Decline causes war.

The present study explores whether and how eco- nomic slowdowns affect the likelihood of international conflict. Do troubled economies make states more aggressive? If so, against which other states? The main argument of this article is that a slowing economy increases the probability of interstate militarized conflict. It also finds that economically troubled autocracies tend to target democracies whose economies are dependent on them. However, although political unrest contributes positively to conflict initiation, there is little evidence that politically troubled states are more likely to target their economic partners. Before World War II, for exam- ple, Nazi Germany was a rising but anxious power with deep concerns about its economic potential and territor- ial autonomy, which prompted it to attack its neighbor- ing trade partners (Tooze, 2008). Similarly, when faced with an economy that had been deteriorating since the late 1930s, imperial Japan attacked Pearl Harbor to save the Greater East Asia Co-Prosperity Sphere (Beasley, 2011[1990]: chs 7 & 8).

The next section of this article presents a critical review of theorizing on diversionary conflict, focusing on the implications of slower growth and domestic unrest. It then advances a theory as to how economic slowdowns lead to diversionary conflict, providing a cau- sal mechanism for (autocracy’s) strategic targeting of (democratic) economic partners with testable hypoth- eses. Next, it tests these hypotheses by conducting a statistical analysis of directed dyads from 1950 to 2010. A discussion of the results follows. The concluding section considers the implications for current and future international relations, focusing on the possible of con- sequences of major powers’ economic slowdowns, if not economic depressions, and their relations with neighbors in an era of a globalized economy and overall decline of democracy.

An economic theory of diversionary conflict

Economic unrest and international conflict

Economic growth is an important factor in studies of diversionary conflicts. The domestic approach to international conflict emphasizes internal unrest – economic as well as political – as a cause of international and domestic conflict (for a review, see Jung, 2019; Levy, 1989). Indeed, many historians and commentators have attributed serious international and domestic violence to internal unrest that challenged political leadership. By initiating hardline policies or aggressive actions toward foreign states or minority groups, domestically embattled leaders can divert their constituents’ attention away from internal ills and create rally-round-the-flag effects. The Falklands War (1982) is a typical example. After Argentina failed to achieve economic reforms, its junta took military action to occupy the Falkland Islands, thereby gaining ardent albeit short-lived support from its domes- tic audience (Levy & Vakili, 1992; Oakes, 2012). The ‘ice-cold bunch of rocks’, as US President Ronald Reagan characterized the islands, seemed strategically unimportant, but it was politically critical to the unpopular leaders of Argentina.

This approach views a bad economy as a driving force incentivizing political leaders to initiate diversionary conflict as a way of proving their ability to lead the country (e.g. Bennett & Nordstrom, 2000; Brule ́, Marshall & Prins, 2010; Fordham, 1998; Heldt, 1999; Solt, 2011). Of course, the leaders could also gain support by rebooting the economy, but they often feel unable to do so. International conflict can yield two distinct benefits to domestically threatened leaders (Haynes, 2017). First, an external threat increases internal cohesion. Many studies of intergroup psychology have found that favorable attitudes toward in-group members become stronger when individuals perceive a more severe threat from (or com- petition with) out-group members. Thus, a conflict with a foreign state can help an incumbent leader maintain political control (Theiler, 2018). Second, when facing domestic concerns about political leadership, leaders can engage in military conflict to demonstrate their competence (Richards et al., 1993). If they meet public expectations or even achieve a glorious outcome, their popularity can be restored.

The effects of economic conditions on international conflict have long been debated, at least since Ostrom & Job’s (1986) study of US presidents’ use of force. Based on statistical analyses in large-N studies, some scholars have argued that the diversionary theory of war is supported only by anecdotal evidence, and that there is no significant relationship between domestic unrest and international conflict (Leeds & Davis, 1997; Meernik & Waterman, 1996). Even if some historical cases are classified as diversionary conflicts, these scho- lars contend, we cannot generalize a causal theory from them. Indeed, few insist that domestic unrest is a suf- ficient condition for international conflict, though it may cause political leaders to contemplate conflict initiation. Some studies even argue that domestically troubled states tend to be targets, not initiators, of international conflicts due to their weakened capability and unwillingness to fight (Blainey, 1988; Clark, Fordham & Nordstrom, 2011; Foster, 2006). Thus, during the last four decades, studies of diversionary conflict have aimed to identify (1) the conditions under which such conflicts occur, (2) the types of domestic unrest that produce international conflict, and (3) alter- native options to the diversionary use of force against foreign states, as discussed below.

