The logic of the security dilemma

The security dilemma is a structural theory that purports to explain why wars can happen even among states that seek nothing more than their own security.Footnote 2 It asserts that ‘the means by which a state tries to increase its security decrease the security of others’ (Jervis, Reference Jervis1978). This is an obvious consequence of the fact that conventional military power is relative, not absolute, so that how much security a given inventory of military force provides depends on what others possess. Moreover, how much security is gained by the acquisition of more weapons depends on how others respond. Here, the decrease in others’ security is an inadvertent and unavoidable, not intended, effect. The core logic of the security dilemma is grounded in its view of security as a good that is both: (1) positional (or zero-sum), such that gains for some mean corresponding losses for others, and (2) scarce; how scarce depends on a number of factors that determine the offense–defense balance.Footnote 3 If security can be mutually shared and simultaneously increased, then there is no security dilemma; security is, instead, a plentiful, if not free, good.

These features of the security dilemma nicely illustrate the more general phenomenon known as perverse system effects. Complex systems, like the international system, often generate outcomes that do not follow actors’ intentions (systems produce unintended consequences). Because states operate within a setting of strategic interdependence (given the interconnectedness that defines systems), they can never do just one thing; for example, increase their security without decreasing the security of others. Their fates, therefore, are strongly influenced by complex and often unpredictable interactions (Jervis, Reference Jervis1997: Ch. 1 and 2). As Jervis puts it: ‘When the security dilemma is at work, international politics can be seen as tragic in the sense that states may desire – or at least be willing to settle for – mutual security, but their own behavior puts this very goal further from their reach’ (Jervis, Reference Jervis2001). This claim, however, seems to violate the basic premise of the security dilemma that security is, in fact, a zero-sum good. My view, one that I have consistently voiced since 1996, is that security, under most conditions, is not a positional good; it can be enjoyed and shared by all who want it.Footnote 4 The only condition under which security cannot be shared by pure security seekers is when there are intense preemptive incentives, that is, when offense has a large advantage over, and is indistinguishable from, defense. When these conditions obtain, the only route to security and survival is through attack. Because the future must be discounted entirely (there is no future if the other attacks first), it is essentially a one-shot Prisoners’ Dilemma (PD) game, wherein players must defect.

In addition to the relative nature of conventional military capabilities and the systemic constraint that prevents states from doing just one thing, security dilemmas exist, according to those who champion the theory, because weapons can be used for both offensive and defensive purposes, which makes it difficult, if not impossible, for status quo states to signal their defensive intentions through force postures and weapons acquisitions. The ambiguity of dual-use weapons means that others cannot be certain whether the increase in arms was for defensive or offensive purposes. When this uncertainty over the intent of the arms build-up is combined with the certain loss in relative power that results, mutual suspicions and fears can arise, and given the inherently unpredictable interactions of complex systems, escalate to war. It is the very nature of the international system and the means by which security is gained under these conditions that generate intense fears and uncertainties, which, the theory claims, explain how unwanted and unforeseen wars can, nevertheless, arise among pure security seekers.

To be perfectly clear, I am simply attempting here to faithfully explain the theory according to the logic expressed by those who have championed it as a major cause of conflict and war. I do not agree with all or even most of these premises and claims. For instance, I do not accept the premise that, under most conditions, uncertainty and mutual suspicions and fears among pure security seekers will generate arms races and war. To get such an outcome, one must add variables to the theory; for example, the fundamental attribution error, prospect theory, and other psychological theories about decision-making. I will return to this topic later in the paper.

That noted, Wagner reduces the security dilemma to just two premises: ‘An increase in one state’s ability to protect itself from attack by others will diminish the ability of other states to protect themselves from an attack by the first state’; and ‘No state can ever be certain that another state will not use force against it’ (p. 26, emphasis in original). The latter premise is not specific to the security dilemma and the former premise, Wagner believes, ‘merely states the obvious fact that only relative, not absolute, military capabilities are important for a state’s security’ (p. 26). Even this premise, he points out, might not be true because ‘the condition that came to be known as mutual assured destruction (MAD) seems to imply that it is not true of nuclear weapons’. Surely Wagner realizes that Jervis – who spent most of the 1980s attacking the dangerous rationalizations behind the ‘countervailing strategy’ and the nostalgic belief that nuclear weapons could be brandished, as conventional weapons once were, as instruments of foreign policy – understands that relative power concerns do not affect nuclear weapons. After all, this basic distinction between nuclear and conventional weapons was captured in the title of the very first book on nuclear strategy, The Absolute Weapon: Atomic Power and World Order (Brodie, Reference Brodie1946).Footnote 5

