the case of dispute duration’, Conflict Management and Peace Science, 21, 255–267. Carter, D. and Signorino, C.S. (2010) ‘Back to the future: modeling time dependence in binary data’,Political Analysis, 18, 271–292. Choi, J.K. (2010)‘Sunshine over a barren soil: the domestic politics of engagement identity formation in South Korea’, Asian Perspective, 34(4), 115–138. Chong, D. and Druckman, J.N. (2007) ‘Framing public opinion in competitive democracies’, American Political Science Review, 101(4), 637–655. Clarke, K.A. and Stone, R.W. (2008) ‘Democracy and the logic of political survival’, AmericanPolitical Science Review, 102(3), 387–392. Colaresi, M. (2004) ‘Aftershocks: Postwar leadership survival, rivalry, and regime dynamics’,International Studies Quarterly, 48(4), 713–727. Dixon, W.J. and Senese, P.D. (2002) ‘Democracy, disputes, and negotiated settlements’, Journal of Conflict Resolution, 46, 547–571. Entman, R.M. (2004) Projections of Power: Framing News, Public Opinion, and U.S. Foreign Policy. Chicago: University of Chicago Press. Gelpi, C., Reifler, J. and Feaver, P. (2007) ‘Iraq the vote: Retrospective and prospective foreign policy judgments on candidate choice and casualty tolerance’, Political Behavior, 29(2), 151–174. Ghosn, F., Palmer, G. and Bremer, S.A. (2004) ‘The MID3 data set, 1993–2001: Procedures, coding rules, and description’, Conflict Management and Peace Science, 21, 133–154. Domestic institutions and conflict in East Asia 87 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Gibler, D.M. (2012) The Territorial Peace: Borders, State Development, and International Conflict. New York: Cambridge University Press. Gleditsch, N.P., Wallensteen, P., Eriksson, M., Sollenberg, M. and Strand, H. (2002) ‘Armed conflict 1946–2001: A new dataset’, Journal of Peace Research, 39(5), 615–637. Goldsmith, B.E. (2006) ‘A universal proposition?: region, conflict, war, and the robustness of the liberal peace’, European Journal of International Relations, 12(4), 533–563. Goldsmith, B.E. (2007) ‘A liberal peace in Asia?’, Journal of Peace Research, 44 (1), 5–27. Goldsmith, B.E. (2013) ‘Different in Asia?: developmental states, trade, and international conflict onset and escalation’, International Relations of the Asia Pacific, 13(2), 175–205. Goldsmith, B.E. (2013) ‘International Trade and the Onset and Escalation of Interstate Conflict: More to Fight About, or More Reasons Not to Fight?’ Defence and Peace Economics, Published online 20 February 2013, Doi 10.1080/10242694.2013.763637, in press. Henderson, E.A. (2009)‘Disturbing the peace: African warfare, political inversion and the universality of the Democratic Peace Thesis’, British Journal of Political Science, 39(1), 25–58. Huth, P.K. and Allee, T.L. (2002a) The Democratic Peace and Territorial Conflict in the Twentieth Century. Cambridge, UK: Cambridge University Press. Huth, P.K. and Allee, T.L. (2002b) ‘Domestic political accountability and the escalation and settlement of international disputes’, Journal of Conflict Resolution, 46(6), 754–790. Jentleson, B.W. (1992) ‘The pretty prudent public: post post-Vietnam American opinion on the Use of Military Force’, International Studies Quarterly, 36(1), 49–73. Johnston, A.I. (2012) ‘What (If anything) does East Asia tell Us about International Relations Theory?’, Annual Review of Political Science, 15, 53–78. Kant, I. ([1795] 1917) Perpetual Peace: A Philosophical Essay. London: George Allen and Unwin Ltd. King, G., Tomz, M. and Wittenberg, J. (2000) ‘Making the most of statistical analyses: improving interpretation and presentation’, American Journal of Political Science, 44, 347–361. Kivimaki, T. (2001) ‘The long peace of ASEAN’, Journal of Peace Research, 38 (1), 5–25. Lake, D.A. (1992) ‘Powerful pacifists: Democratic states and war’, American Political Science Review, 86(1), 24–37. Lind, J. (2011) ‘Democratization and stability in East Asia’, International Studies Quarterly, 55(2), 409–436. 