The Origins of Major War ilensified arms racing and crisis initiation, but only as policies sucn thev recognize the risks of provoking war by an inadvertent escalation. The mode! thus helps to predict shifts in the probability of major war even when leaders are loath to fall into major war, as they are in the nuclear age. The empirical tests of the book's argument are to be found in chapters 3 through 8. In chapter 9,1 conclude by examining the theoretical and practical implications of the argument.8 Rethinking Realist Theories of Major War ■ The three most prominent realist explanations for major war among great powers, as I noted earlier, are classical realism, structural neorealism, and hegemonic stability theory.' Classical realism argues that major war is likely when one state is preponderant and unlikely when great powers are relatively equal. A balance of power keeps the peace by convincing potential aggressors that war will have both high costs and a low probability of success. An imbalance provides the key condition for major war, since the superior state is more likely to expand in the belief that war can pay.2 As for this superior state's motives, classical realists would agree with Hans Mor-genthau that the preponderant state initiates war for unit-level reasons— for greed, glory, or what Morgenthau saw as its "lust for power" manifested in "nationalistic universalism."3 Given the propensity of superior states to attack, classical realists argue that multipolarity should be relatively more stable than bipolarity. Since exact equality cannot always be ensured, alliance restructuring (external balancing) in multipolarity can create the requisite balance of power between blocs, even if individual states are unequal. As long as flexibility is maintained, such that great powers can easily shift alliance ties in response to a stronger power, preponderant states can be deterred from aggression. Conversely, bipolar systems are prone to be unstable since if any inequality between the two great powers opens up, no large alliance partners exist to forge an effective balance of power.4 Classical realism's strength is its emphasis on power differentials, which provides a fine-grained sense of the relative weights that go onto the scales of the balance of power.5 It also highlights a flaw in theories that argue for the stability of bipolar systems: states in multipolarity have another mechanism—alliances—in addition to arms racing to help deter an aggressor [10] [11] The Origins of Major War when rough equality cannot be maintained, in bipolarity, great powers have no viable alliance option to fall back on when arms racing alone is not enough. Classical realism confronts two main empirical problems, however. First, it cannot explain how multipolar systems with tight alliances against the potential hegemon, such as the one that existed before 1914, can still fall into major war. In such cases, the overall balance of power between the blocs fails to keep the peace.6 Second, in the key bipolar cases in history—Sparta-Athens, Carthage-Rome, and France-Hapsburgs—war ensued when the two great powers were essentially equal. Here, the balance of power between the individual great powers did not deter war. The theoretical problem behind these empirical anomalies lies with the largely static nature of classical realism. The theory derives predictions primarily from snapshots of the international systems. The result is the familiar picture of great powers as billiard balls of varying sizes.7 The importance of dynamic trends in the differentials of power is understated in such an analysis. Some classical realists such as Morgenthau recognize in passing that preventive wars—wars for fear of decline—are a significant problem in history. Yet the conditions under which preventive war motivations are invoked remain theoretically undeveloped.8 The second approach, neorealism, focuses on two enduring structural features of the international system, anarchy and polarity. Anarchy—the absence of a central authority to protect the great powers—produces the recurring pattern of conflict seen in international politics over the millennia.9 Across anarchic realms, neorealists assert that bipolar systems are less likely to experience major war than multipolar systems. Three main reasons are given: in bipolarity, great powers avoid being chain-ganged into major war by crises over small powers; they also stand firm, however, to prevent losses on the periphery, thus enhancing deterrence; and finally, the great powers are less inclined to neglect internal military spending that might allow a superior military power to arise.10 Neorealism's strength is its isolation of the structural effects of anarchy and polarity. This leaves us with a profoundly tragic view of international relations: even when states only seek security, they may still fall into devastating wars that threaten their survival." In its Waltzian form, however, neorealism suffers from the same deficiency as classical realism: it is not dynamic enough. With polarity as the key structural variable, there is nothing to vary within either a bipolar or multipolar system to explain why any system should move from peace to war.12 As Waltz explains, "within a system, [systemic] theory explains recurrences and repetitions, not change." If changes in state behavior "occur within a system that endures, their causes arc found at the unit level."1'1 This is an unnecessarily limiting view of the explanatory power of a systemic theory. As I explore below, changes and [12] (Rethinking Realist Theories of Major War trends in the differentials of relative power between states-—a systemic variable going beyond the mere number of great powers—can have marked effects on behavior without necessary consideration of unit-level i' changes. jr Other neorealists have sought to go beyond Waltz by incorporating the classical realist point that power inequalities increase the likelihood of major war.14 Still others emphasize that states face a security dilemma, whereby the actions one state takes to enhance its security end up reducing the security of its adversaries. As a result, states sometimes have preventive i incentives to eliminate a growing adversary before it becomes too strong.15 I build these additional elements into my synthetic argument. Since 1 examine the effects of power trends in bipolar versus multipolar systems, however, I reach conclusions different from those of neorealism on the conditions for war in the two system-types. Moreover, chapter 2 moves beyond existing structural arguments by fusing within a leader's decision-making logic both the risks of decline and the risks of an inadvertent spiral to major war. Finally, by considering different forms of decline, including power os- 1 dilations and decline driven by inferiority in economic and potential power, I provide additional conditions constraining the rational response to decline. The third realist perspective on major war is the security variant of hegemonic stability theory,16 represented most prominently by A. F. K. Organski and Robert Gilpin. Turning classical realism on its head, they argue that a hegemonic system with one powerful actor will be stable because of the hege-mon's self-interest in maintaining the political-military order. When a second-ranked state rises to near equality with this now former hegemon, however, this ascending state is inclined to initiate war to receive the status and rewards denied by the traditional system.17 Hence, contrary to the classical realist view that a balance of power keeps the peace, major war is the result of a growing equality of power between the two most powerful states in any system. The strength of this approach lies in its dynamic nature. Hegemonic stability theory thus provides a more extended analysis of the impact of power changes on great power behavior than is offered in either classical : realism or neorealism. Two main problems remain, however. First, hege- monic stability theory has no deductively consistent theory of war initiation. There is no logical reason why a state should attack while it is still rising, since by simply waiting, the state will be able to achieve its objectives i more easily, and at less cost.18 This argument holds even if rising states have goals other than security, such as status and prestige, as hegemonic stabil-I ity theorists assume they do. Waiting until the state has maximized its I power ensures the maximum return on its war investment. After all, even t more status and rewards are obtained by fighting when one stands the best I chance of winning quickly and at low cost.10 Hegemonic stability theory mus cannot expiam wny German leaders in both World Wars did not initiate war until after they saw that Germany was declining. Moreover, it cannot account for the fact that in the seven other major wars prior to 1900, it was the declining great power that brought on the hostilities. A similar pattern is seen in each of the key crises of the early cold war: it was the state foreseeing decline that initiated the dangerous crisis period. The second limitation is that hegemonic stability theory's core logic for major war is confined to the two most powerful states in any system—the leading state and the rising challenger. The theory thus minimizes the importance of third-, fourth-, and fifth-ranked great powers on the calculations of the other two.20 This might make some sense in a bipolar system, as I show, but it makes little sense for the multipolar cases of European history from 1556 on. Empirically, for example, the theory has a hard time explaining how one state in each of the two major wars of this century—Germany—was able to take on a coalition of second-, third-, and fourth-ranked great powers, fight a long war, and nearly emerge victoriously, if indeed Germany was only equal to the formerly dominant state in military power.21 Interestingly, the evidence provided by hegemonic stability theorists confirms that Germany was in fact preponderant when it took on the system. Organski and Jacek Kugler conclude that by 1913, "Germany [had] clearly surpassed the United Kingdom," the formerly dominant state, while by 1939, Germany had a "significant advantage" over Britain.22 Kugler and William Domke, to explain how Germany could have come so close to winning both wars, show that Germany in 1914 and 1939-40 was significantly superior in actualized military power. In 1914, Germany was almost as powerful as Britain, Russia, and France combined. In 1939-40, Germany was almost twice as strong as France and Britain combined; in 1941-42, it matched the Soviet Union on the eastern front even as it continued to wage war in the west.23 To accommodate these facts, hegemonic stability theorists adjust the theory: they argue that although equality between individual great powers may not be associated with major war, relative equality between their alliance blocs is. Organski and Kugler conclude: "it is clear that [the World Wars] occur after the intersection when the two nations fight alone (which is contrary to what the power-transition model leads us to expect), but before the coalition of the challenger overtakes the coalition of the dominant country."2' Woosang Kim, in an important statistical reworking of Organ-ski's argument, shows that major wars occur at points of essential equality only when power is adjusted to incorporate alliance partners.25 This reformulation still allows hegemonic stability theorists to challenge classical realism: as noted, classical realists cannot explain why in cases like World War 1 war occurred despite the relative equality between two tight alliances. Overall, however, the primary challenge ot hegemonic stability theory has been dissipated. Classical realists and hegemonic stability theorists now essentially agree that in 1914 and 1939 one state—Germany—was significantly superior to any other individual state, even if Germany (along with minor partners) was opposed by a coalition of equal strength. Military historians, as I discuss in chapters 3 through 5, would agree.2' The agreement between classical realists and hegemonic stability realists on German military superiority in the twentieth century simplifies the task ahead. Yet we