Despite this fundamental difference in nuclear and conventional forces, the security dilemma still applies in the nuclear age, and so we need the following additional logic to explain how this can be true: The security dilemma varies in intensity depending on ‘whether defensive weapons and policies can be distinguished from offensive ones, and whether the defense or the offense has the advantage’ (Jervis, Reference Jervis1978: 186). One of three fundamental changes, in addition to mutual kill and speed of kill, wrought by nuclear weapons is that offense became defense, and defense became offense. Hence, the logic of ‘MAD’ requires both superpowers to forego first-strike capabilities in favor of a secure ‘assured destruction’ second-strike capability. This formula for stable strategic nuclear deterrence turns the conventional concepts of offense and defense on their head: offensive capability (the ability to assure the other’s destruction and thereby take its population hostage) strengthens deterrence stability and the status quo; while defensive capability (the ability to protect the state against total destruction and free its population from being held hostage by the other side) weakens deterrence and thereby creates instabilities that can serve offensive purposes.

Leaving nuclear weapons aside, there are two other core conditions of the security dilemma that Wagner overlooks. The first is a scope condition that is key to understanding the theory and its limitations: the security dilemma only applies to pure security seekers, whether or not they can signal their intentions as such. If states are arming because there is an actual aggressor or because of real, not imagined, non-security conflicts of interest among them, then there is no security dilemma. When there is a true aggressor or non-security conflicts of interest, the observed reciprocal increase in arms and hostility is merely a reflection of underlying and irreconcilable conflict, not a cause of it. These situations are explained by the theories of balance of power and deterrence, not the security dilemma. This scope condition, however, raises the question: if all states must be pure security seekers, how can a PD payoff structure arise among them? In other words, why would any of them prefer DC to CC? The reason is that, under certain conditions, they cannot signal their true intentions as pure security seekers; and the external environment is so vicious that they must assume the worst about each other because guessing wrong could mean extinction. Accordingly, even though they are pure security seekers (and must be in order for security-dilemma logic to apply), they can prefer DC to CC – to attack rather than mutually cooperate – for purely defensive reasons; there is no need to assume that they choose to defect for greedy, non-security benefits. All that is required is a positive probability that the other side is indeed a non-security seeker.

Put differently, the security dilemma requires that there be some fear that greedy types exist among the actors; it also requires that, in fact, there are not greedy types among them. It is this misperception among the actors that there is a positive probability that some may be greedy actors that drives the security dilemma. As I wrote years ago: ‘At bottom, the security dilemma…rests on the assumption that some states are misperceived to be either currently harboring aggressive designs, or that they may become aggressive in the future. In a hypothetical world in which states are known to be status quo and cannot be otherwise in the future, it would make no sense to say that offensive advantage dictates “that the only route to security lies through expansion” ’ (Schweller, Reference Schweller1996: 119).

The second logic is more complex: the security dilemma is a PD when and only when offense has the advantage and is indistinguishable from defense. At all other times, the security dilemma is a Stag Hunt and, therefore, not a dilemma at all. The logic behind this widely unrecognized but crucial deduction is explained at length below.

Problems with security dilemma logic

Before proceeding, let me emphasize at the outset that Wagner and I agree but for very different reasons that there are problems with the theory’s logic. In one of the seminal works on the security dilemma, which Wagner rightly calls ‘one of the most influential articles about international politics ever written’ (p. 27), Robert Jervis clearly states that the security dilemma is not a Stag Hunt: ‘The third problem present in international politics but not in the Stag Hunt is the security dilemma’ (Jervis, Reference Jervis1978: 169). Instead, Jervis models it as a PD. This has become the standard view of the security dilemma ever since, and it is why, in my view, the theory as typically presented is not entirely logical. Wagner claims that Jervis believes that ‘a repeated PD could be represented by the 2 × 2 game often called the Stag Hunt’ (p. 30), and that this transformation of the situation from a PD to a Stag Hunt is a possible solution to the security dilemma. He does not provide a citation for these claims, however, and I cannot find any passage where Jervis says that a repeated PD is a Stag Hunt or that Stag Hunt provides a solution to the security dilemma.Footnote 6 Indeed, Jervis has assured me in a personal correspondence that he never said that a repeated PD can be treated like a Stag Hunt. One of the problems with Wagner’s discussion of realism is that it does not appear that he circulated the manuscript for comments to any of the contemporary realists that he critiques in the book. Had he done so, errors and misunderstandings like this one could have been easily avoided.