First, many studies have shed light on dyadic or domestic conditions that facilitate diversionary conflict. Domestically challenged leaders are prudent in choosing their diversionary targets so as to remain in office. Targeting a territorial or strategic rival can revive domestic support because there is little need to justify the military action to the domestic public (Heldt, 1999; Mitchell & Prins, 2004; Tir, 2010). Also, political leaders may target states with rising power and/or a different identity to exploit their domestic audience’s fear of these rivals (Jung, 2014a), whereas they avoid attacking states that are either too powerful or too weak, because either win- ning is impossible or it will not strengthen their domestic standing (Tarar, 2006). Human rights violators may become democratic regimes’ diversionary targets because a liberal audience readily supports humanitarian inter- vention (Tokdemir & Mark, 2018).

Another condition for the onset of diversionary conflict is regime type. Some scholars claim that democratic leaders facing hard times domestically are more likely than authoritarian ones to initiate diversionary conflict (Gelpi, 1997; Smith, 1996). These researchers have con- cluded that a bad economy, an upcoming election, and low approval ratings may encourage democracies to pur- sue international conflicts. But others oppose this argument, countering that democratic leaders are more constrained by other political actors such as opposition parties and legislatures (Miller, 1995) and less capable of hiding their diversionary incentives (Bell, 2013). Follow- ing this logic, some would expect authoritarian regimes to be more likely diverters than democratic ones. Indeed, non-personalist leaders of juntas and one-party regimes are concerned about their states’ political and military elites, so they sometimes exploit international events to maintain domestic control (Weeks, 2014).

Second, some studies have examined how distinct types of domestic unrest may push leaders to engage in international conflict. Because diversionary incentives cannot be measured directly, many studies focus on a particular type of domestic unrest: political, economic, social, or military. Little consensus has been reached, however. Some use economic variables such as unem- ployment and inflation rates or the Gini index as proxies for diversionary incentives (Fordham, 1998; Solt, 2011). But this measurement is invalid if the domestic public blames previous leaders or foreign actors for a bad economy. In fact, when a leader has recently assumed office or when a global economic crisis occurs, the domestic public typically shows support for its leadership even while demanding economic reforms.

Other analyses have measured political unrest in terms of congressional opposition (Brule ́, 2006), partisan support (Morgan & Bickers, 1992), upcoming elections (Gaubatz, 1991), presidential scandals (Hendrickson, 2002), low presidential approval ratings (Li, James & Drury, 2009), and coup risk (Miller & Elgu ̈n, 2011). However, political events vary in terms of the number of participants, intensity of opposition to leadership, and effects on the central government in charge of foreign affairs. Therefore, a few studies have focused on the political leader’s party and his or her support base to measure the extent to which the leadership was shaken (Brule ́ & Williams, 2009; Fordham, 1998). In addition, collecting data on political events across nations and times is difficult, so only relatively recent information on developed states is reliably attainable and usable for testing the diversionary hypothesis.

Third, some studies have pointed out alternatives to targeting foreign states in response to domestic chal- lenges. Political leaders can resolve domestic problems directly, rather than divert public attention from them, by carrying out domestic reforms. Oakes (2012) described the failure of economic reform by the junta as a precondition for its occupation of the Falklands. Moreover, political leaders sometimes choose to repress opposition voices and strengthen internal control (Davies, 2016; Enterline & Gleditsch, 2000). If reform and repression are unavailable, domestically challenged leaders may target ethnic or religious minority groups or refugees and exploit the majority’s in-group cohesion without risking high costs (Oakes, 2012: ch. 5; Savun & Gineste, 2019; Tir & Jasinski, 2008). These studies highlight a menu for political leaders facing domestic problems, which includes foreign diversion as well as domestic reform, repression, and internal diversion.

To summarize, domestic unrest has been widely seen as causing international conflict. When the economy worsens, for example, political leaders often struggle to maintain domestic control and may view conflict with a foreign state as a way to divert public attention and gain domestic support. However, there is mixed evidence to support this theory of diversionary conflict. Therefore, various types of domestic unrest and their interactions with other variables should be considered to provide a more satisfactory explanation of international conflict based on sound empirical evidence.