still lack a theory that can explain, without invoking ad hoc unit-level factors like "lusting for power" and "dissatisfaction with the status quo," why preponderant states in multipolarity attack the system in the face of the staggering risks and costs. Moreover, how the pressures to initiate major war change between multipolar and bipolar systems is still un-derspecified. Providing a comprehensive systemic theory of major war, one that synthesizes the strengths of current realist approaches, is the objective of the rest of this chapter. Dynamic Differentials Theory The core causal or independent variable of the argument is the dynamic differential: the simultaneous interaction of the differentials of relative military power between great powers and the expected trend of those differentials, distinguishing between the effects of power changes in bipolar-ity versus multipolarity.27 In addition, I break the notion of power into three types—military, economic, and potential—to show how decline in the latter two forms affects the behavior of states that may be superior in military power. The theory makes three main assertions. First, in any system, assuming states are rational security-seeking actors which remain uncertain about others' future intentions,28 it is the dominant but declining military great power that is most likely to begin a major war. Second, the constraints on the dominant state differ in bipolar versus multipolar systems. In multipolarity, major war is likely only if the declining state has a significant level of military superiority. In bipolarity, however, the declining state can attack even when only roughly equal, and sometimes even if it is second-ranked. Third, the probability of major war increases when decline is seen as both deep and inevitable. A consideration of overall economic power and potential power is thus necessary, since the levels and trends of these two other forms of power are crucial in determining the extent and inevitability of military decline. The first proposition is relatively straightforward: because major wars are so costly, and because they risk the very survival of the state, the initia- The Origins of Major War Rethinking Realist Theories of Major War for of war is more likely to be the dominant military power; smaller military powers simply lack the capability to "take on the system." Moreover, it is irrational for any great power to begin a major war while still rising, since, as noted, waiting allows it to attack later with a higher probability of success, and at less cost. All major wars, if actors meet the requirement of rationality, therefore must be preventive wars.29 The second proposition requires more explication. To state it slightly differently, while near equality between individual great powers is likely to be stabilizing in multipolarity even when some states are declining, near equality in bipolarity can be very unstable when either of the great powers, but especially the dominant power, perceives itself to be declining. Thus, the conditions for major war in multipolarity are less permissive than those in bipolarity, meaning that for any given set of power differentials and trends, war is less likely in multipolar systems.30 The logic behind this assertion is as follows. In multipolar systems, if all states are relatively equal in military power, no state will make a bid for hegemony against the system, for four main reasons. First, even if a state expects the others to remain disunited—that is, even if it does not expect a counter-coalition to form against it—equality with its rivals will likely mean long and costly bilateral wars, wars that will sap the state's ability to continue the fight until hegemony is achieved. If complete hegemony is not achieved, those states sitting on the sidelines will emerge in a stronger position relative to the state that initiates war. Hence launching all-out war in the first place is irrational.31 Second, to the extent that a coalition does form against the challenger, there is even less probability that the initiator could emerge in a stronger and more secure position after the war. Coalitions in multipolarity, since they are made up of states with "great power," become formidable fighting forces as their unity increases.32 The third reason follows from the other two. A declining but only equal great power in multipolarity has reason to think that a rising state, as long as it does not grow too preponderant, will also be restrained in its ambitions simply by the presence of so many other great powers. Therefore, a preventive war for security is less imperative. Fourth, to the extent that an equal but declining power can form alliances against the state that is rising, it will have less concern about being overtaken. This restates classical realism's insight that states in multipolarity, compared to bipolarity, have recourse to an additional means to uphold their security besides internal balancing, namely, external balancing through alliances. Because of the collective action problem that may be present, however, my deductive logic as to whv an equal but declining state does not initiate war in multipolarity does not depend on this state's ability to form a tight alliance for its security (although such alliances certainly reinforce the argument). Rather, the argument revolves around the [16] I' state's recognition that even it no alliances form against it il it begins a mall jor war, it will not have enough power to win a victory against all the oth-m. ers; and even if no alliances form with this state if it chooses to "decline m gracefully," the presence of many actors should help deter the rising state jK;. from attacking later. Consequently, in multipolarity, only when a given m state is clearly superior to any other individual state in military power can ~m it contemplate waging a war for hegemony. B In bipolar systems, however, these arguments push in the opposite direc- § tion, and therefore preventive war is likely even when states are near equals, if First, a declining and near-equal state realizes that it has to face only one 1 other great power, not many, and therefore even if the war is long and diffi- S cult, there are no additional opponents to defeat after the bilateral victory is 1 achieved. A successful bid for hegemony is thus easier to achieve. Moreover, 1, even if the declining state fears a stalemated and inconclusive war with the ft rising state, it does not have to worry about a relative loss to third party ac- m tors that sit on the sidelines to avoid the costs of war. Such actors, since small, m are unlikely to gain enough to raise themselves to the top of the system. ■ Second, the declining state knows that even if a coalition forms against S its attack, the small states joining the rising great power are unlikely to alter M the expected outcome significantly. In comparison with multipolarity, indi- ■ vidual coalitional partners simply have far less weight to throw against the m initiator of major war.33 Third, because the declining state realizes these two m factors are in its favor when it is slightly superior, it knows that the rising ■ state will not be terribly constrained after it achieves superiority. Fourth, X the declining state knows that the other states in the system, even if some M are willing to ally against the ascending state, are not substantial enough to m shore up its waning security. Hence preventive war before the point of S overtaking makes rational sense. X Note that because of the absence of significant third parties, even the B second-ranked state in bipolarity can initiate major war when in steep deli cline. The core logic applies: it has to beat only one other great power, and ■ there is little concern about stalemated wars that allow sideline-sitters to S rise to the top.34 Of course, the greater the second-ranked state's level of in-B feriority, the less confidence it will have in a hard-line policy. H< The argument I have outlined is summarized visually in figures 1 and 2. » These heuristic diagrams present the main systemic situations that might mt- be faced in either multipolarity or bipolarity.35 Note that at times tv f4, and jB t5, the probability of major war should be low for both system-types, since X the trends in the military balance are stable; with no state experiencing de-cline, there is no imperative to go to war for security reasons.36 At time f,, m however, the impending decline of the dominant state in the bipolar situa- H tion (fig. 2) means the likelihood of major war is high, while in tire multipo- m. lar situation the likelihood is low because of the restraining presence of the ll l2 £3 U l5 l6 S l8 Relative military power Probability of major war Time Fig. l. Relative military power curves and the probability of major war: multipolarity ll l2 S U S l6 l7 l8 Relative military power Probability of major war low high moderate low low high high high Time Fig. 2. Relative military power curves and the probability of major war: bipolarity The Origins of Major War other equal great powers.37 At times th and t-, when there is marked inequality in both bipolar and multipolar systems, impending decline should make major war highly likely in both system-types. At f8, however, while the probability of war is again high in the bipolar case, instability in the multipolar case should be tempered somewhat by the existence of the third-, fourth-, and fifth-ranked powers (although since these latter powers are weaker than at time t2, the probability of major war is still "moderate")-38 In both multipolarity and bipolarity, it is a declining state that initiates war. When it does so depends greatly on its estimation of the inevitability and the extent of its fall: the higher the expectation of an inevitable and deep decline, the more the state will be inclined to preventive war simply for security reasons. If decline is caused by entrenched stagnation relative to the rising state, this will certainly be worrisome; the fewer internal measures available to overcome the stagnation, the more the state will see decline as deep and inevitable. Of even greater significance for the declining state's calculus is its level of economic power and its overall potential power compared to its military power.39 A state in either bipolarity or multipolarity that is superior but declining in military power, but also superior and growing in the other two power dimensions, is unlikely to be that anxious about decline. After all, given that its economic and potential power is strong and ascending, this state should be able to reverse the downward military trend simply by spending more on arms in the future. A state, however, that is superior in military power but inferior in economic and especially potential power is more likely to believe that, once its military power begins to wane, further decline will be inevitable and deep. This is especially so if the trends of relative economic and potential power are downward as well. The state will believe that there is little it can do through arms racing to halt its declining military power: it would simply be spending a greater percentage of an already declining economic base in the attempt to keep up with a rising state that has the resources to outspend it militarily. Moreover, economic restructuring is unlikely to help, since the potential power that is the foundation for economic power is also inferior and declining. Under these circumstances, a dominant military power is likely to be pessimistic about the future and more inclined to initiate major war as a "now-or-never" attempt to shore up its waning security.40 The Contribution of the Argument As noted, this book's goal is to build a theory with greater explanatory and predictive power by synthesizing the strengths of current realist approaches. The resulting theory offers two new contributions. First, the theory provides a deductively consistent argument for how changes in rel- [20] Rethinking Realist Theories of Major War ative power should have differing effects in bipolar versus multipolar systems. Classical realism and neorealism emphasize the importance of polarity and occasionally consider dynamic trends, but they do not analyze the effect of power trends across the