Jervis’s solution resides in the offense–defense balance, which does not alter the preference structure (the game itself) but rather the payoffs within a PD. Defensive advantage increases the chances for mutual cooperation in a PD because: (1) the gains from DC decrease and are not much higher than CC; (2) the costs of CD decrease and are not much worse than DD; and (3) both the costs of DD and the gains from CC increase. Offensive advantage decreases the chances for cooperation because: (1) the gains from DC increase and are far higher than for CC; (2) the costs of CD increase and are much higher than for DD; and (3) there is little difference between the gains from CC and the costs of DD (presumably because CC cannot hold for long).Footnote 7

Of course, cardinal utilities do not change the nature of the game. A PD structure is a PD structure, regardless of the cardinal utilities. Jervis is suggesting that, if the game is iterated, then moves are, in a sense, reversible. Thus, when defense has the advantage and can be distinguished from offense, players can learn to cooperate because they will be less inclined to heavily discount future gains from mutual cooperation. Conversely, when offense has a huge advantage and cannot be distinguished from defense, all players will defect as if it is a single-shot game. This is an interesting formulation of the security dilemma, but I am not persuaded that it accurately models the dynamics of the situations. I believe, instead, that the game itself changes from PD to Stag Hunt depending on the offense–defense balance and whether offense can be distinguished from defense. When the structure of the game is Stag Hunt, which it typically is, there is no security dilemma; instead, there is a security issue. This has not been understood in the literature. Here, Wagner and I seem to agree. I will return to this subject.

In Wagner’s view, there is no relation between the security dilemma and the PD: ‘Unfortunately, the only connection between the security dilemma and the PD is that they both have the word dilemma in their names. Therefore, like his discussion of the offense–defense balance, the additional plausibility that Jervis’s use of the PD gave to the idea that anarchy made peace among independent states unlikely was quite unwarranted’ (p. 29). Wagner gets this entirely wrong, I believe. Security dilemmas can, indeed, be modeled as one-shot PDs, though most security dilemmas are, as I argue below, Stag Hunt situations and, when PD is the correct model for the security dilemma, war is virtually inevitable.

The problem is not that the PD version of the theory is illogical but rather that PD-type security dilemmas are historically very rare and should not be generalized.Footnote 8 The security dilemma is only a PD when there are huge first-strike advantages or, more broadly, when security can only be achieved through the pursuit of superiority and domination – situations that rarely obtain in the real world.Footnote 9 In all other situations, the security dilemma is not really a dilemma; it is, instead, a problem of mistrust under conditions of uncertainty described by Rousseau’s Stag Hunt parable. Let me explain.

In theory and practice, the security dilemma varies in intensity depending on whether state-of-the-art military technology favors offense or defense. How one views the ‘normal state of affairs’ with respect to the intensity of the security dilemma has, rightly or wrongly, become the fault-line distinguishing offensive from defensive realism. Defensive realists view the security dilemma as mild to moderate in intensity, such that security is relatively plentiful under normal conditions and can be shared by security-seeking states. Thus, the idea that security can be enhanced through expansion is a counterproductive myth that results from unit-level pathologies, not the structural logic of the international system (Snyder, Reference Snyder1991: Ch. 1). Strategies of conquest, whether motivated by greed or survival, are ultimately self-defeating because they provoke powerful counter-coalitions, which, after victory, severely punish and sometimes eradicate aggressors, especially predators.

