The ORE of Politics will be available for subscription in late September. Speak to your Oxford representative or contact us to find out more.

Show Summary Details

Page of

PRINTED FROM the OXFORD RESEARCH ENCYCLOPEDIA, POLITICS ( (c) Oxford University Press USA, 2016. All Rights Reserved. Personal use only; commercial use is strictly prohibited. Please see applicable Privacy Policy and Legal Notice (for details see Privacy Policy).

date: 21 September 2017

Arms Races: An Assessment of Conceptual and Theoretical Challenges

Summary and Keywords

An “arms race” is a competition over the quality or quantity of military capabilities between states in the international system. The arms race phenomenon has received considerable attention from scholars over many decades because of the ubiquity, throughout history, of states building arms as a means of deterring enemies, but disagreement persists over whether that policy is effective at avoiding war.

The Latin phrase si vis pacem, para bellum, meaning “if you want peace, prepare for war,” dates back to the Roman Empire but the sentiment is likely much older. That states should rapidly build up their militaries in the face of potential threats is a common thread that runs through much of the modern international relations scholarship influenced by realism and deterrence theory. Meeting force with force, the logic went, was the only way to ensure the security or survival of the sovereign state. These states faced a paradox, however, best articulated by the “security dilemma.” Anything a state does in the name of defense, like a rapid military buildup, decreases the security of other states and will be viewed with hostile intent. This set up a debate over competing expectations regarding the relationship between arms races and war (peace). On one hand, deterrence theory posits that rapid arming is necessary to raise the cost of an adversary attacking and, consequently, preserves peace. On the other hand, the spiral model argues that the reality of the security dilemma means that arming produces mistrust, hostility and, thus, increases the likelihood of war. Scholars set out to test these competing hypotheses using large data sets and statistical techniques, but there was widespread disagreement on how to measure arms races, appropriate research design, and the statistical findings were somewhat mixed.

Critics of this approach to studying arms races note a number of important weaknesses. First, scholars primarily focus on the consequences of arms races—whether they lead to war or peace—at the expense of understanding the causes. Those who advance this position believe that a theory of arms race onset might well inform our understanding of their consequences. Second, security dilemma, taken as the primary motivation for arms races, suffers from significant logical flaws. Third, assessment of the arms race-war relationship consists of comparative theory tests of deterrence theory and spiral model, yet these ideas are underdeveloped and expectations oversimplified. More recently, scholarship has shifted the focus from the consequences of arms races to developing theories and empirical tests of their causes. These efforts have been informed by insights from bargaining models of war, and their application to this context holds promise for better future understanding of both the causes and consequences of arms races.

Keywords: arms races, security dilemma, spiral model, deterrence theory, bargaining, uncertainty, interstate war, empirical international relations theory


“Arms races” are competitions over the quality or quantity of military capabilities between sovereign states in the international system. An arms race is an interdependent event—two states competing over their relative power—but this outcome is produce by domestic political decisions of two independent states (Anderton, 1989). How states move from independent decisions related to maintaining a standing army to engaging in an interdependent, competitive arms race, and the consequences of these arms races, are topics long engaged in by foreign policy and international relations scholars. One reason for this focus is because maintenance of a standing army is viewed as necessary for general deterrence against potential and realized enemies. Indeed, the existence of the state is driven in part by the desire of people to have security: security most often provided by the presence of a state sponsored military. Yet decisions to surge state power through accelerated arms acquisitions and expansion of armies, even in the name of defense, can have unintended consequences relative to other states in the international system. Research has associated rapid military buildups with the destabilization of bilateral relations and even war. Despite considerable attention, our understanding of why states move to rapidly increase their military power, how arms races emerge, and the consequences of those competitions is still incomplete.

The purpose of this article is to provide a review of a subset of the arms race literature with particular emphasis on the conceptual and theoretical challenges of this research agenda, to include efforts to operationalize (i.e., measure) the arms race phenomenon. The point of this article is not to represent a comprehensive overview of all theoretical and empirical research conducted in the area of arms races. Such a review is well beyond the scope of a single article. Rather the focus is on some of the critical areas where additional theoretical research is needed to move forward our understanding of the arms race phenomenon.