system-types. Hegemonic stability theory and preventive war arguments are fundamentally dynamic, but they do not include polarity as a critical boundary condition. Second, the book divides power, for theory-making purposes, into three categories—military, economic, and potential. By considering the differentials and trends in each realm, we can determine when declining military power will lead directly to war and when it will lead to measures short of war.41 Even theories emphasizing the problem of relative decline, such as preventive war arguments, have trouble explaining why some situations of decline are more destabilizing than others. Polarity, of course, plays a significant role here, as I have emphasized. Decline is more likely to lead to major war in bipolarity, since the declining state does not have to possess marked military superiority and may even be somewhat inferior. Yet equally important is the declining state's military power in comparison to its economic and especially potential power. A dominant military state that is inferior in economic and potential power is much more likely to expect decline to be both deep and inevitable, inclining it to risky actions. Negative trends in the latter two forms of power will only make things worse. The theory thus helps answer two long-standing questions: Is major war more likely when great powers are equal or unequal? And can major war occur between states seeking only security, or must there be actors with inherently aggressive motives? The answer to the first question is clear: it depends on the polarity of the system. Major wars in multipolarity require a preponderant military power, but they can occur in bipolarity whether the two great powers are equal or unequal.42 This helps explain why in the three bipolar cases before 1945 (Sparta-Athens, Carthage-Rome, and France-Hapsburgs), it was the declining and formerly dominant great power that initiated major war against the rising adversary, despite relative military equality between the states. In the early cold war, there was great instability whenever one of the superpowers feared serious decline, even though the United States re-, mained militarily superior throughout. The multipolar systems before 1914 and 1939, on the other hand, were destabilized only after one state (Germany) came to possess significant military superiority over other states taken individually. In three of the four major wars before 1900—the Thirty Years War, the wars of Louis XIV, and the Napoleonic Wars—the conflict J was initiated by the power with marked military superiority. Only in the C Seven Years War was the declining state essentially equal, an anomaly I j, discuss in chapter 8. I The answer to the second question is equally clear: innately aggressive i [21] The Origins of Major War Rethinking Realist Theories of Major War actors, even though they may exacerbate the likelihood of major war, are neither necessary nor sufficient conditions for such wars. Aggressive unit-level motives are not a sufficient condition, since even the most hostile leaders will be deterred from initiating major war unless power conditions make the bid for hegemony feasible. For the period of the last five hundred years, we might identify a number of European leaders who would have wanted hegemony purely for glory or greed-driven reasons. Yet there were only seven clear cases of major war during this time.* That so few major wars occurred is explained by a simple fact: few states ever achieved the military superiority needed to take on the system.44 (And as I show in the empirical chapters, these wars were driven primarily by fears of decline, even when unit-level factors were also present.) Aggressive motives are also not a necessary condition for major war. A purely security-seeking state may initiate war in either bipolarity or multi-polarity solely because of its fear of inevitable and profound decline. Needless to say, a rising state showing signs of hostile intentions will make this declining state even more likely to attack.45 But the initiator's attack is still a function of security motives, not unit-level aggressive designs. Perhaps the clearest case is Sparta's initiation of major war against Athens. The Spartans feared revolt at home if the soldiers were away fighting a large-scale war. Yet fear of the rise of Athens forced the Spartans into war, even as these domestic factors inclined them to peace.46 As we shall see, Germany in 1914 faced a very similar situation: the key German leaders believed that war would only exacerbate domestic instability at home; yet it had to be chosen to prevent the rise of the Russian menace. Even more to the point, the declining security-seeking state may initiate major war even if all other states in the system, including the rising state, are also only security-seekers. The declining state, given the anarchic environment, will be inclined to doubt the present intentions of other states, despite their best efforts to show their peaceful desires.47 Indeed, rising states have every incentive to misrepresent their intentions as peaceful to reduce the possibility of preventive attack. The declining state will therefore have a hard time sorting out those states that are genuinely peaceful from those that are not.4" Even today, for example, it is unclear whether or not Czar Nicholas II privately desired hegemony but was only postponing a bid until Russia grew stronger. Hence, German leaders, despite Nicholas's efforts to communicate his benign intentions, still felt compelled to initiate major war, Russia faced the same problem in 1939-41, when, despite his best ef-lovts, Stalin could not convince Hitler of his good intentions. Hruillv, she declining state may oven know with certainty that the other has peaceful intentions., but still initiate war tor security reasons. The prob-!■ in is a profound one: the other's intentions might change in the future after it reaches a position of dominance, perhaps because of a change in gov- [22] I eminent or leadership or simply because of its stronger power position.'" Depending on the likelihood of this domestic change in the other, preven-l tive war or preventive measures that knowingly increase the likelihood of major war through inadvertent means can become rational even against an J. adversary recognized to be peaceful. Truman acted in mid-1945 to contain ' the Soviet Union, raising the risk of an undesired war, even though he be- l lieved Stalin was a reasonable individual. Truman feared not Stalin's inten- se tions per se, but rather the intentions of those who would take over after his I death. In 1962, Kennedy initiated a crisis over missiles in Cuba, not because ; he thought Khrushchev wanted nuclear war, but because he could not be \ sure of Russian intentions down the road should the Soviets achieve a per- 1 I ceived measure of nuclear superiority.50 Questions Regarding the Theory's Logic I I now address three questions that arise about the logic of the argument (presented in this chapter. First, does the preponderant state in multipolar-ity (state C in fig. 1) need to possess military power greater than all other great powers combined to make major war a rational option?51 The answer: not necessarily. A state in multipolarity can make a potentially successful bid for hegemony with less than 50 percent of the system's military power by taking advantage of the difficulties coalitions have in coordinating their military actions. In short, there is good reason to expect that coalitional i strength will be less than the sum of its parts.52 States that are considering whether to form a coalition against a preponderant power are faced with two contrary sets of incentives. On the one hand, no great power smaller than the preponderant state has an interest in seeing this state defeat the others, since it knows it could be next.53 Accordingly, smaller great powers are pushed toward a coalition to prevent the elimination of their partners. On the other hand, owing to anarchy and the concern for relative power, great powers in multipolarity face an intense form of the collective action problem. Individual states in a coalition against a challenger have an incentive to sit on the sidelines or to contribute less than their fullest effort, so as to maximize their relative position after the war ends.54 Hence, coalitional tightness is likely to vary between the ex-'• tremes of concerted effort when the potential hegemon appears particu- larly strong, to coalitional disunity when the threat to the system appears : minimal." Challengers can take advantage of this disunity by trying to eliminate other great powers one at a time. This argument helps explain 1 why states may make bids for hegemony in multipolarity when tivy arc- larger than anv single state but smaller than the combined resources of all j other powers. [ [23] The Origins of Major War Because of the incentive to "hang together or hang separately," however, challengers that take on the system cannot depend on complete allied disunity. Hence, coalitional unity and fighting power will likely be some- ' where between strong and nonexistent. What we would expect is that the J more evident the preponderant state's military superiority versus any 1 other individual great power, the higher the others' incentive to band to- .1 gether against this threat. Conversely, the smaller the leading state's level 1 of superiority, the greater the others' incentive to sit on the sidelines. It may seem contradictory to argue that a declining preponderant state may make a bid for hegemony because the collective action problem keeps others disunited, and also to argue that a declining near-equal state in multipolar! ty avoids war for fear of provoking a coalition or from a belief that it can form alliances. It is not. In both cases, the leading state recognizes that others are torn between two contrary incentives: to unite for fear of defeat, or to sit on the sidelines to maximize power. Thus any state taking on the system will likely have to contend with some coalitional effort, but it can also expect some disunity. The key difference is that a near-equal state has no way of defeating every other power singlehandedly even if the coalition against it is nonexistent; costly bilateral wars will sap its strength. A preponderant state, however, stands a chance of defeating the system even with a coalition against it, and especially if the level of coalitional disunity is fairly high. Bilateral victories will be quicker, with lower costs, and thus will not deplete its strength for future attacks. This discussion helps resolve a second major question: why does the rising and increasingly dominant state in multipolarity (state C in fig. i from ■f2 to f4) not fear a preventive attack by an offensive coalition of the other states, even though it represents an obvious and growing threat? The likelihood of such an attack is low precisely because of the intense collective action problem. Even if the coalition can be formed, each partner fears that its allies will deliberately hold back their full commitment to shift the costs of preventive war to others. When states are in a purely defensive alliance, the collective action problem is moderated by the fact that the alliance does not automatically mean war; thus, actors can be convinced to ally solely to deter the growing state from war. Coordinating an offensive coalitional strike is more difficult: attacking a rising dominant power guarantees war, thus instantly raising each state's fear that it will bear the brunt of the rising power's fury. A "you first" mentality will ensure disunity in any offensive coalition that is suggested.56 Accordingly, the rising state in multipolarity will likely grow unmolested, with each declining state hoping either that the rising state will not achieve sufficient power to take on the system, or f that their own internal and external measures can deter it from doing so.57 f The third issue is the most complicated. Notwithstanding the earlier \ discussion, one may still wonder if the theory ultimately needs unit-level > Rethinking Realist Theories of Major War drives for major war, such as greed, glory-seeking, or ideological hostility, to be workable. The short answer is yes.5s