Conversely, offensive realists portray a world in which the security dilemma is always intense or threatening to become so. In this world, security is at all times scarce and states can never have enough power because all great powers possess some offensive capability, because intentions are inherently unknowable and can change, and because states can never know who, how many, and how strong their future rivals will be. These structural uncertainties and the self-help, competitive nature of international politics compel states, even pure security seekers, to act like revisionist states, increasing their power until they have achieved a measure of hegemony. Conquest is, therefore, an essential activity of great powers that pays generous rewards when not done recklessly.Footnote 10

Unlike defensive realists, offensive realists portray normal international politics as an intense security dilemma, which engenders powerful preemptive incentives and spiraling arms races. It is the proverbial powder-keg situation, where any spark can explode into war.Footnote 11 The security dilemma is most acute when offense has the advantage and is indistinguishable from defense. Here, the impossibility of signaling one’s own benign intentions combined with the staggering costs of guessing wrong about the intentions and future actions of others explains how uncertainty, via security dilemmas, can lead to aggressive behavior and preemptive wars that no one wants. Such conditions give rise to the ‘reciprocal fear of surprise attack’ and other preemptive incentives that compel all states to act like aggressors and assume the worst of others.Footnote 12 States operating in an environment of large first-strike advantages confront the same problem faced by two gunslingers in a small town lacking a capable sheriff. Both gunslingers may prefer a bargain whereby each leaves the other alone, but neither side can credibly commit not to shoot the other in the back.Footnote 13

Furthermore, when offense has the advantage: wars will be short, decisive and, profitable for the winner; ‘there will be incentives for high levels of arms, and quick and strong reaction to the other’s increases in arms’; ‘states will have to recruit allies in advance’ and form tight alliances; and statesmen ‘will be quick to perceive ambiguous evidence as indicating that others are aggressive. Thus there will be more cases of status-quo powers arming against each other in the incorrect belief that the other is hostile’ (Jervis, Reference Jervis1978: 189–190). This preference structure is clearly a PD, in which all sides prefer to defect even if the other side cooperates. It is not structural uncertainty that drives preemptive behavior but rather certainty that: (1) the gains from exploitation (DC) are very high; (2) the costs from getting suckered (CD) are intolerable (certain defeat and possible extinction); and (3) pre-war bargains (CC) cannot be enforced or relied upon to hold. Under these conditions, the core problem for security seekers is that huge first-strike advantages do not allow them to share lasting security, which can only be achieved by attacking first and defeating the other side (DC). While all may desperately seek to avoid war, none can credibly commit not to strike first.

Wagner not only rules out the PD as a model for the security dilemma but goes further to rule out the security dilemma as a cause of war: ‘But even if the security dilemma could explain why war occurs (which, as we have seen, it cannot)’ (p. 35). Later, he writes:

What does this mean? Why cannot wars occur among states that prefer the prewar status quo? Where did we see ‘that this is not true’? The problem with Wagner’s analysis is that his discussion of the security dilemma ‘assumed that what is at issue is the distribution of valuable territory among predatory rulers’ (p. 184). We know, however, that predatory goals and real non-security conflicts of interest are inappropriate assumptions with respect to the security dilemma. Security dilemma theory is not about states that simply prefer the prewar status quo. As Wagner himself points out in a footnote that ends the above quote, ‘even predatory rulers competing for valuable territory might both be satisfied with the status quo’ (fn. 12: p. 182). Of course, this is how deterrence theory and balance of power work when they prevent war (which, by the way, is not the main aim of balance of power, since war must be a legitimate tool of statecraft for the theory to operate). What distinguishes the security dilemma from deterrence theory and balance of power is precisely that the former involves only states that seek security aims; there can be no predatory states that use or threaten military force when an opportunity arises to achieve their non-security aims. Under certain conditions, as Wagner points out, even predatory states with non-security goals will prefer the status quo to the costs of trying to change the status quo by means of force. This is obvious – even to contemporary realists – and obviously not a situation that applies to the logic of the security dilemma. While later on Wagner acknowledges that the contemporary literature aims to be more general than his focus on territorial disputes among predatory rulers (p. 201), this caveat cannot excuse his use of predators in the context of the security dilemma; the theory is not just more general than this focus; it rules it out altogether.

In the second part of the passage above, Wagner says: ‘However, it [wars among status-quo states] can be true of both preemptive and preventive wars (though it need not be) – the state whose attack is preempted might actually not have intended to attack, and the state whose increase in power is prevented might never have challenged the status quo’ (p. 182). Is not this the logic of the security dilemma? So does he agree that security dilemmas can cause war or disagree with this statement? One never knows. In addition, this is precisely how the book reads throughout. One moment, the link between the security dilemma and war is not true; then, in the very next sentence, it appears to provide a logical explanation for how preemptive and preventive wars can occur when offense has the advantage. But then, on the very next page, Wagner says (note the length of the first sentence in this paragraph):

There are several problems with this passage. First, the security dilemma is not a theory about contests of disarmament. It is about how states that seek nothing more than their own security (which is not the same as saying that they want to maintain the prewar status quo) sometimes must ‘act’ like aggressors when the only path to security is through aggression. Second, the security dilemma does not claim that both sides prefer the status quo to war. Indeed, the security dilemma predicts that, when offense has a large advantage over defense and cannot be distinguished from defense, security seekers must attack in order to survive; it is the unsustainable and vicious nature of the status quo that compels them to seek to replace it with a more stable situation. Third, the last sentence in the above passage reiterates the essential logic of the security dilemma; therefore, once again, does Wagner agree that the security dilemma can cause war or disagree with this statement?