Although there is widespread agreement in the literature on the theoretical definition of an arms race, there is important disagreement on how to measure it. Additionally, the theoretical and empirical research focuses primarily on the consequences of arms competitions (e.g., do they lead to war or peace) at the expense of understanding why they start. And the theories that do exist for the causes and consequences of arms races, grounded primarily in variants of realism and underscored by the security dilemma, are underspecified or logically flawed, rendering them of limited use. The road forward for arms race research must be based on robust theorizing, an increased emphasis on the causes of arms races, and more consistent empirical strategies for testing implications.

Defining an Arms Race

Conceptually, the idea of an arms race is fairly straightforward. Most definitions, at their core, have similar features. An arms race must involve a minimum of two states but could involve more actors in multilateral competitions. There must be mutual targeting, whereby both states acknowledge that they are building up their militaries for the purpose of competing with the other state. And there must be some effort to increase the quantity or the quality of military assets, whether people or equipment, in competition with the other state(s). These basic elements are ubiquitous to definitions found in the literature. The key components of Richardson’s (1960) canonical mathematical model of arms races includes these key conceptual features—he models competition between two actors characterized by the presence of threat, grievance, increases in defense, a fatigue component, and rapid acceleration.

Huntington (1958, p. 41) defines an arms race as “a progressive, competitive peacetime increase in armaments by two states or a coalition of states resulting from conflicting purposes or mutual fear.” Gray’s (1971, p. 40) classic treatment defines an arms race as “two or more parties perceiving themselves to be in an adversary relationship, who are increasing or improving their armaments at a rapid rate and structuring their respective military postures with a general attention to the past, current, and anticipated military and political behavior of the other parties.” Or Kydd’s (2000, p. 229) definition, “a situation in which two or more states involved in a conflictual relationship compete with each other over the strength of their armed forces.”

The similarity in conceptual definitions is a relative strength of the literature. Intuitively, scholars mutually agree on the fundamental character of an arms race. The real point of departure has been how scholars move from those conceptual definitions to operationalization. Although there have been many disagreements, this article focuses on the following: multilateral arms races, where more than one state is arming against a common enemy; what constitutes a rapid military buildup; establishing interdependence, and choosing the appropriate sample population.

The first problem is one of differentiating between bilateral and multilateral arms races. Conceptually, almost all definitions allow for the possibility that an arms race may take place between two or more states. But it is not clear how this should be handled when identifying arms races for empirical analysis. Take, for example, World War I.1 Prior to the outbreak of war, all three members of the Triple Entente (United Kingdom, France, and Russia) are engaged in arms competitions with Germany, and two of its members (France and Russia) are in arms races against Austria-Hungry. Given that Germany and Austria-Hungry are allied through the Triple Alliance (along with Italy), one could reasonably code this as a single arms race between the Triple Entente and the Triple Alliance.

Such a decision would be complicated by at least two factors. First, not all alliance members are arming against each other. Depending on how arms races are measured, the United Kingdom is not arming against Austria-Hungry and none of the Triple Entente members are arming against Italy, a member of the Triple Alliance leading up to the outbreak of war. Second, two intra-alliance partners, Italy and Austria-Hungary, are engaged in an arms race against each other. One could conceivably, and reasonably, code the competitions surrounding these two alliances as a single coalition versus coalition arms race, two arms races (coalition versus coalition and the one intra-alliance arms race between Italy and Austria-Hungry) or as many as six if you disaggregated all of the arms races into bilateral events.

How to treat multiple states arming against a common foe was certainly one of the main points of contention between Wallace (1979), who disaggregated all of the World War I and II arms races into bilateral and independent events, and Weede (1980), or Diehl (1983), who believed multiple states arming against a common enemy, particularly when resulting in a single war, should be treated as a single event.2 Both positions can be justified. Individual states are making decisions domestically to arm with purpose against an enemy. But their decisions are clearly influenced by alliance partners, and in statistical models, which assume independence across observations, it is problematic to treat all of these arms races and resulting wars as independent events. The convention established in the empirical literature has been to disaggregate these arms races, treating them as separate and independent events. But there are clearly statistical and substantive consequences associated with this decision that have not been adequately addressed by scholars.

The second challenge has been determining what constitutes a rapid military buildup. On one hand, this seems simple—look for evidence that states are expanding their military spending or size of their military at a rate that represents an outlier from the norm. On the other hand, any threshold chosen to measure “rapid” increases will be arbitrary, will be difficult to differentiate from alternative explanations (e.g., increased spending for domestic reasons unrelated to the other state), and may fail to capture differences in the quantity or quality of arms possessed by each state. Initial efforts to measure arms races involved developing complex indices that tracked changes in spending (Wallace, 1979, 1982) but were quickly subject to criticism that they were difficult to replicate, lacked transparency and unintentionally captured unilateral buildups rather than just bilateral competitions (Diehl, 1983).