The possibility of such drives arising in the future is a necessary condition for the theory's causal argument. That is, the fact that human beings have the ability to use violence for nonsecurity objectives is required to make preventive war fully rational. Imagine a planet where the beings were hardwired by nature never to strike unless they were immediately about to be struck—that is, a planet where the actors were physically incapable of using violence for nonsecurity motives like greed and glory. The basis for the security dilemma would, at least over time, disappear. Leaders would understand that even if they allowed rising states to become preponderant, these states would never attack later at their peak for aggressive unit-level reasons. Nor would such states attack later for their own preventive reasons, since they as well would know that they had nothing to fear from rising states. In Rousseau's terms, all great powers would be permanent staghunt actors, preferring peace to all other outcomes.5" Major wars could occur only through preemption. But since there would be no reason to threaten another, beliefs that the other was readying itself for immediate attack—a necessary condition for preemption between staghunt actors—would not arise. This argument does not mean, however, that on planet earth major wars require actors that are presently driven by aggressive, nonsecurity motives. As discussed, a declining state seeking only security could attack a rising state which also pursued security, simply because it was uncertain that the other was currently a security-seeker or would still remain one later at its peak. Yet the dilemma for the declining state does not end there. Even if it knows for sure that the other will still be a security-seeker later, it has reason for preventive war now. It knows that the rising state, once it peaks in power, will be faced with the same problem of uncertain intentions that the declining state faces now. So if the rising state is likely to launch a defensive preventive war later, with much more power, for fear of decline, then preventive war now by the presently declining state makes sense. Figure 3 shows this dilemma for the bipolar situation."0 State A at time t0 knows that after time f, state B will consider preventive war for purely security reasons, since B cannot trust A's future intentions at time t2. The causal sequence is clear but tragic: because A may turn into, an innately aggressive state by t2, B has incentives for preventive war at tv but because it does, A has incentives for preventive war at f0. And of course even if A knows that it will still be a security-seeker by t2—and knows it can convince B of this fact—B still has preventive motives at tt because it cannot be sure A won't launch a preventive attack at t2 for fear of B's motives at f,."1 This argument, while logically consistent, may seem too convoluted to be borne out in practice. Yet as will we see, U.S. leaders in the cold war feared that if [25] the Soviets ever reached temporary superiority, tney mignt anacK preventively out of their own fears for future security. For states to reason in this way in bipolarity and multipolarity, there must be the possibility that eventually one of the states will attack for nonse-curity, aggressive objectives. Yet we can see that major war can still occur even when there are no presently aggressive actors or even the likelihood of such actors on the horizon. Most tragically, the declining state may know these facts and still have an incentive for preventive war. Methodology: The Definition of Major War Major wars, as noted in the introduction, are wars that involve all the great powers, that are fought at the highest level of intensity, and where there is a strong possibility that great powers may be eliminated as sovereign states. As an ideal type, this definition is designed to establish boundaries for the analysis. Efforts to build universal theories of war are likely to be confounded by the diversity of wars through history. It is unlikely, for example, that civil wars, local conflicts like the Vietnam war, European imperialist wars of the nineteenth century, and major wars such as World Wars I and II will have a common cause.62 By starting with the most profound conflicts—general wars that threaten the survival of the system's most powerful actors—I am seeking a more manageable theoretical goal. Yet since such wars, or the avoidance of them, so determine great power behavior, a strong theory of major war will often have much to say about the causes of smaller conflicts through history. I purposefully do not define major wars according to their duration or number of casualties.63 To do so would be to presuppose the type of war based on its outcome rather than its nature. By my definition, therefore, World War II would still have been a major war even if Germany had quickly dispatched France, Britain, and Russia, as seemed likely after France's defeat in June 1940; it was Germany's effort to destroy the other great powers that made it a major war. Not defining major wars in terms of duration or casualties has an important benefit. It allows one to measure the military balance between great powers in the most objective way possible, namely, by looking at how the war played itself out on the battlefield. As all military analysts recognize, overall military power, as a state's relative ability to fight and win in a war, reflects the combined influence of a whole host of quantitative and qualitative factors. Thus only in actual war do we see each side's true military strength (after adjusting for fortuitous factors such as weather).64 But to use this technique to measure military power while defining major wars according to duration and battle deaths would intro- The Origins of Major War Rethinking Realist Theories of Major War ciuce an important bias: measures for the dependent variable (major war) would also be factors influencing the measure of the independent variable (relative power). My definition thus requires only that the states fight at the highest level of intensity—that they are fully mobilized. Thus a state like France in 1940, which gets beaten easily despite complete mobilization, was still involved in a major war. Yet the details of the battle are useful in establishing the objective military balance between the two sides. This definition establishes only the ideal-typical major war; no actual war will ever fit the criteria perfectly. The criteria provide a standard for separating wars of fundamentally different types, but they should not be held to religiously. Debating whether World War I was a major war until 1916 because the United States was not involved is hardly worthwhile: the war was so obviously close to the ideal-typical form as to require its inclusion. Nevertheless, lines do have to be drawn. To ensure a focus on system-wide wars, I exclude bilateral wars within multipolar systems. Thus while Organski and Kugler include the 1904-5 Russo-Japanese conflict as a major war, such a war must be excluded from my study since it involved only two of seven or eight major powers, had little possibility of escalation to systemic conflict, and was aimed not to eliminate either power but to control Korea. Research Method Six principal questions may be asked about the research that underlies the theory presented in this book. Should the theory be deductively or inductively driven? Under what conditions is the theory falsified? Should the evidence be evaluated primarily quantitatively, through statistical tests, or qualitatively, through intensive diplomatic-historical case studies? What are the best measures for the independent and dependent variables, relative power and the probability of major war? What criteria guide the selection of cases? How can one mitigate some common methodological obstacles, including selection bias and the omitted variable problem? Dynamic differentials theory is a deductive systemic argument modeled on microeconomic theory. Microeconomics starts by assuming exoge-nously determined tastes for consumers and firms (as represented by fixed indifference curves), and it posits rational, self-regarding actors seeking the best means to their ends (the maximization of utility and profit). With this micro-foundation, variables such as prices and costs of inputs are introduced to predict deductively how behavior should change with changes in these external factors."' Similarly, I begin by assuming actors with singular goals—the maximization of their security rather than utility or profit—who calculate the rational steps to reach these ends. Then, by holding unit-level [28] factors constant, I derive predictions as to how behavior should change in response to changes in a core systemic variable: the size of the differentials of power and the trend of those differentials. Figure 4 provides a list of the theory's assumptions. Each assumption is a variable that is, for theory-making purposes, fixed at a particular point. This specification of boundary conditions permits one to conduct a controlled mental experiment isolating the expected causal impact of changes in the independent variable on the dependent variable.61' In this book, some assumptions are particularly important. I assume, for example, that states are uncertain about the other's future intentions.1'7 A declining state therefore does not know whether the rising state will attack it later, giving it reason to worry about its future security.68 The assumptions that states are "neutral" in terms of their tolerance for costs and risks allows the theory to posit actors who do not shy away from war if it is the best means to security (nor do they embrace war as the only means). These assumptions help to reinforce a picture of highly rational actors who choose options that best maximize their long-term likelihood of survival. Building a deductive structure in this way has important advantages. It allows one to build a stronger theory of the systemic pressures on actor behavior, even while recognizing that leaders will sometimes be influenced by domestic- and individual-level factors. Variations in such unit-level factors or in systemic factors like the offense-defense balance or geographic position act as "disturbing causes," pushing actor behavior away from what might be expected from looking at power differentials and their trends alone.65 Theoretically, one can anticipate deviations from a theory's narrow predictions by simply relaxing the assumptions. That is, one allows assumptions to vary away from the fixed points, to determine deductively the impact on expected behavior. Thus when the theory is taken to the real world, where facts rarely conform perfectly to assumptions, behavior and outcomes can not only be predicted, but explained. I assume, for example, that states are the same distance from one another. Relaxing this assumption implies that distant rising states should be less frightening than rising states that are nearby (for any given rise). Thus, Britain's greater concern for Germany's rise after 1895 versus the rise of Japan or the United States can be explained within the theory's logic. In chapter 9,1 relax the assumption that declining states are uncertain about the other's future character. This permits the theory to predict the likely effects of regime-type (e.g., rising democracies versus rising authoritarian states) on the declining state's behavior. This discussion helps set the boundaries for falsifying the theory. A deductive systemic theory is not falsified simply by pointing to cases where unit-level factors drove actor behavior. These may simply be cases where unit-level factors, as disturbing causes, were so powerful that they over- [29] (Note: each assumption is a variable thai could vary in practice, but thai is, for theoretical purposes. Iixecl at a certain point, as indicated by the dotted line.) Unit-level assumptions 1. Rationality: actor calculates best means low high to desired ends, given information available <--.- -,—!_>. 2. Ends: actor seeks security above all else other ends <_--.- security --;> 3. Direction of ends: to self only altruistic <- egoist -■— 4. Nature of actor: unitary nonunitary <- fully unitary - ; >■ 5. Risk-tolerance: neutral risk-avoidant <-; risk-acceptant -_-->. 