A slowing economy and foreign aggression

Does an economic slowdown lead to diversionary con- flict? As discussed above, research on diversionary con- flict has highlighted different types of unrest. The present study uses a negative trend in economic growth in recent years as evidence of a slowing economy, rather than static values of per-capita income and unemploy- ment rates in any given year. A slowing economy causes many individuals to struggle economically, fear for the future, and, as a consequence, be less supportive of their government, even if their state is growing faster than or catching up with neighboring or rival states. To take a specific example, this is why some experts on China worry that its middle-income trap makes the Chinese Communist Party ‘feel compelled to act strongly and forcefully abroad’ (Pence, 2020). Although some studies argue that individuals use an international rather than historical benchmark to assess their government’s eco- nomic performance (e.g. Kayser & Peress, 2012), others conclude that the public punishes incumbent leaders for a worsening economy (Hayes et al., 2015; Lewis-Beck & Stegmaier, 2000). When expecting losses, or negative changes from their reference point, individuals prefer a risk-accepting option (changing or opposing leadership).

<<TEXT CONDENSED NONE OMITTED>>

Even in troubled economies, some leaders become popular in the event of small economic improvements, which may or may not have resulted from their reform programs. Since ordinary people, especially poorly edu- cated ones, use an historical reference point when asses- sing leader competence (Aytaç, 2018), economic trends (improving or worsening) are more important than eco- nomic outcomes (good or bad) for domestic popularity and legitimizing the incumbent’s leadership. A sound but worse economy does not aid an incumbent leader’s popularity (Aytaç, 2017). For instance, South Korean President Kim Dae-jung’s high approval rating in the wake of the Asian financial crisis (Leem, 2001: 54) can be explained by the country’s economic improvement as reflected in the declining unemployment rate, which fell from 6.96% in 1998 to 6.34% in 1999 and 4.4% in 2000 (World Bank, 2020). Before the global pandemic, former US President Donald Trump gained strong domestic support because the economy was better, in terms of unemployment and stock market performance, than during the second term of the Obama administra- tion (BBC News, 2019). As illustrated in Figure 1, the present study explores economic problems that cause political and social unrest, and it explores the causal path linking an economic slow- down that challenges incumbent leadership and military conflict. This study does not insist that other illnesses <<FIGURE OMITTED>> and troubles have little or no effect on leaders’ diversion- ary incentives. Instead, it assumes that domestic chal- lenges to political leaders result from a variety of political, social, and military situations. Given the long debate on the relationship between domestic problems and foreign aggression as discussed above, this article tests the hypothesis that economic slowdowns tend to cause states to initiate militarized aggression in a systema- tic and rigorous way (H1). H1: A state is more likely to initiate a military conflict when it experiences a higher degree of economic slowdown. Economic dependence and attractive targets Even if economically troubled states tend to be aggressive, unpopular leaders are not necessarily reckless and will not target other states indiscriminately. They can be prudent in choosing their military targets to restore their political influence. These leaders hope to avoid wars with high costs and unsuccessful outcomes. They need rally effects, not hard-won victories, to protect and prolong their political lives. Therefore, diversionary incentives should generate small-scale conflicts in terms of intensity and duration (Weisiger, 2013). Because long and bloody wars are not popular, leaders search for non-aggressive targets that are unlikely to lead to serious reprisals (Jung, 2014b). For example, since the first North Korean nuclear crisis, some experts have expressed the view that Pyongyang does not want a large-scale war but repeated crises that provide material benefits through the resulting negotiations with Washington and Seoul, which create a siege mentality among the North Korean elites and pub- lic. Kim Jong Un knows that a war with the United States would be a catastrophe, but also that provocations short of war may well be advantageous. How do political leaders choose diversionary targets and gain domestic support through popular conflicts? This analysis highlights a potential target’s (PT) eco- nomic dependence on a potential initiator (PI) among others, which may affect political leaders’ decisions regarding diversionary conflict. Economic interdepen- dence has been a prominent topic in international con- flict studies, with some suggesting that it has pacifying effects. Liberal scholars have argued that ‘trading states’ that were once ‘territorial’ now make money, not war (Angell, 2012; Rosecrance, 1986). In a globalized econ- omy, states see one another as business partners and believe that trade and investment with these partners will increase prosperity and stability, at least in the long term. Although one-time defection may provide larger short- term benefits, repeated games encourage