Disagreements also persisted over how much of a spending increase was necessary to code something an arms race. Researchers used thresholds of 10% (Wallace, 1979, 1982), 8% (Diehl, 1983) or an assessment of whether spending was at any point in time higher than that state’s historical average (Horn, 1987). Additionally, it was not clear whether spending just needed to rapidly increase for a specified duration or, based on the Richardson (1960) model, also needed to accelerate over time (Horn, 1987). Although Diehl’s 8% threshold has gained traction in some circles (Gibler, Rider, & Hutchison, 2005; Rider, 2009, 2013; Rider, Findley, & Diehl, 2011), the Horn (1987) measure, relying on comparisons to the historical average and Richardson’s acceleration criteria, has been preferred by those doing research in the steps-to-war tradition (Sample, 1997, 1998, 2002; Senese & Vasquez, 2008).

The third challenge is related to establishing or measuring interdependence. Conceptual definitions generally include, either explicitly or implicitly, some sentiment that the two states acknowledge they are in a competitive relationship and that the decision to arm is motivated or “caused” by and aimed at the other state. Yet few operational definitions include such a requirement. In some ways, this is understandable. Identifying pairs of states rapidly increasing their spending or military personnel is fairly straightforward with available data. Establishing that the effort is directed and interdependent, however, requires time and resource-intensive examination of large numbers of cases using primary and secondary sources focused on government documents and leader statements that may or may not be readily available depending on the time period, transparency of the regime, and strategic incentives of actors. Convention had scholars instead selecting on pairs of states involved in military conflict (Diehl, 1983; Horn, 1987; Sample, 1997, 1998, 2002; Wallace, 1979, 1982), and identifying whether they were both increasing arms prior to the dispute. But this allows for the possibility that rapid arming between the two states was coincidental or aimed at another state (Diehl, 1983; Gibler et al., 2005). Of arms race data sets covering considerable temporal and spatial domains, few other than Gibler et al. (2005) establish independence.

Alternatively, interdependence may also be measured with statistical tools. Using estimation techniques like time-series analysis scholars sought to established evidence that the defense spending of one state “causes” the deference spending of another. In general, the strongest predictor of future defense spending of a state is that state’s past defense spending. State A’s defense spending can be said to “cause” State B’s defense spending, however, if past values of State A’s defense spending provide predictive information about State B’s spending beyond what past values of State B’s spending provide (and vice versa). It was believed that such evidence of interdependent causality in spending would support the basic conceptual assumption of the arms race models: namely, that states were responding to arming decisions of their adversaries (i.e., there is interdependence), a key component of the Richardson action-reaction model.

Despite the intuitive promise of this approach, efforts to use statistical techniques to establish causal interdependence through examination of defense spending have found little supporting evidence (Anderton, 1989). Most research either finds no support for the Richardson action-reaction model (Cusak & Ward, 1981; Smith, Sola, & Spagnolo, 2000) or only mixed support with evidence that arms races potentially work differently than the Richardson model predicts (Majeski, 1985). Majeski and Jones (1981), for example, examine 12 arms races. In the majority, there is no statistical evidence of interdependence of defense spending. And in those where there is interdependence, it is asymmetric; that is to say, only one of the two states is explicitly responding to the spending decisions of the other. One challenge to this line of research, however, suggests methodological misspecification renders some findings suspect, and when properly specified there is evidence for interdependence (Freeman, 1983).

Another challenge to these findings is that defense spending is potentially a poor indicator of arms race activity and, more specifically, arms acquisitions (Stoll, 1982). The argument is that rather than reacting to spending, states typically react to the balance of forces or weapons stockpiles (Ward, 1984). And defense spending does not always capture well changes in stockpiles; indeed, spending can even fall when stockpiles are rising if states are acquiring additional weapons systems rather than replacing (Anderton, 1989, p. 352). Scholars, thus, need to focus on relative balances of arms rather than pure spending. This could take the form of matching arms for arms, as in the classic Anglo German naval arms race, or, and maybe more likely, the matching of an adversaries arms acquisitions with weapons that can defensively counter them (McCubbins, 1983). The limitation of these balance of forces studies is that they have focused almost exclusively on the U.S.-USSR arms race during the Cold War. The difficulty of collecting systematic data on weapons systems and relative balance of forces, especially compared to the ease of defense spending measures, has inhibited scholars from doing research on a large population of states over an extended period of time.