Finally, nowhere in Wagner’s discussion of the security dilemma is there a mention of whether offense can be distinguished from defense. How can this be so? After all, whether offense can be distinguished from defense is one of only two dimensions that determine the severity of the security dilemma (the other being the offense–defense balance). This second dimension is extremely important because the ability to distinguish offensive from defensive weapons and force postures goes to the very core of whether states can make their intentions known to others. If everyone knew that everyone else was merely a security seeker and nothing more (which must be true for security dilemma logic to apply), and if they knew that none could become a future predator (at least, not without giving others sufficient advance warning to protect themselves), then the security dilemma would cease to exist, regardless of the offense–defense balance. Precisely, because states do not know that they are all security seekers and they cannot make known their benign intentions, extreme offensive advantage often winds up in preemptive war. In addition, this is why realists view wars that arise from security-dilemma dynamics as tragedies; it is not that these wars are inadvertent or accidental but rather that they are the logical consequence of the structure of the situation, which means that wars can occur among states even in the absence of predators with greedy intentions.

Curiously, Wagner seems to dismiss (or does he?) the logic of the security dilemma as a route to preemptive war even though this ‘PD version’ of the security dilemma exemplifies the formal bargaining approach to war and international politics that he so strongly advocates. Formal bargaining theories of war explain, among other things, how various commitment and informational problems prevent perfectly rational actors from reaching ex ante bargains that both sides would prefer to the costs and risks of fighting.Footnote 14 Within this rationalist framework, it is not anarchy and uncertainty per se that cause war but rather actors’ incentives: (1) to misrepresent private information, which prevents them from sharing it (hence, agents reach conflicting estimates about the outcome of war) and (2) to renege on unenforceable bargains given certain state preferences and opportunities for action (the inability to make credible commitments).Footnote 15 When offense has the advantage and cannot be distinguished from defense, status quo states are unable to signal their benign intentions and guessing wrong about others’ intentions is fatal. The core problem is that, in an anarchic environment of self-enforcing agreements, states cannot credibly commit not to attack. Accordingly, when the security dilemma is at its most vicious, ‘the only route to security lies through expansion [or attack]. Status-quo power must then act like aggressors; the fact that they would gladly agree to forego the opportunity for expansion [or attack] in return for guarantees for their security has no implications for their behavior’ (Jervis, Reference Jervis1978: 187).

When defense has the advantage, the reverse is true: wars will be long, indecisive, and unprofitable; there is no need to match or respond quickly to the other side’s increase in arms; states will not seek tight alliances in peacetime (they may steer clear of alliances altogether); and leaders can afford to interpret ambiguous information in a benign way because guessing wrong about others’ intentions is not especially costly, much less fatal. There is nothing about these incentives that suggests exploitation (DC) will or should be preferred over mutual cooperation (CC), just the opposite. When defense has the advantage, the preference structure of a security dilemma is a Stag Hunt, not PD. Yet, this essential fact has gone unrecognized in the literature.

In his most recent work on the security dilemma, Jervis introduces what he calls a ‘deep’ security dilemma. This is what I have been referring to as the PD variant of the security dilemma. Note that, in his description of a deep security dilemma, Jervis differentiates it from a security dilemma ‘based on mistrust’ – the Stag Hunt problem:

In other words, a security dilemma is ‘deep’ when security requires or is thought to require superiority over, not equality with, the rival; that is, when security can only be achieved by means of domination and exploitation of the rival, which is a PD situation because DC is preferred to CC. It is structural situation driven by huge preemptive incentives, where mutual security cannot be achieved and is, therefore, not an option.