interstate cooperation through linkage politics and international organizations (Axelrod & Keohane, 1985). According to some studies, democratic peace (i.e. the absence of war between democracies) can be attributed not to the normative and institutional characteristics of democracies but to the economic interdependence among democracies, most of which have capitalist economies (Gartzke, 2007; Mousseau, 2002). However, some scholars have questioned the ability of economic interdependence to yield interstate peace and stability (e.g. Barbieri, 2002; Copeland, 2015; Grieco, 1988). Most if not all interdependent relations are asym- metric. When two states become economic partners, one relies more on the other because of the former’s smaller economy and/or limited access to markets, resources, and products. As a result, the senior partner has leverage over the junior partner, which, in turn, is always suspi- cious of the senior. This is why realists such as Waltz (1970) argue that economic interdependence is a cause of international conflict rather than of international peace. Stronger economic relations can stoke mistrust and provide more reasons for disputes. Indeed, wars such as World War I and the Pacific War turned partners in trade into enemies. This article suggests that leaders of economically troubled states will prefer vulnerable economic partners as diversionary targets. Due to its economic dependence, the junior partner can be expected, at least by the leader facing a domestic challenge, to respond timidly to mili- tary aggression. The lack of harsh retaliation by the junior partner will enable the aggressor’s leader to benefit from rally effects without embarking on a protracted, costly conflict. If an economically troubled state experi- enced a harsh response to an act of aggression, the aggres- sor’s leader would face domestic criticism over the high cost of the military action. Targeting an economically vulnerable state can be less risky because the target is less committed to and has less capacity for escalation of a crisis. Given that leaders facing an economic slowdown prefer limited conflicts to all-out war, the present study hypothesizes that a PT’s economic dependence on a PI increases the possibility of the PI’s diversionary action against the PT (H2). Some may ask why certain economically dependent states do not respond to foreign aggression. Because of their dependence on other economies, their leaders may have high incentives to divert domestic attention to for- eign affairs in politically or economically challenging times. However, this analysis does not make assumptions about the PT’s economic or political situation, but only about its level of economic dependence on the PI. Lit- erature on the diversionary use of force has examined the effects on a PI’s use of force of aspects of the PT’s domestic situation such as unemployment and inflation (Clark, 2003; Clark, Fordham & Nordstrom, 2011), presidential popularity (Davies, 2007), and legislative opposition (Foster, 2006). Chiozza & Goemans (2004) argue that leaders at a higher risk of losing office are less likely to be military targets, based on a statistical analysis of global leaders from 1919 through 1999. This study shifts our attention to another feature of a potential target, namely economic dependence. H2: A state is more likely to initiate a military conflict against another state when the former’s economic slowdown is greater and the latter’s economic dependence on the former is stronger. How do economically dependent states respond to aggression by states with slowing economies? If economically troubled states expect a harsh response from their PTs, they may choose repression and reform rather than foreign diversion. The present study hypothesizes that economically dependent and politically democratic states encourage diversionary action. Democratic leaders should be accountable to their citizens and interest groups, whose main concern is economic interests and prosperity. As McGillivray & Smith (2008) demon- strate, democratic leaders maintain cooperation with for- eign partners without risking harm to the long-term interests of societies that have the power to replace them- selves. Most business sectors prefer peace over war with respect to their country’s relationships with major eco- nomic partners, although companies that are less com- petitive in the global market or wish to protect their technology and intellectual property sometimes press for hardline policies like economic sanctions. Often, mili- tary conflict with senior economic partners creates domestic discontent and social unrest for junior partners. This is why easily replaceable leaders appear to be non- aggressive targets that invite diversionary action, but not large-scale conflict (Jung, 2014b). However, democratic leaders can become attractive targets for autocracies, though not for fellow democra- cies. According to Oneal & Russett (2001) and Oneal & Tir (2006), among others, democracies generally view each other as military and economic partners rather than military targets. Even in times of economic downturn, democratic leaders are unlikely to target other democra- cies, mainly because such a decision would rarely gain popular support in liberal societies, which vehemently espouse democratic ideas and goals in international soci- ety such as the promotion of democracy and human rights protection. As shown in the