Finally, conceptual definitions of arms races generally include some element of disagreement or enmity. Whether Richardson’s (1960) grievance criteria, Huntington’s (1958) “conflicting purposes” requirement, or Kydd’s (2000) “conflictual relationship” component scholars recognize that arms races occur between states with preexisting hostile relations or, at the very least, an ongoing disagreement over the status quo that could generate the need for arms to compete. But, again, this element of the conceptual definition has either been absent from operationalization or clumsily applied. For example, one method has been to select a sample population of states engaged in militarized disputes with each other (Diehl, 1983; Horn, 1987; Sample, 1997, 1998, 2002; Wallace, 1979, 1982) and then determine if there was rapid arming prior to the dispute. Such a method ensures the states have some conflict of interest, consistent with the conceptual definition, but causes other problems. It excludes states in arms races but not experiencing disputes (Gibler et al., 2005), and creates problems of selection bias for those interested in evaluating the deterrent effects of arms races (Diehl & Crescenzi, 1998). Another solution, similarly problematic, selects on a sample of states with disparate foreign policy preferences (Rider et al., 2011). Probably the most promising approach, suggested by Diehl and Crescenzi (1998) and implemented by Gibler et al. (2005), is to use rivalry as a part of the operational definition. In other words, for there to be an arms race states must be rapidly building arms and engaged in long-term rivalry relations, which ensures the presence of an underlying disagreement, competition, and enmity required by most conceptual definitions.

In summary, scholars share widespread consensus on the conceptual elements that should be present in order to classify a phenomenon an arms race. A similar consensus did not historically exist among those who wished to study arms races empirically, with widespread disagreement over how to measure the various conceptual components. Although debate over measurement is important, the absence of consensus did stifle progress in the study of arms races. After years of debate, some agreement has emerged, with variations of the Diehl (1983) and Horn (1987) measures preferred by a plurality of scholars. Consistent use of the same measure has allowed arms race research to move beyond discussions of measurement and research design to better develop and engage questions of how arms races emerge and their consequences, something discussed in the next sections.

How Arms Races Emerge

The foundational concern of all states is basic security. Survival of the sovereign state necessitates protecting the integrity of the state’s territory. Ceding territory can mean the loss of natural resources, populations, and eventually the state itself. Inadequate security can also leave states at the mercy of others for policy concession. Consequently, most states maintain a standing army as a disincentive to potential adversaries or revision-minded states. The behavior states engage in to dissuade would be revisionist states from altering the status quo, like maintaining a military, is the foundation of general deterrence. In trying to establish a general deterrence posture, states must determine how much in resources to dedicate.

Resources are finite. Governments must choose how to allocate a limited pie among various pots—the classic guns versus butter trade-off. Tax dollars may be spent on domestic priorities such as education, infrastructure, pensions, health care, and other similar needs (i.e., butter). Or revenue may be spent on national defense (i.e., guns). Resources are generally divided between the two depending on the demands of the state at any moment, but an increase in spending on one often necessitates a decrease for the other. As a general rule, we might expect states to economize on security. That is to say, in order to maximize spending on domestic priorities, which are popular rewards to constituencies for patronage-minded politicians, states may wish to spend just enough on defense to satisfy domestic demands for security and deter any potential aggressors.

Buzan and Herring (1998, p. 80) call this the maintenance phase, where states simply attempt to maintain their military status quo. At other times, when the threat environment increases or a leader becomes revision minded, states may feel the need to reallocate spending, with redistribution of resources toward defense. It is when a single state rapidly shift resources toward defense that the potential arises for other states to respond, generating what we might call an arms race.

The action-reaction model represents the conventional logic behind the emergence of arms races (Buzan & Herring, 1998, p. 83). State A initiates a rapid military build aimed at increasing state power. Some other state, State B, observes State A’s efforts to increase military capabilities through this rapid arming, and must decide how to respond. There could be no response, in which case we would just observe a unilateral military buildup by State A. If, however, State B believes State A’s actions are threatening, then State B may reciprocate the behavior by initiating a rapid military buildup of its own.