6. Cost-tolerance: neutral cost-avoidant <-L cost-acceptant --*- Systemic-level assumptions 1. Certainty regarding other's present intentions: largely uncertain fully uncertain <-!- fully certain -> 2. Certainty regarding other's future intentions: fully uncertain fully uncertain <_j- fully certain --»- 3. Certainty regarding one's past and present power levels: largely certain fully uncertain <- fully certain -!-> 4. Offense-defense balance: neutral def. dominant <-L off. dominant -—-->■ 5. Geographic positions of states: equidistant van-distant <-,--. equidistant -!_> 6. Technological Costs of War: moderate low <-L high -,---> Fig. 4. The assumptions of the theory rode the effects of the theory's systemic variable. Moreover, a good theory, as noted, will help predict deviations from narrow predictions as one relaxes core assumptions. The real standard for falsification must therefor." be that the actors, in terms of the impact of systemic factors, did not act for the reasons the theory hypothesizes, but rather according to another systemi-cally driven logic.70 The theory in this book is thus not competing directly against unit-level theories, but only against other realist theories focusing on the role of power. Most evidently, if we find that leaders who believe their states are rising are initiating wars and crises, my systemic logic would be discontinued and the logic of hegemonic stability theory upheld. Likewise, if we see states in multipolarity taking on the system despite only essential military equality, then the theory would again be wrong and hegemonic stability theory correct.71 Unit-level factors do not falsify a systemic theory, nor do systemic factors falsify unit-level theories. There is, however, the separate issue of salience. If unit-level variables are almost always more dominant in historical cases, then we have reason to doubt the causal importance of systemic theories. Accordingly, in the empirical chapters, I test the theory not only against other systemic arguments, but also against domestic- and individual-level theories. The theory will gain in credibility if it can be shown that power differentials and trends often overrode unit-level factors. The clearest tests are examples where unit-level factors push toward peace, but the actors choose conflict anyway because of the systemic pressures. Yet the theory must be seen as less salient (even if logically sound) if actors evaluate power in the expected way but end up acting for primarily unit-level reasons. The ideal evidence necessary either to falsify different theories or to establish their salience comes from internal documents. To determine causality, correlations between objective factors and actor behavior are insufficient. We must ultimately examine how the actors thought, that is, we have to see if they acted according to the logic of the theory. Many scholars wrongly assume that systemic theories must avoid examining actor perceptions, since this seems to require a descent to the unit level. Hence many systemic theories rest only on measures of objective factors, such as the number of tanks or troops. Yet social science is about human behavior, and human beings act only from their beliefs about phenomena.72 In fact, what a scholar might establish as the objective balance in tanks fifty years after a war is largely irrelevant in testing any theory, since if decision-makers did not or could not observe such evidence, they could not have acted upon it. No one would develop of theory of fire-behavior by analyzing, after the fact, whether there really was a fire; what explains behavior is whether the actors believed a tire was in progress, and its perceived severity. Similarly, all theories of international relations, to establish causality and not just car The Origins of Major Wat- Rethinking Realist Theories of Major War relation, time. mntely must be tested against the beliefs of the actors at the To test any social science theory, therefore, two questions must be asked. Did the actors act for the reasons the theory hypothesized? And if they did, were the beliefs underpinning their actions reasonable given the information available? Answering yes to the first question establishes the causal power of the argument. If leaders take their nations into major wars or dangerous crises because they perceive their states are in severe decline, and not for other reasons, then the book's theory explains why they acted the way they did. If they acted according to another power logic, the theory is falsified. If they acted as a result of the overwhelming influence of non-power variables, then the theory lacks salience. The second question tests the "realistic-ness" of the rationality assumption. If, given the information available, leaders should have reasonably seen that the state was rising, not declining, then social and psychological pathologies might have to be brought in to explain why the leaders chose preventive war. Note, however, that even in this case dynamic differentials theory, not hegemonic stability theory, explains why the actors did what they did. Its causal logic is thus upheld, while hegemonic stability theory would be disconfirmed. "Objective" evidence an investigator can collect might still have a role to play. It would help one answer the second question, by showing that the actors' beliefs were reasonable or unreasonable. But even here, to demonstrate irrationality, it must be shown that the actors had access to this information and chose to ignore it, or they could have had access but because of psychological barriers, performed an incomplete information search. In my case studies, I do make reference to objective evidence on the military, economic, and potential power levels of nations. As noted, the most objective measure of relative military power is what happens on the battlefield, where quantitative and qualitative factors come together as one. Yet to test causality, documented evidence of actor beliefs about the power balance is still the key measure.74 In sum, examining perceptions does not automatically mean that a theorist has dipped down to the unit level. If the perception is of a phenomenon external to the state and the phenomenon is evaluated rationally, then the primary cause of behavior is at the systemic level.75 Causality is at the unit level only if the perceptions driving behavior are of a domestic-level phenomenon, such as internal instability or bureaucratic maneuvering. Note that examining perceptions also does not mean one is necessarily a "con-structivist." Constructivism focuses on how state-to-state interaction can reshape identities and interests.76 My focus is simply on leaders who learn about the other's material power. Such leaders are not changing core values and interests, only their beliefs about an external reality (indeed, mv de- i ductive logic assumes that actor ends are fixed). Moreover, no interacti0ri between states is required: states can collect information by means of spy_ ing, satellites, published documents, and so forth.77 This leads to a discussion of why this book uses qualitative case studjes rather than the quantitative approach of some realists.78 Most important |s the above point: that quantitative studies are a second-best approach to es_ tablishing causality when internal documents are available (although I use statistics to reinforce that leader perceptions were for the most past r©^_ sonable). Moreover, quantitative studies lack "a good way to measure rt^y dependent variable, the probability of major war. Perhaps because of tl-»fs they almost invariably examine the dichofomous dependent variable war/not-war. I estimate changes in the probability of major war by chan|>es in the extent of the states' planning for major war, changing internal estimates of the likelihood of such a conflict, and fluctuations in trie*" general level of hostility between states, shown most clearly by major crises like th,e Cuban missile crisis. This approach may not be fully satisfying, but it does capture our intuitive notion that even when major war is not occurring some situations are relatively more stable than others. Making the dependent variable continuous, both in theory-building and theory-testing, offers a number of advantages. It forces any theory toexplaii^ more than just the historical outbreak of actual major wars. This means that times of crisis and conflict, such as the Cuban missile crisis, can be brought within the purview of the theories (see chapter 2). In doing so, we greatly expand the size of the data set (the number of "observations").™ Instead of examining just two data points in the twentieth century, 1914 and 1939, which are then compared to the many years of peace, the analyst can explore variations in the likelihood of major war over time. Periods of relative calm, such as the 1920s, can be compared to the 1930s, and the 1930s themselves can be studied for yearly shifts in the level of tension and crisis. Fluctuations in the level of crisis within the cold war can also be brought within the theory's purview. The bulk of my empirical work focuses on three main periods: the run-up to World War I; the interwar years up to World War II; and the early cold war from 1945 to 1962. In chapter 8,1 also briefly cover seven other key major wars in western history. Aside from their sheer historical importance, the twentieth-century cases satisfy the most important criterion for establishing causality: a huge number of declassified documents. Moreover, since Germany lost both wars, the German documents-provide a relatively complete and unbiased view of-internal decision-making.80 The openness of the American declassification system also offers such a view into the logic of the cold war. The cut-off of 1962 was established for one main reason; given the "thirty-year rule" for U.S. declassification, documents released after this date are less numerous and more politically suspect. Despite the focus on these three periods, since 1 examine the changes in [32] [33] i ne unguis of Major War I state behavior across time and space, the number of observations is far f larger than three. This expansion of the realized "cases" helps to avoid the f, bias caused by selecting on the dependent variable, namely, looking only at | the years when major wars broke out.82 Analysis of the run-ups to major | wars and crises provide us with variations on both independent and de- | pendent variables.'" In addition, the consideration of seven other major I wars prior to 1900 provides a check against the possibility that the twenti- I eth century is biased toward my argument. In this way, the generalizability of the argument across time and space can be tested.84 It is also worth noting that the twentieth-century cases are not easy ones for the theory. Almost every study of the origins of World War II emphasizes the unit-level causes of the war, most obviously the characteristics of the Nazi state and its leader. If I can show the salience of my argument in such a case, there is greater reason for confidence in the theory.85 Arguments on the origins of the cold war and the crises of the 1945-62 period also tend to emphasize either the ideological roots of superpower conflict or the role of misperception regarding the other's intentions. Although arguments on the impact of shifting power may be more common here than for World War II, they still face strong competition from unit-level theories."6 The run-up to World War I seems on the surface to be the easiest period to make my argument stick, since a number of scholars have documented German fears of a rising Russia. Yet this case is still, in a different way, a hard one for the theory. World War I seems to support practically every theory of major war out there, at both the systemic and unit levels. To show, as I seek to do, that only dynamic differentials theory provides a comprehensive explanation for the war is thus a difficult undertaking. A problem for many studies is the problem of omitted variables—variables that are left out of an empirical test but have a strong causal relation to both the dependent and independent variables.871 deal with this problem in a straightforward way. For each of the