Iraq War, democratic leaders undertaking aggressive military action prefer to claim that their actions are necessary to save ordinary people living under dictators and brutal regimes. Dem- ocratic leaders can hardly legitimize a decision to initiate conflict with another democracy, even if it is short-lived, managed, and beneficial. To secure domestic and inter- national support, democratic leaders of slowing econo- mies may either choose non-democratic targets or opt for domestic solutions. Thus, this article’s final hypothesis is that an autocracy with a slowing economy is more likely to initiate a conflict with an economically dependent and politically dissimilar state. H3: A state is more likely to initiate a military conflict against another when the former’s economic slow- down is greater and the latter’s economic depen- dence on the former is stronger, if the potential initiator is an autocracy and the potential target is a democracy. Even rising powers cannot avoid domestic disgruntle- ment due to economic slowdowns. Growing economies experience forms of economic stagnation such as the middle-income trap at times and cannot achieve high growth rates forever (Aiyar et al., 2013). Even political leaders in states with remarkable records of economic growth cannot avoid domestic problems, because rapid growth cannot solve all their internal problems and sometimes accelerates social divisions. However, when a slowing economy derails public expectations, the blame tends to be directed toward the government and racial, ethnic, or other groups. In this situation, as Organski (1968: 373) observed, political leaders ‘are led to provocative statements and actions toward other nations in an effort to distract attention from internal difficulties and fix the responsibility for internal trou- bles on the outside’. Research design To test these hypotheses, this study examines directed dyads from 1950 to 2010. The unit of analysis is a directed dyad-year. The dependent variable, conflict initiation (Initiation), is measured by coding whether a PI initiated a hostile militarized interstate dispute (hos- tility level 1⁄4 4 or 5; ‘use of force’ or ‘war’) against a PT in each year. Gibler, Miller & Little’s (2016) Militarized Interstate Disputes (MID) dataset, a modified version of the Correlates of War (COW) Project’s MID dataset, is used for this measurement. Due to concerns about reverse causality, there is a one-year lag between the dependent variable (year t) and the independent variables (year [t–1]). The independent variable of GDP growth rate (Growth) is measured by averaging a PI’s annual GDP growth rates over the previous three years (i.e. years [t–3], [t–2], and [t–1]). A lower value represents slow economic growth, whereas a higher value shows rapid economic growth. As a robustness check, the growth variable is alternatively measured by averaging annual GDP growth rates (relative to all other states) over the previous two or five years. The GDP data are available at the Maddison Project Database v2018 (Bolt et al., 2018). The intervening variable, economic dependence (Dependence), is measured by dividing bilateral trade volume by a PT’s GDP in a given year. A higher value means that a PT has a larger economic dependence on a PI. For measurement purposes, this study relies on the economic variables of the Historical Bilateral Trade and Gravity Dataset (TRADHIST) (Fouquin & Hugot, 2016), which contains both bilateral trade and GDP data with a large temporal and geographical coverage. Six control variables are added to the model to exam- ine the net effects of the independent variables and their interactions. First, power parity (PowerParity) between a PI and a PT is a natural log of 1 plus a smaller Composite Index of National Capabilities (CINC) score divided by a larger CINC score. The CINC scores provided by the COW’s National Material Capabilities v5.0 (Singer, Bremer & Stuckey, 1972) provide each state’s relative power in a given year. It is expected that the more power parity there is, the more likely a PI is to initiate a conflict (Bremer, 1992). Second, an alliance (Ally) between a PI and a PT is coded as 1 if they had a defense pact in a given year, based on the Formal Interstate Alliance Dataset v4.1 (Gibler, 2009). Since common threats produce alliances (Walt, 1987), an alliance relationship has a negative impact on conflict initiation. Third, the democracy variable (JointDemocracy) is coded as 1 if both states in a dyad are democracies and 0 otherwise, using the polity scores available in the Polity IV Dataset (Marshall, Gurr & Jaggers, 2019). A state is coded as a democracy if its polity score is greater than or equal to 6, as an autocracy if its polity is less than or equal to –6, and as mixed if its polity is between 6 and –6. It is widely known that democracies cooperate with each other more than other types of dyads, even in the pres- ence of domestic challenges (Oneal & Tir, 2006). Fourth, the rivalry variable (Rivalry) is coded as 1 if the two states are rivals and 0 otherwise, using the rivalry dataset provided by Thompson and Dreyer (2012). Stra- tegic and territorial rivals are known to participate in military conflicts more frequently, especially when faced with domestic problems (Mitchell & Prins, 2004; Tir, 2010). Fifth, whether a PI shares a border with a PT (Border) is measured with the COW’s contiguity dataset (Stinnett et al., 2002). A state