88 Benjamin E. Goldsmith at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Marshall, M.G., Gurr, T.R. and Jaggers, K. (2010) Polity IV Project: Political Regime Characteristics and Transitions, 1800–2009, Dataset Users’ Manual, Center for Systemic Peace. www.systemicpeace.org/polity/polity4.htm (29 March 2011, date last accessed). Miller, S. and Gibler, D.M. (2011) ‘Democracies, territory, and negotiated compromises’, Conflict Management and Peace Science, 28(3), 261–279. Morrow, J.D., de Mesquita, B.B., Siverson, R.M. and Smith, A. (2008) ‘Retesting selectorate theory, separating the effect of w from other elements of democracy’, American Political Science Review, 102(3), 393–400. Pevehouse, J.C., Nordstrom, T. and Warnke, K. (2004) ‘The COW-2 international organizations dataset version 2.0’, Conflict Management and Peace Science, 21 (2), 101–119. Regan, P. and Leng, R.J. (2003) ‘Culture and negotiation in militarized interstate disputes’, Conflict Management and Peace Science, 20(2), 111–132. Reiter, D. and Stam, A.C. (2002) Democracies at War. Princeton, NJ: Princeton University Press. Reiter, D. and Stam, A.C. (2003) ‘Identifying the culprit: democracy, dictatorship, and dispute initiation’, American Political Science Review, 97(2), 333–337. Reiter, D. and Tillman, E.R. (2002) ‘Public, legislative, and executive constraints on the Democratic Initiation of Conflict’, Journal of Politics, 64(3), 810–826. Russett, B. and Oneal, J.R. (2001)Triangulating Peace: Democracy, Interdependence, and International Organizations. New York: W.W. Norton & Co. Schultz, K.A. (1999) ‘Do democratic institutions constrain or inform? Contrasting two institutional perspectives on democracy and war’, International Organization, 53(2), 233–266. Schultz, K.A. (2001) Democracy and Coercive Diplomacy. Cambridge, UK: Cambridge University Press. Schumpter, J.A. (1950) Capitalism, Socialism and Democracy. London: George Allen & Unwin. Signorino, C.S. and Ritter, J.M. (1999) ‘Tau-b or Not Tau-b: Measuring the similarity of foreign policy positions’, International Studies Quarterly, 43(1): 115–144. Simmons, B., Dobbin, F. and Garrett, G., eds. (2008) The Global Diffusion of Markets and Democracy. Cambridge University Press. Solingen, E. (2007) ‘Pax Asiatica versus Belli Levantina: The foundations of War and peace in East Asia and the Middle East’, American Political Science Review, 101(4), 757–780. Svensson, I. (2011) ‘East Asian peacemaking: exploring the patterns of conflict management and conflict settlement in East Asia’, Asian Perspective, 35(2), 163–186. Domestic institutions and conflict in East Asia 89 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Svensson, I. and Lindgren, M. (2011) ‘From bombs to banners? The decline of wars and the rise of unarmed uprisings in East Asia’, Security Dialogue, 42(3), 219–237. Tang, C.-M. (2012) ‘Southeast Asian peace revisited: a capitalist trajectory’, International Relations of the Asia-Pacific 12(3), 389–417. Tønnesson, S. (2009) ‘What is it that best explains the East Asian peace since 1979? A call for a research agenda’, Asian Perspective 33(1), 111–136. USAID. (2012) Overseas Loans & Grants [Greenbook]. http://gbk.eads. usaidallnet.gov/(22 October 2012, last date accessed). Vogel, E.F. (2011) Deng Xiaoping and the Transformation of China. Cambridge, MA: Belknap Press. Wilkenfeld, J. and Brecher, M. (2009) Codebook for ICB2 International Crisis Behavior Project. Version 9.0. http://www.cidcm.umd.edu/icb (25 March 2010, last date accessed). Yahuda, M. (2004) The International Politics of the Asia-Pacific. 2nd edn. London: RoutledgeCurzon.