Conversely, when the security dilemma is less intense (when defense has the advantage and/or is distinguishable from offense), then the preference structure is Stag Hunt, not PD.Footnote 16 The problem is not a difference in preferences over outcomes (both states prefer CC to DC) but rather mistrust: security seekers cannot trust each other to cooperate to achieve their most preferred outcome, mutual security. In these situations, which I will call ‘shallow’ security dilemmas, the main barriers to cooperation are incomplete information about the other’s intentions (e.g. is its first choice CC, DC, or DD? is CC preferred to DD, etc.?) and the implications of cheating due to inadequate monitoring and enforcement mechanisms. When defense has the advantage or is not entirely trumped by offense, there is no reason to believe that status-quo powers will prefer DC over CC or will have to act like aggressors. Indeed, security seekers have powerful incentives not to exploit or dominate others, viz., they lose their reputation as status-quo powers and recklessly provoke the formation of powerful coalitions against them. Instead, mutual security can be achieved as long as states trust others to reciprocate cooperation for cooperation (Kydd, Reference Kydd2005). This is not a dilemma but rather a Stag-Hunt coordination problem that can and should be solved – given the high incentives for status-quo states to avoid unnecessary wars – through communication and signaling.Footnote 17

Why, then, would pure security-seekers under conditions other than extreme offensive advantage wind up in a security dilemma? What prevents them from solving the Stag Hunt problem and achieving their shared preference for mutual cooperation? After all, cooperation allows each to gain security and avoid a costly and dangerous arms race – one that is entirely pointless given the actors’ true intentions. Why would they allow their behavior to put this very goal further from reach?

Shallow security dilemmas and the spiral model of conflict

To explain how shallow security dilemmas can have deep security dilemma effects, realists bring in unit-level variables from cognitive psychology. The familiar spiral model of conflict is essentially the security dilemma with a psychological overlay that explains how conflict and wars can arise among pure security seekers under normal security dilemma conditions, that is, when structural factors are not intense enough to determine state behavior (to use Wolfers’s famous phrase, when the house is not on fire) (Jervis, Reference Jervis1976: Ch. 3). The spiral model’s prime causal mover is what cognitive psychologists call, The Fundamental Attribution Error, which alleges that people tend to make situational attributions to explain their own disagreeable behavior and dispositional attributions to explain others’ unwanted behavior. This has two important consequences for arms spirals. On the one hand, leaders understand their own arms increases as appropriate defensive responses to a dangerous external environment and expect others similarly to attribute their behavior to this motivation. On the other hand, they interpret the arms buildups of their neighbors as clear indications of aggressive intent. In Jervis’s words: ‘If the state believes that others know that it is not a threat, it will conclude that they will arm or pursue hostile policies only if they are aggressive. For if they sought only security they would welcome, or at least not object to, the state’s policy…. The benefit [of explicating these psychological dynamics] is in seeing how the basic security dilemma becomes overlaid by reinforcing misunderstandings as each side comes to believe that not only is the other a potential menace…but that the other’s behavior has shown that it is an active enemy’ (Jervis, Reference Jervis1976: 71, 75). This suggests that, as a security dilemma unfolds and becomes more intense, it is driven not by uncertainty but by misplaced certainty that the other is a true aggressor.

The relationship between the security dilemma and spiral model, therefore, may be one of sequencing, in which affective content and cognitive closure increase over time. Sometimes, leaders hold longstanding but mistaken beliefs that the other side is an enemy and interpret its behavior exclusively through this lens. When it builds arms, therefore, it must be for offensive reasons. A spiraling arms competition unfolds as a self-fulfilling prophecy grounded in misplaced certainty that the other intends to do harm and acts solely with this purpose in mind. At other times, leaders are initially uncertain about the potential rival’s intentions. As political pressure builds to stand firm and respond with costly actions, however, these leaders are increasingly tempted to rid the situation of ambiguity by attributing the other’s unwanted behavior to dispositional rather than situational factors. As sincere belief and tactical arguments become blurred by subconscious psychological processes, leaders will genuinely misinterpret the other’s arms increase as a clear sign of aggressive intent and discredit information that disconfirms this core belief. One wonders how a costly and intense arms race could occur among pure security seekers without these certain judgments about the target’s intentions – judgments that must be incorrect for the security dilemma to apply. In other words, leaders must be guilty of misplaced certainty for normal security dilemmas to cause intense conflict and war: they must be certain (or near so) that the other is an aggressor, and they must be wrong. None of this logic appears in Wagner’s discussion of the security dilemma. Moreover, the spiral model’s psycho-logic defies the rationalist foundation of formal bargaining theories of war that Wagner champions.