is more likely to target a neighboring state, especially if a pre-existing territorial dispute exists. Sixth, the distance variable (lnDistance) is the natural log of the distance between the two states’ capitals, which is computed by NewGene software (http://www.newge nesoftware.org/home/4587292843) using a method from EuGene software (Bennett & Stam, 2000). States located closer to each other are more likely to be involved in a conflict because of territorial issues and convenient power projection. Last, to control time dependence in a cross-national time-series dataset, the variable of peace years (PeaceYears) is added. It represents the number of years since the last MID, and its square and cube are also included (Carter & Signorino, 2010). <<TABLE OMITTED>> Results and discussion In general, the logit analysis of the binary dependent variable provides strong support for the first hypothesis that an economic slowdown contributes positively to conflict initiation (H1), but little support for the second hypothesis that economically troubled states are more likely to target economically dependent states (H2). Model 1 (see Table I), which tests whether economic growth rate affects the initiation of military conflict, finds negative and significant effects at the 0.01 level. However, Model 2 finds that the interaction between growth and dependence (Growth(t–1) 􏰁 Dependence(t–1)) and the variable of dependence (Dependence(t–1)) have nonsignificant effects, while the growth variable (Growth(t–1)) has stronger effects on hostile conflict initia- tion. In both models, the control variables – power parity, rivalry, border, and distance – show the expected effects at the 0.01 significance level, while the variables of joint democracy and alliance have non-significant effects. Next, to test whether the regime type impacts the effects of the combination of growth and dependence (H3), Model 3 restricts their cases to the dyads in which PIs are autocracies and PTs are democracies. In this model, the interaction term has negative effects at the 0.01 level, and its constitutive terms – Growth(t–1) and Dependence(t–1) – have significant effects at the 0.01 or 0.05 level. This suggests that economically troubled autocracies tend to target states that are economically dependent and politically democratic. According to the models’ log likelihood and scores of AIC and BIC, Model 3 has the most explanatory power, even after considering the number of explanatory variables. In addition, to see whether the effects of growth, depen- dence, and their interaction hold for cases of any level of conflict, rather than only high-level conflict, Models 4 and 5 change the dependent variable from hostile con- flict initiation to conflict initiation and yield the results similar to those from Models 1 and 3. To understand the interactive effects of growth rate and economic dependence, marginal effects and pre- dicted probabilities based on Model 3 were calculated and illustrated.1 Following Berry, Golder & Milton’s (2012) recommendations, here I illustrate the marginal effects of each of the two constitutive terms of the inter- action term (Growth and Dependence) across the range of the other term while controlling other variables in accor- dance with a dyad of China (PI) and Japan (PT) in 2010.2 Figure 2(a) shows that marginal effects of a PT’s economic dependence decline as a PI’s economic growth increases from –20% to 20%. The dependence’s positive marginal effects are statistically significant only when the growth is slow (Growth(t–1) <1⁄4 0 [11.34 percentile]). The marginal effects of economic growth also decrease as the level of dependence increases (Figure 2(b)). The marginal effects of the economic growth are negative and statistically significant in the most range of economic dependence (Dependence(t–1) <1⁄4 0.05 [99.12 percen- tile]). In sum, these two graphs show that when a PI’s economy slows down, a PT’s dependence has positive effects on hostile conflict initiation. When a PT’s eco- nomic dependence on a PI is not so extreme, the PI’s economic growth has negative effects. Also, the probability of an autocracy’s (PI) initiation of hostile conflict against a democracy (PT) across the range of a PI’s growth and a PT’s dependence are pre- dicted. By controlling other variables of power balance, ally, rivalry, border, distance, and peace years, we can determine how much effect the growth rate and eco- nomic dependence have on the probability of conflict initiation. As illustrated in Figure 3(a), the probability of conflict initiation against a potential target whose economic dependence is .02 decreases from .972 to .001 when a PI’s growth rate increases from –20% to 20%. Figure 3(b) shows the increased probability of conflict initiation by a PI whose economic growth is 0% from .231 to .971 as a PT’s economic dependence changes from 0 to .15. As a robustness check, the measures of growth, depen- dence, and democracy are changed. First, economic growth is replaced by political unrest. Many studies of diversionary conflict have measured domestic unrest based on the number of dramatic political events such as military coups, political crises, and mass demonstra- tions. Relying on the Cross-National Time Series (CNTS) Dataset (Banks & Wilson, 2020), the variable of political unrest (Unrest(t–1)) for this study is the natural log of 1 plus the number of ‘major government crises’ and ‘purges’ in