twentieth-century cases, I include all the primary causal arguments that have been established to explain actor behavior. I then use the documentary evidence to test the explanatory power of these arguments against this book's theory. In this way, we can be sure that there is no hidden variable causing the result. Moreover, this method helps deal with the problem of overdetermination. If the documentary record supports dynamic differentials theory, but also calls alternative hypotheses into question, then there is greater confidence that shifting power differentials and not other variables are driving the observed behavior and outcomes. Thus, in cases like the First World War, I am not simply adding yet another hypothesis to the many that exist but rather seeking to j show flic limitations of competing arguments as 1 support the plausibility I of it; v 0 ho C C CD T3 CD ■£ £ S »-» N £ (D Si -Q D) CD [42] The Origins of Major War his actions seem largely driven by greed and glory. Glory and ideology also shaped Hitler's calculus in 1939. My argument does not claim to explain the origins of such unit-level factors, and they may play an important and independent role in causing major war. The issue for the empirical chapters, however, is their relative historical salience versus my theory's more systemic factors. Pathways three to five are situations where a state initiates a crisis without necessarily desiring major war; that is, the crisis is not a mere justification for war, but serves another purpose. In each of these three paths, the foundational motives can be either preventive (security-driven) or aggressive (unit-level based). My concern is with the former. In the third path, the initiator has deadlock preferences prior to the crisis because of steep decline that makes the continued status quo unacceptable. It provokes a crisis, however, in the hope that it can coerce sufficient concessions from its adversaries to ameliorate decline and make major war unnecessary. Should the adversaries fail to offer adequate concessions, the state, given its deadlock preferences, will then move to initiate major war as per the first pathway. For example, before the Peloponnesian War, Sparta, a state reluctant for domestic reasons to fight a major war, sought to coerce Athens into giving up its empire. Such a concession would have reversed the decline that was pushing Sparta toward war. Yet once Athens refused to comply, Sparta was forced to attack. The fourth and fifth pathways capture the two primary means of inadvertent war. As noted, inadvertent war is a war that occurs despite the fact that prior to the crisis period, no state preferred major war to continued peace.23 The processes of the crisis itself, therefore, play a significant role in causing one state to initiate war despite these pre-crisis preferences. And war is indeed chosen; it is a conscious act. Although the literature often refers to accidental war,24 implying that neither side made a decision for war, in practice major war is too significant to "just happen." Figure 6 thus reexpresses within a cohesive framework the various elements of inadvertence in the crisis literature. Pathway four is war through preemption. Each side, or at least the initiator, has "staghunt" preferences, where peace is preferred to getting in the first blow. But because of first-move advantages (an offensive-dominant system), striking first is better than an outcome where both attack essentially simultaneously and is certainly preferred to being hit first.25 In such a situation, since peace is preferred to all of the war outcomes, a state will initiate a war only if it believes that the other is preparing to strike.26 Thus even though the two are often conflated, preemptive motives and preventive motives remain distinct. In preventive war, the initiator has deadlock preferences (preferring war to peace) not because it fears immediate attack against its homeland, but because its declining situation means that the ris- [44] Foreign Policy Choices and the Probability of Major War ing state will be able to attack later with more power (independent of whether there are first-move advantages). In preemptive war, the actors still prefer peace to war. But since they believe that the other is preparing to attack, they cannot afford to be caught unaware, given the offense-dominance of the system.27 Crisis periods thus have a significant impact on the motives for preemptive major war. Most important, crises typically involve the increased mobilization of each side's military forces, the placing of these forces on alert, and hostile rhetoric (demands and threats). These measures ensure that the state will not be caught unprepared, while also serving as signals of a state's resolve. Yet they can also produce a mutual fear of surprise attack, since each side is uncertain as to what the other will do with its mobilized forces. As figure 6 indicates, accidents within the crisis will greatly heighten this fear. In the Cuban missile crisis, for example, as both sides moved to higher states of alert, each side feared that accidents like the U2 downing over Cuba might reinforce perceptions that the other was readying itself for attack.28 Preemptive wars in general, not to mention preemptive major wars, are rare events in history.29 The Seven Years War resulted from Prussia's preemption of Austria and Russia's impending preventive attack, but this is the only clear preemptive major war of which I am aware. A plausible story about preemption and World War I can be told (indeed, it remains the standard account). As I discuss in chapters 3 and 4, however, preemptive motives drove only Russian behavior, and German leaders exploited these Russian fears to bring on the preventive major war that only they wanted. Yet the rarity of preemptive major wars in practice is not surprising, given leaders' awareness of the dangers of things getting out of hand.30 The risks of preemption, therefore, cannot be dismissed; indeed, the very act of downplaying these risks would make such wars more likely. Moreover, the Seven Years War and the fact that preemptive wars on a lesser scale do occur—the clearest examples being the U.S.-Chinese war in Korea in 1950 and the 1967 Arab-Israeli war31—show that preemption at the level of major war is always a real possibility. In the modern era, in fact, nuclear weapons so fundamentally alter the costs of war that preemption is one of the few pathways to major war which can still be seen as "rational."32 Any theory of major war hoping to be applicable to the post-1945 period must therefore explain why states would ever initiate crises that significantly increase the risk of a preemptive total war. If the fourth pathway shows how a crisis can reshape beliefs, the fifth pathway shows the effect of crises on preferences. Crises force states to put their reputations on the line. Actions such as crisis mobilization, harsh rhetoric, and military moves against third parties often have high audience costs: should a state take such actions and then back down, domestic and in- [45] l he Origins of Major War ternational audiences may perceive it as weak.13 As work on incomplete-information games shows, actions with high audience costs send "costly signals" revealing a state's true resolve. Since weak actors are less likely to take such actions, challengers can update their beliefs about a defender's toughness, and thus back away before things go too far.34 There is a downside to this, however: actions with large audience costs make it harder for states to make concessions within a crisis, thus narrowing the bargaining space of negotiated solutions both prefer to war. In short, reputations may get over-committed. The crisis then transforms actors who had chicken or staghunt preferences before the crisis into deadlock actors, who prefer war not only to the old status quo but also to any negotiated deal within the crisis.35 Reputational commitments help explain Russian behavior in 1914 and British/French behavior in 1939. In the July crisis, Russian leaders did not want major war and thus sought a negotiated solution. Yet as the crisis unfolded, the reputation loss entailed in abandoning Serbia shifted Russian preferences toward conflict. Likewise, in August to September 1939, the reputational costs of backing away from promises to defend Poland meant that war for London and Paris was preferred to any negotiated deal. World Wars I and II were not inadvertent, since, as chapters 3 through 5 show, Germany brought on war despite knowing that its adversaries would have to respond. Still, these situations do show how the commitment of one's reputation can pull states into major war. Although there are few cases where overcommitted reputations were the sole cause of major war, such a pathway is clearly a profound risk, especially in the nuclear era. In the Cuban missile crisis, for example, U.S. leaders understood that a strike on Cuba would force Moscow, for reputational reasons, to attack Berlin or the U.S. missiles in Turkey. Yet as Robert McNamara noted on 27 October, this would then require U.S. retaliation, perhaps against Russia's Black Sea fleet. Each round of tit-for-tat would have made it increasingly difficult to retreat from general war. Crisis Initiation versus Accommodation / Conciliation The discussion of pathways to major war clarifies the logic of the decision-making model. In maximizing security, declining states will not jump into preventive war if crisis initiation can reverse the decline. Still, leaders know that crises put both sides on a slippery slope to major war through either preemption or reputation-driven escalation. Indeed, it is precisely the risk of things getting out of hand that permits leaders to believe they can coerce concessions from the other and thus mitigate decline.36 Leaders understand the risks of inadvertent war. In thirteen of the sixteen crises examined by Glenn Snyder and Paul Diesing, policy-makers feared thai the crisis might spin out of control.1 This result should not be f Foreign Policy Choices and the Probability of Major War 5 surprising: it is intuitively evident that great power crises like the Cuban $ missile crisis involve a higher risk of all-out war.3S In fact, the rarity of crises \ in history is easily explained by states' fears of provoking their adversaries, I knowing that inadvertent escalation to war might occur.39 i Critical to explaining why a state might choose crisis initiation over more moderate options is understanding the specific source of the state's decline: I where is the decline coming from, and how "fixable" is it by crisis mea-: sures? For the initiation of crisis to be rational, given the escalatory risks, a 1 leader must expect that a crisis can coerce geopolitical concessions of the t kind that will address the root of decline. i If decline is rooted in a state's inferiority in economic power and poten- l rial power, there is a problem. It is generally difficult to compel large concessions on such things as GNP and population/territorial mass through crisis initiation. This was part of the German predicament before 1914 and 1939. It was highly unlikely that Russian leaders could have been coerced by a crisis into relinquishing huge amounts of territory and population or into forgoing industrialization. Yet barring such concessions, Russia was s destined to overwhelm Germany over the long term. Exploiting short-term military superiority, German leaders plunged into two devastating wars. h Even when states have more equitable distributions of territory, popula- tion, and raw materials, entrenched relative stagnation can still set in. Depending on the cause of stagnation, crisis initiation may or may not over- '. come it. Spain's steep decline after 1600, for example, was driven by deep structural problems within the Spanish economy. Initiating crises against ; France, Spain's primary adversary, would not have solved these problems; preventive war to buy breathing space for reforms seemed the only way (chapter 8). Likewise in the case of Soviet stagnation in the 1980s, as I discuss shortly, crises would not have resolved Russia's need for modern technology. Fortunately, accommodation held out the possibility of securing this technology, while