The next phase may be a reinforcing process, whereby State A observes State B’s actions and respond by continuing the rapid military buildup. At which point State B, observing State A’s sustaining behavior, may feel justified in their initial response (to build arms) and continue their own arming efforts. Through this back and forth process, we would observe the emergence of a competitive arms race.

Why would we expect this action-reaction process that produces arms race outcomes? This logic is rooted in expectations derived from variants of realism, and is undergirded by two key concepts: security dilemma and prisoner’s dilemma. Classic security dilemma, as defined by Herz (1950, p. 157), asserts that states are naturally compelled to increase power (e.g., arms building) for security, but the act of doing so makes other states insecure by definition, triggering them to reciprocate. Anything a state does, even in the name of defense, can and often is viewed as a threat. Consequently, the act of increasing security motivates other states to do likewise, increasing the original state’s insecurity, perpetuating the cycle. The key principle underlying security dilemma is then uncertainty over preferences or intentions—you can never be sure the action you observe is security or revision motivated. It is that uncertainty that compels states to arm in response.

Prisoner’s dilemma has been used as an analogy to explain why individual arming and arms races are a rational outcome in international relations. This is done by exploring a simple 2x2 framework of preferences, outcomes, and payoffs, where two states may choose to arm or not arm without knowledge of what the other state is doing. The logic of the prisoner’s dilemma is widely available elsewhere; thus it is unnecessary to belabor the point. But this article will walk through some of the basic logic in brief, as it relates to arms races (Snyder, 1971). Two states, A and B, know they are jointly better off if they can agree to not engage in competitive arming against each other (i.e., engage in arms control rather than an arms race), allowing them to use those resources elsewhere. But each state realizes they get their best outcome if they choose to arm, while their competitor does not, because it allows them to use the power advantage to exploit the other state. And each state realizes they get their worst outcome if they choose to not arm, while the other does arms, because they will be vulnerable to exploitation by the state increasing its power. Thus, the rational outcome is for states to arm, no matter what their adversary is doing, as they can potentially get their most advantageous outcome (arming while their adversary is not), but at worst avoid their most feared outcome (not arming while their adversary arms). Based on this logic, arms races are a natural product of the incentives and constraints states face in the international system.

We see these ideas of the security dilemma and prisoner’s dilemma articulated in various iterations of realism, particular when it comes to this impetus to arm. States wish to dominate other states, while at the same time avoid being dominated. They do this by seeking power, and balancing against potential challengers. As Morgenthau (2006, p. 192) posits, the “principle means . . . by which a nation endeavors with the power at its disposal to maintain or reestablish the balance of power are armaments.” But the natural dilemma that states face in doing so is that the “armaments race is a constantly increasing burden of military preparations devouring an ever greater proportion of the national budget and making for ever deepening fears, suspicions, and insecurity” (Morgenthau, 2006, p. 192). Similarly, Waltz (1979, ch. 6) argues that states naturally seek armaments because “fear of . . . unwanted consequences stimulates states to behave in ways that tend toward the creation of balance of power.” Finally, Mearsheimer (2001, pp. 156–157) asserts that states “assume direct responsibility for preventing an aggressor from upsetting the balance of power” with the primary goal being to “deter the aggressor” through “internal balancing” (e.g., arms building). For realism and its variants, competitive arming is a natural and rational response to the conditions of the international system.

The Arms Race to War Debate

Consequently, and relying on the tenets of realism, most early arms race research does not ask why states arm because it is framed as self-evident. States arm as a response to this security dilemma. Indeed, realism advances arming in the face of perceived threat as the only logical response. As is a state’s decision to reciprocate rapid arming they see in other states. The actions of another state to arm makes you less secure, you cannot guarantee that they will not use those arms against you, and the consequences for not doing so in the face of a state with aggressive intentions is likely destruction. Thus, scholars saw little value in exploring the “why arm” question. Rather, scholarly emphasis was placed on evaluating the consequences of arms races.