a given year. Like the economic growth variable, political unrest has significant effects on conflict initiation (Models 6, 7, and 8 in Table II). However, the interaction between political unrest and economic dependence has non-significant effects in Models 7 and 8, while the economic dependence loses significance in Model 8. This result suggests that politically troubled states tend to be aggressors, irrespective of their regime type, but that their PTs’ economic dependence has little effect on the military action. Second, a PT’s dependence is measured by its depen- dence on the global economy. When a state is more involved in global trade and foreign investment, it is less likely to respond harshly due to the likely cost to its national economy and political leadership. Even for a state whose dependence on an economically troubled state is not great, the former can be seen as a non- aggressive target by the latter’s leaders. In this case, a globalized economy may invite aggression from the slowing economy by initiating a low-scale conflict for rally effects. The variable of global dependence (Global- Dependence(t–1)) is measured by a PT’s annual trade volume divided by its GDP in a given year. The results of Models 9 and 10 show that the variable of global dependence has significant but negative effect on con- flict initiation. The significance of the interaction term between growth and global dependence disappears in the model with all dyads (Model 9), as well as in the model with dyads of autocracies (PIs) and democracies (PTs) (Model 10). Those states that are closely inte- grated into the global economy tend to avoid rather than to provoke conflict. Third, PI and PT regime types (democracy, mixed, and autocracy) are measured by the polity score’s different thresholds (þ7 and –7) and by V-Dem’s liberal democ- racy scores (Coppedge et al., 2021; see the Online appen- dix, Table A1). When using the liberal democracy index, this study uses two thresholds (.11 [33 percentile] and .41 [66 percentile]) to classify states into three categories: autocracy, mixed, and democracy. Models 11 and 12 show that the significant effects on conflict initiation of growth and its interaction with dependence remain con- stant independent of the measurements of regime types. Fourth, this study measures the variable of economic growth rate in different ways. First, it averages annual growth rates over the previous five or two years rather than over the previous three years (Models 13 and 14 of Table A2 in the Online appendix). Next, it averages the previous three-year annual growth rates relative to the global average or global median (Models 15 and 16 of Table A2 in the Online appendix). The results of Models 13 and 14 are similar to those of Model 3, whereas a PI’s relative growth rate measurements in Models 15 and 16 show positive or insignificant effects of economic growth. Finally, this study examines the Cold War and post- Cold War periods separately (Models 17 and 18 of Table A3 in the Online appendix), and it restricts cases to geographically close or major dyads, each of which is composed of two nearby states or at least one major power (Model 19 of Table A3 in the Online appendix). Economic growth rates have had negative effects in the Cold War and post-Cold War periods. However, after 1990, economic dependence loses its significant effects and its interaction with growth has positive, not nega- tive, effects on conflict initiation. Future research should take this difference into account and compare regions in which states interact at different levels of economic dependence and political confrontation. When the set of cases changes from all dyads to geographically close or major dyads, the increased explanatory power is found from the AIC and BIC scores of Models 3 and 19, while the economic dependence variable loses its significance. In sum, this statistical analysis shows that an eco- nomic slowdown increases the likelihood of initiating military conflict. Specifically, economically troubled autocracies act more aggressively toward democracies whose economies depend on them. Political unrest also increases conflict initiation, as many studies of diversion- ary conflict have argued. However, autocracies suffering from political troubles do not seem to target economi- cally dependent democracies. Democracies with smaller economies seem to be attractive targets for economically troubled autocracies whose leaders may blame the polit- ically different partner’s ‘unfair’ trade and currency prac- tices and domestic policies for the deteriorating economy. The autocracies will present their military decisions as inevitable to correct international wrongs and reboot the national economy. Meanwhile, leaders facing political unrest may seek diversionary targets, initiating a conflict through which they can mobilize their support base (and marginalize opposition groups) by exploiting ideological, religious, and/or ethnic iden- tities of ‘us’ and ‘them’. Future studies should seek to unravel the complex relationship between types and lev- els of domestic unrest and types and combinations of political regimes, in either the presence or absence of diversionary conflict. The findings of this study suggest that an attractive target is not predetermined but is closely related to the domestic and international situa- tion confronted by leaders who are battling for their political survival.