Final thoughts

Leaving aside the security dilemma, realists have theorized about much of what Wagner claims that they have not appreciated, failed to understand, or ignored. While reading the book, I did not find myself disagreeing with most of what it says but rather with what Wagner claims realists have said. Much of it was very familiar. Consider, for example, Wagner’s reconsideration of anarchy and hierarchy. Here, Wagner criticizes Waltz’s notion of anarchy as confused because it conflates decentralized institutional structures with the absence of a leader or ruler. The correct oppositions should not be anarchy and hierarchy, as Waltz maintains, but rather anarchy and some form of ‘archy’ such as monarchy or oligarchy. Is this really a distinction with a difference? Wagner says that anarchy ‘refers to the absence of a leader or ruler, which structural Realists equate with the absence of any institutional structure’ (p. 122). Where does he get the idea that realists equate anarchy with the absence of any institutional structure? Waltz himself writes: ‘Since world politics, although not formally organized, is not entirely without institutions and orderly procedures, students are inclined to see a lessening of anarchy when alliances form, when transactions across national borders increase, and when international agencies multiply’ (Waltz, Reference Waltz1979: p. 114). A few pages later, however, Wagner claims something different: ‘The difference between the existence of a monopoly of the legitimate use of force within states and the absence of such a monopoly among them is what Waltz really had in mind when he distinguished between hierarchy and anarchy’ (p. 127). Now structural realists do not ‘equate anarchy with the absence of any institutional structure’ but with the absence of a monopoly of the legitimate use of force among states. Which is it?

Wagner concludes the section, ‘Anarchy and Hierarchy Reconsidered’, with the claim: ‘A history of modern international politics told as the history of peace settlements would be a history that revealed the institutional development of the European state system, which was eventually extended to encompass the globe…. States are the product of a process by which groups of individuals with well-defined identities use violence to bargain over the institutional structures that will regulate conflicts among them’ (p. 125). How is this story different than the structural realist one told by Robert Gilpin in War and Change in World Politics or the modified version of it told by John Ikenberry in After Victory? (Gilpin, Reference Gilpin1981; G. John Ikenberry, Reference Ikenberry2001).

Wagner ends this chapter on violence, organization, and war with the following observations: ‘Thus not only can states with a monopoly of the legitimate use of force within their territories make war with other states; they can make peace with them as well, which they also do. This is something that organizations without a monopoly of the legitimate use of force are unable to do. The institutional structure of a state system does not tell us why the peace that states make among themselves could not be as lasting as the peace some states, but not all, have made within their territories’ (p. 129). Structural realists would have no problem with the first part of this paragraph. The last sentence, however, does not seem to follow logically from the first two. How does the fact that sovereignty enables states to make peace with each other undermine the other consequences of anarchy that realists have emphasized that distinguish international politics from domestic politics? How does the ability of sovereignty to facilitate peace settlements contradict the realist claim that anarchy is a self-help system? that, under anarchy, agreements are self-enforcing? that there is no 9-1-1 that states can call when they get in trouble, and so force is not only the ultima ratio in international politics but the first and constant one? and, because war always lurks in the background, states must be primarily concerned with their relative gains and losses, which makes cooperation difficult to achieve and harder to maintain? I simply do not understand how Wagner’s logic about the benefits of sovereignty for the creation of peace undercuts realist claims about anarchy and its consequences and how the anarchic realm of international politics distinguishes it from domestic politics. This does not mean that Wagner is incorrect, of course, but I suspect other readers will be scratching their heads as well. Failing to see Wagner’s logic on this critical point, I cannot accept the basic premise of the book, namely, that the study of international politics has been impoverished by this distinction and must, instead, be part of the more general study of the relation between political order and organized violence.

Let me conclude by pointing out that I have not even attempted to provide a comprehensive and balanced review of Wagner’s book. I have, instead, narrowly focused on his critique of contemporary realism, aiming to show that the logics and whatever flaws lie therein are more subtle and complex than Wagner’s treatment of them. That said, serious students of international relations should not only read War and the State but come to terms with its arguments. It is one of those rare works, like Waltz’s Theory of International Politics, Gilpin’s War and Change, Keohane’s After Hegemony, and Wendt’s Social Theory of International Politics, about which one must have a reasoned and informed opinion, whether pro or con.Footnote 18