the preventive war option was very unappealing. When neither inferiority in economic /potential power nor entrenched stagnation is the main problem, the situation is less dire, and states should be reluctant to accept the risks of crisis initiation. They may still do so, however, because of the third form of decline: power oscillations. In an ongoing rivalry, deterrence/containment measures like arms racing and alliance I building are usually preferred to crisis initiation. Deterrence actions are certainly provocative. Yet since they are steps largely internal to a bloc, they usually entail lower risks of inadvertent escalation than crises. Crises are more escalatory because they typically involve a direct challenge to the ;. politico-territorial status quo and thus to the other's position and reputa- i- tion. But deterrence policies are often not enough. Sometimes a state, de- I spite its best efforts, cannot keep up with another in the short term. I Relative geostrategic strength is a matter of the relative success of a state's The Origins of Major War armament and alliance programs. Both sides may be trying vigorously to match the other side's actions. Yet one side may be far more successful, at least for a while, in building actualized military-geopolitical power. For the state less able to keep up in the short term, this creates a dual dilemma. On the one hand, the state fears that the adversary's growing strength may give it the confidence to attack later for nonsecurity reasons. On the other hand, a time lag in the effectiveness of the declining state's deterrence/containment programs means that its very capability to reverse the trends later may give the other a strong security-driven reason for war. In short, the state knows that its very attempt (and its ability) to reverse the short-term trends—to catch up after falling behind—can push the other into preventive war. Thus oscillations in the relative geopolitical balance caused by differing success rates for deterrence policies can push states to more extreme options.40 In the short term, decline will be seen as exogenous in the absence of stronger action. More dangerous policies like crisis initiation will therefore be seen as necessary to mitigate the decline that arms racing and alliance building alone cannot address. As we will see in chapter 7, power oscillations were behind the three key crises of the early cold war. Stalin pressed on Berlin in 1948 in order to compel Washington to reverse its plan to unite the three sectors of western Germany and integrate the result into the western bloc. In 1961, Khrushchev initiated another crisis over Berlin to force an agreement that would stabilize the economic situation of East Germany. In 1962, Kennedy brought on a crisis over missiles in Cuba to avert a significant short-term shift in the balance of power. In each case, two conditions for crisis initiation were present. First, greater arms and alliance activity alone would not have stopped the negative power oscillation in the near term. Second, the leaders in question could reasonably believe that initiating a crisis might compel concessions that would address the cause of decline, albeit at a heightened risk of inadvertent war. In some situations, neither crisis initiation nor hard-line deterrence policies hold out the prospect of averting decline. If the state does not have the military power necessary for a successful war, then accommodation with the rising state(s) may be the rational option.41 This option can be expected to be effective, however, only under certain strategic circumstances. In mul-tipolarity, if one is declining in relation to two or more states, boundary conditions such as geography and technology can play important roles in determining which state one accommodates, and which state one opposes. That England after 1890 decided to align with the rising United States and Japan rather than fighting a preventive war against either is not surprising. Geography meant that America and japan were much lower threats to Britain's existence than a rising Germany. The technology of the day reinforced this tendency. With the emergence of the airplane, the British splen- Foreign Policy Choices and the Probability of Major War did isolation policy was becoming increasingly outdated. Yet since U.S. and Japanese planes could not reach British shores, these states posed little threat from the skies.42 In more bipolar situations, accommodation may be the smart strategy, but only if such a policy stands a good chance of reversing decline. For sound strategic reasons, we are unlikely to find many declining states accommodating just to buy the rising state's goodwill—that is, to buy a promise that it will not attack later when it has become preponderant.43 The problem, as James Fearon notes, is one of trust under anarchy. The rising state may genuinely wish to commit now to not attacking later, but the commitment is not enforceable; there is little to stop it from changing its mind after preponderance has been achieved, particularly if current leaders are no longer around.44 If the accommodation involves concessions that only cause the state to decline even further, accommodation is even less likely to be adopted simply for the sake of security promises. There are those rare circumstances where accommodation in bipolarity holds out the prospect of mitigating one's decline. In the mid-1980s, for example, Gorbachev and the Politburo moved away from a hard-line policy that was only exacerbating Russia's entrenched economic stagnation. By shifting to detente, the Soviets had a clear goal: to secure trade and technology concessions that would revitalize the Soviet economy.45 Without these concessions, Soviet economic and potential power was expected to continue to fall behind the west. For this accommodation strategy to be the rational choice depended on a number of conditions. Preventive nuclear war would hardly have furthered Soviet security, given each side's ten thousand strategic warheads. Initiating a crisis, such as another move on Berlin, would have been fruitless. It would not have dealt with the root problem: the inferiority of Soviet technology as the world moved into the information age. The Soviets also could rest behind their huge strategic arsenal as they made the concessions on eastern Europe, Euromissiles, and the like necessary to secure U.S. promises on trade and technology. In most bipolar situations, accommodation is less likely to be effective. In the Sparta-Athens, Carthage-Rome, and France-Hapsburgs cases, the declining states had no powerful nuclear second-strike to lean upon; preventive war thus emerged as the rational strategy. This discussion reinforces the importance of understanding the source of a state's decline, and to what extent a particular policy will address it. One factor in both bipolar and multipolar systems is the specific foundation for potential power. In the nuclear era, technology and education are critical components of potential power. They are not only the basis for economic growth but are essential to the ongoing modernization of nuclear weapons and their supporting logistical and communications structures. The Soviets in the 1980s, therefore, had reason to fear the widening technological gap. [48] U9] ruitigu runty unu/tes ana irie L'ruuuuauy ujivutpr war In particular, the deployment of such innovations as the Strategic Defense Initiative ("Star Wars") system might have undermined the Soviet second-strike capability. It is thus not surprising that a critical objective in Gorbachev's bargaining posture, in addition to loosening trade constraints, was obtaining restrictions on Star Wars research and deployments. Before 1945, however, territorial size and population were typically the key components of potential power. Superiority in land mass implied the diversity of raw materials and food production needed for a growing economy, which in turn ensured that a large population could be shaped into an effective fighting force. Inferiority implied the opposite. As noted, Germany's core problem in this century was precisely its severe inferiority visa-vis Russia in territory and population. The same was the case in the pre-1945 bipolar cases. Sparta tried to demand Athenian concessions on its empire, the basis for Athens's potential power, but it seems clear that Sparta had little hope Athens would comply. Carthage by the 220s b.c. was dealing with a Rome that controlled the Italian peninsula, Sicily, Sardinia, and the coast of southern France. Having won these territories through costly wars, the Romans were unlikely to give them up just to allay Carthaginian fears of decline. In the French case, the Hapsburgs by 1520 controlled twice as much territory in Europe and had just conquered large parts of Latin America. In each case, preventive war against the rising colossus before it had consolidated its strength thus appeared to be the only rational strategy. The choice between war, crisis, deterrence, and accommodation is a difficult one. Preventive major wars, given their high risks, are options of last resort. Crisis initiation is also a risky option, and will usually be chosen only when less provocative hard-line policies like arms racing and alliance consolidation prove ineffective in sustaining the power balance. But the crisis must hold out the hope of coercing the kind of concessions that can address the source of the state's decline. Hence, when decline is rooted in technological change and certain types of economic stagnation, and when preventive war is infeasible, accommodation to the rising state may be the only feasible strategy. This is especially the case when accommodation is the only way to secure the trade needed to revive one's economic-technological base. The Beginning of Cold War Rivalries We have seen why, within an existing rivalry, a state might shift from ongoing deterrence policies to the more risky strategy of initiating a crisis. But why might states in a period of peaceful relations shift from a policy of calm engagement to one of hard-line deterrence/containment? The di- pt lemma is clear. Moving to a hard-line policy to contain the other is likely to • undermine any trust between the relevant great powers. Yet it may be the ; only way to prevent the other's steady rise. \ In this sense, cold wars (or "enduring rivalries") do not just happen.4" They are initiated, in the same way crises are initiated. And like crises, cold wars are recognized to be events that raise the probability of an inadvertent slide into major war (thus our collective sigh of relief in 1989-91).47 Cold [- wars may not have the immediate salience of events like the Munich crisis or the Cuban missile crisis. But by raising levels of suspicion, provoking arms races, and encouraging states to put forces on higher levels of alert, they not only make such crises more likely, they also increase the risks that 1 one side will see preventive and preemptive major war to be in its interest. i In this section I reiterate the logic of the model as it applies to the begin- ; ning of cold wars and compare it to the relevant alternative theories. The argument represents a synthesis of spiral-model reasoning on the risks of I cold war escalation and deterrence-model reasoning on the importance of ; maintaining power. I thus seek to bring the two models together within one common causal framework.48 A rising great power in a situation of peaceful relations has little problem deciding on its policy: it has no reason to disturb the current situation as t long as engagement continues to facilitate its growth.49 The declining state is in a more uncomfortable position, since it faces trade-offs similar to those : confronting a state having to decide whether to initiate a crisis. In choosing between staying with a peaceful, soft-line posture or moving to hard-line deterrence, a state will have its behavior shaped by the six variables/parameters outlined in figure 5. The greater the depth and inevitability of long-term decline in the absence of stronger action, and the more such action is expected to mitigate or reverse this decline, the more likely the state is to switch to a harder-line policy. Yet the state must also take into account