<<PARAGRAPH BREAKS CONTINUE>>

Summary and implications

Since the 2000s, many pundits and scholars have attrib- uted two rising powers’ instances of aggression to their domestic problems. Russia’s invasion of Ukraine and annexation of Crimea and its deployment of troops to Syria, along with China’s assertive actions in the South China Sea, have been seen as aggressions arising out of domestic weakness (Kaplan, 2016). Some have attributed China’s recent provocative actions against India, Japan, and Taiwan to its struggling economy in the wake of the pandemic (Myers, 2020). In the post-Cold War period, these two great powers, often regarded as revisio- nist powers, have regained their national capabilities to some extent and have sought higher status in interna- tional politics (Larson & Shevchenko, 2010). What made them more aggressive in the 2010s? One promising explanation is their decreasing economic growth rates. Due to their abundance of natural and human resources, the two former communist countries had maintained their increased status relative to their neighboring states since the 2000s or even earlier. However, falling energy prices and the reform-incapable structures of their economies [undermined] retarded their further growth and increased the burden on their authoritarian leaders, who also had to manage domestic calls for polit- ical freedom and maintain the legitimacy of their rule. According to this article’s causal explanation, Russia and China became more revisionist actors in the 2010s at least partly because their economies went downhill, and their aggressions were directed at smaller and more vul- nerable economies.

A slowing economy and economic dependence have both individual and interactive effects on military con- flict in international politics. Slow growth rates contribute positively to conflict initiation, especially when an autocracy suffering from a slowing economy targets a democracy whose economy is dependent on the auto- cracy. Whereas economic interdependence has been regarded by many liberal scholars as a reason for inter- national cooperation, this study shows why and how asymmetric interdependence contributes to interstate conflict by articulating its interaction with an economic slowdown in a dyadic relationship. Both political unrest and economic problems have destabilizing effects on international politics, but their effects differ when eco- nomically dependent democracies are the target. Future research should compare different types of domestic unrest and their effects on international relations with sophisticated logic and strong evidence.

This analysis shows that economic growth has two distinct effects. Slower growth makes one state weaker than others and its leaders more vulnerable to domestic challenges. A state’s economic slowdown not only causes a change in the balance of economic power so as to favor its competitor but also increases the troubled state’s foreign aggression, which can harm its adversary. Therefore, it is uncertain whether one state’s economic slowdown benefits its rival’s security. An autocracy’s slowing econ- omy can be a warning, rather than good news, to its democratic opponents if the latter maintain economic relationships with the former. In this regard, economic sanctions may provoke the target leader’s diversionary tactic of blaming its poor economy on the sanctioner. We can see this pattern in Tokyo’s surprise attack on Pearl Harbor at a time of US sanctions against Japan, which had continued since the late 1930s.

The current global economic outlook is uncertain in the wake of COVID-19. Although we do not yet know the pandemic’s long-term economic consequences, all states will likely struggle with slower growth and the erosion of democracy and will engage in more protec- tionist policies to reboot their economies and save major industries. In this context, we should be especially concerned about the risk that troubled economies pose to international security. As this study’s findings suggest, slower growth causes domestic problems, which often lead to the diversionary use of force against foreign states. If we cannot prevent struggling leaders from using diversionary tactics, worsening economies in authoritarian states will likely drive such states into a more confrontational or aggressive stance toward their economic part- ners, if not toward political and ideological partners. We can all hope world leaders will realize that this is a time to beat a global pandemic and an economic recession, not to beat up on a foreign state.