the extent to which a hard-line policy will increase the probability of major war through inadvertent escalation. As the spiral model would emphasize, hard-line deterrence cannot be implemented in a vacuum. Given the security dilemma, such policies tend to be misintepreted as preparations for possible expansion.50 They therefore provoke a destabilizing action-reaction cycle.51 In making a rational decision, therefore, a security-maximizing state ; must recognize the upsides and the downsides of both soft-line reassurance . and hard-line deterrence. We can therefore anticipate that when decline is not expected to be deep or inevitable great powers will choose peaceful en-• gagement over containment. In the 1990s, for example, the United States |, pursued engagement toward China, and, as chapter 9 discusses, this was ? the logical and predictable policy. Yet, when profound decline can be anticipated should the state fail to switch to containment, then such a switch is The Origins of Major War likely. Truman's move to containment in mid-1945, even at a time when he felt that Stalin had relatively moderate intentions, shows the profound impact of such dynamic power trends. Critical to a state's determination of the prospects for long-term decline are the initial differentials and trends of economic power and especially potential power. In general, we can predict that the stronger the state is in economic and potential power, the more sanguine it will be about the future, and thus the less willing it will be to accept the risks of inadvertent escalation inherent in a cold war rivalry. The United States in the 1990s, for example, possessed advantages over China in technology, raw materials, and education. Thus, despite China's economic growth, there was no immediate reason to believe that it could overtake the United States in overall power. This situation was quite a bit different from the U.S. position versus Russia in 1945. America certainly was not in the dire position of Germany earlier in the century (when Germany had about one-third of Russia's population and one-fortieth of its land mass). The Soviet Union in 1945, however, already possessed a strong educational/technological base and a high level of industrialization. Hence there was good reason to worry that this state might overtake the United States if nothing was done to restrain its growth. In short, the degree of a state's superiority or inferiority in economic and potential power has much to do with its willingness to accept the risks inherent in hard-line policies. This analysis builds on the insights of both the deterrence and spiral models. It is important to note, however, that when realists of different stripes (and liberals) debate the strengths and weaknesses of these models, they tend to focus on the actors' military policies, not on the differentials and trends of economic and potential power. For defensive realists and liberals who emphasize the value of the spiral model, the security dilemma is fundamentally a function of each side's arms and alliance policies; it is one state's efforts to improve its security through such military means that is so frightening to the other side.52 My argument supplements this view by considering how states deal with actors who are not currently developing the kind of military power needed to attack but who might be able to build this power later, should their growth in economic and potential power be allowed to continue. In the traditional security dilemma underpinning the spiral model,'if state B is not building its military strength, state A should be relatively sanguine. In my model, even this situation can provoke state A to move to containment if the trends of economic and potential power are against it. The United States shifted to containment in 1945 even though it was known that the Soviet Union was drastically demilitarizing; the fear of Soviet growth in economic and potential power drove the hard-line policy. The argument 1 have outlined shows why spiral modelers and deterrence modelers need not be at odds with one another.53 Both sides, I believe, can [52] Foreign Policy Choices and the Probability of Major War accept the variables/parameters outlined in this chapter and the basic causal logic linking them. Both could agree, for example, that if a state is declining deeply, can do little about it through either hard- or soft-line policies, and will almost certainly be attacked later, preventive war, however unpalatable, is the option that will probably maximize the state's expected probability of survival. If decline is not at all inevitable, the other's intention to attack later not clear, and a hard-line policy would pose a fair risk of inadvertent war, then soft-line engagement is typically the best option. Alternatively, if decline would be deep and inevitable in the absence of stronger action but could be reversed by a hard-line posture, such a posture will likely be preferred so long as the risks of inadvertent war are not abnormally high and there is a fair chance the other might attack later if allowed to rise. The more severe the decline and the fewer the lower-level options, the more likely the state is to accept significant risks of inadvertent escalation. The disagreement between deterrence and spiral modelers theoretically—and between hawks and doves in the real world—is therefore less about causal logic, and more about the exact values of the variables and parameters.54 Consider U.S. policy versus China at the turn of the new millennium, for example. Hawks and doves could agree that if it could be known for sure that China would become clearly preponderant in twenty years and would be aggressive at that time, then preventive containment would be advisable now, notwithstanding the risk of sparking a new cold war. Yet doves reject both premises. They would thus conclude that peaceful engagement is preferred, at least until such time as China's long-term growth and future aggressiveness seem inevitable. In sum, both the deterrence model and the spiral model have part of the puzzle right. Yet the models remain disconnected and incomplete. The deterrence model correctly notes that hard-line policies can avert decline, but it ignores the spiral-model point that these policies can increase the probability of war through inadvertent escalation. The spiral model captures this risk. Yet it minimizes the potential downside to reassurance policies, namely, that they may permit exogenous decline to continue, thereby leaving the state less able to defend itself later. By integrating the risk of inadvertent spiraling into a primarily power-driven model, this book specifies more clearly under what conditions states will remain peacefully engaged, and under what conditions they will fall into cold war. This chapter has sought to show how incorporating the insights of the crisis and security-dilemma literatures can help build a more powerful dynamic realist approach to major war. Leaders are not ignorant of escalation effects. Most obviously, they know that if they provoke a crisis by challenging another's vital interests, things may get out of hand. Crises are thereto! The Origins of Major War m fore likely only when less risky options will not achieve the state's ends. | Leaders also know that if they adopt hard-line policies during periods of relative calm, the other may interpret these acts as hostile, leading the actors into a cold war rivalry. U.S. leaders today are aware of the risks of trying to contain China, just as Truman was in 1945. Dynamic differentials theory helps answer the question of why leaders, knowing the risks of hard-line policies, would ever choose them over more accommodative strategies. Hence, one need not fall back on domestic-level forces or leader misperception to see why states sometimes take such gambles. Rational security-seeking actors pursue the option that maximizes the state's expected probability of survival. When a state is not in decline, or is in fact rising, peaceful engagement is usually the best strategy. Declining states, however, face different constraints and therefore have different induced preferences. The more severe the state's deteriorating position, the more likely it is, all things being equal, to adopt severe policies with a high risk of inadvertent escalation in order to avert further decline. This analysis thus allows us to predict changes in the probability of major war over time through a power-driven systemic logic. Summary of the Competing Hypotheses In the following six chapters, I test the hypotheses about major war from dynamic differentials theory both against current realist theories and against other explanations in the literature. But first it will be useful to summarize the main competing realist hypotheses. Evidence that would falsify a theory or reduce its salience is specified in parentheses.55 Dynamic Differentials Theory 1. Leaders perceiving their states to be in decline will be the initiators of major wars or of crises/cold wars that increase the risk of inadvertent major war. The more severe a state's decline will be in the absence of strong actions, the more likely the state is to initiate such actions. In general, therefore, the greater the declining state's inferiority in economic power and potential power, the more likely it is to pursue highly risky policies. (Falsified if rising states are the initiators of major wars and crises/cold wars. Falsified if the greater a state's rise, the more it is willing to take such risks. Falsified if steeply declining states with the requisite military power [points 2 and 3 below] do not initiate war or crisis/cold war, even when such policies can avert decline. Salience reduced if states begin major wars or ; crises/cold wars for unit-level reasons.") 2. In multipolarity, the declining state will be more likely to initiate ma- • Foreign Policy Choices and the Probability of Major War jor wars or crises/cold wars that increase the risk of inadvertent major war when it possesses marked military superiority versus the other great powers taken individually. The more marked its military superiority, the more likely it is to initiate such actions. (Falsified if the state initiates major war or highly risky policies with only military equality or less.) 3. In bipolarity, the declining state will be more likely to initiate major wars or crises/cold wars that increase the risk of inadvertent major war if it is either preponderant or near-equal in military power. The second-ranked great power may also initiate such policies if it is declining, but the greater its level of inferiority, the less likely it is to do so. (Falsified if the second-ranked state attacks even though significantly inferior and when decline is not that severe.) Classical Realism 1. Major wars and crises/cold wars raising the risk of inadvertent major wars are most likely to occur when one state has a preponderance of power. Such events should not occur when there is a balance of power between individual states or between alliance blocs. (Falsified if these events occur when individual great powers are in relative balance or their alliance blocs are in balance.) 2. Major wars and crises/cold wars are started by a state with aggressive unit-level motives for expansion. (Salience is reduced if the initiator is only or primarily seeking security.) Structural Neorealism 1. Major wars and crises/cold wars raising the risk of inadvertent major war are more likely to occur in multipolar systems than bipolar systems. (Falsified if such events are as common or more common in bipolar systems.)56 Hegemonic Stability Theory 1. Major wars and crises/cold wars raising the risk of inadvertent major war are most likely to occur when the two most powerful states in the system are roughly equal and one state is overtaking the other.57 (Falsified if such events occur when one state is preponderant.) 2. The initiator of major wars and intense crises/cold wars should be the rising state. It will initiate only when, because of unit-level factors, it is dissatisfied with the status and rewards provided by the system. (Falsified if such events are initialed by the declining state. Salience is reduced if initiator is solely or primarily seeking security.)