Instead of focusing on the causes, research assumed an ongoing arms race, and focused on the outcome—were arms races followed by war or peace. This was a test of the standard deterrence theory versus spiral model. Deterrence theory, as used in the arms race literature, is uncomplicated. If states want to deter other states, they must meet aggression with aggression (Achen & Snidal, 1989; Snyder & Diesing, 1977). A failure to respond to the actions of others can and will be perceived as weakness. When one state, the initiator state, begins a rapid military buildup, there is a chance they have revisionist or aggressive intentions. If the other state, the defender state, fails to reciprocate that buildup, the initiator state will take this as a sign that the defender lacks the willingness or the capabilities to resist the initiator’s challenge. Consequently, the initiator state will be emboldened to make threats, increase existing demands, or even attack. Alternatively, if the defender chooses to reciprocate the military buildup of the initiator state, the initiator may rethink their revisionist intentions. By reciprocating the arms buildup, the defender not only conveys to the initiator willingness or capabilities to compete, they also raise the cost (by increasing power and preparedness) of action for the initiator. Expectations from deterrence theory are that arms races should lead to peace by demonstrating resolve, capabilities, and raising the cost of conflict.

Alternatively, spiral model suggests a different outcome; arms races will produce war, not peace. Under the spiral model, states reciprocate arming observed by another state because of the basic security dilemma (Jervis, 1976, p. 66). When another state engages in an arms buildup, it both increases its probability of winning conflict (by increasing power and preparedness) and engages in an activity (rapid arming) that appears to have only one purpose (hostility). Even if a state itself may arm for defensive purposes it will not likely interpret the actions of other states as purely defensive. Instead, it will view those actions as nefarious and for aggressive, revisionist intentions.

This is in part because of the difficulty in differentiating weapons as defensive or offensive (Levy, 1984). Leaders have a tendency, however, to interpret their own actions as benign or justified, assume that leaders of other states should interpret them the same way, but are unable to view the actions of their potential adversaries similarly (Jervis, 1976, p. 69). In this situation, the initiator may arm for defensive purposes, and believe that observing states will know their intentions. An observer, the defending state, will see the arms buildup, assume the arms are for hostile purposes, and reciprocate the buildup. The initiator, observing the defenders response, and despite having armed for defensive purposes, may similarly believe the defender can only be arming with hostile intention. The initiator may then respond by increasing her arming efforts. The defender state may observe the initiator’s response, feel justified in her initial reaction, and redouble arming efforts. This reciprocal behavior by both states serves only to reinforce their perceptions of the need for increased security. Through this process, enmity, distrust, and hostility may grow. As tensions increase, and both states continue to arm, reinforcing feelings of enmity and distrust, the likelihood of war increases. War, in this way, occurs not because it was desired but out of a spiral of fear (Jervis, 1976, p. 67).

In summary, research on arms races has relied heavily on security dilemma for the theoretical logic of why arms races start. And spiral model and deterrence theory have been leveraged for understanding the potential consequences. As argued in the next section, however, these conventions are limited in their ability to help explain arms race behavior and advance an understanding of their effect on international relations outcomes.

Weaknesses of Existing Approaches

Theorizing about arms races focuses primarily on the consequences of these competitions (i.e., war or peace), assuming an arms race and emphasizing evaluation of how arms races impacted relations between competitors. In some ways this makes sense, given the general consensus that security dilemma adequately accounts for the arms race phenomenon. And one could reasonably argue that deterrence and spiral models imply the reason for the start of the arms race, if not explicitly theorize about it. States build arms out of either fear of potential threat or out of a desire to revise the status quo. States reciprocate, producing arms races, because they can never been sure of the intention of the originating state. The severity of outcome for not reciprocating, exploitation or state death, are so salient that states must respond. Thus, the arms race outcome is portrayed as a rational and inevitable outcome of an anarchic international system (Waltz, 1979). The more interesting question for scholars, then, was not why arms races occur but an evaluation of divergent expectations on the consequences of arms races. Unfortunately, there are two major problems with this traditional approach, which are explored in the following section.

First, the logic of why states arm is not as self-evident as implied by realism and its variants. Security dilemma and prisoner’s dilemma underlie this traditional arming logic but seem inadequate for fully understanding the process. In some ways, prisoner’s dilemma, as a model of failed cooperation, acts only as an analogy to illustrate how security dilemma is expected to operate. Indeed, Jervis (1976, p. 67) suggests that prisoner’s dilemma makes its way into the debate as an imperfect model of the security dilemma logic. Wagner (2007), however, suggests this is misguided as these two models have little in common. The security dilemma logic is driven by uncertainty over preferences or the intentions of actors. The preferences of actors are well known in the prisoner’s dilemma, and thus outcomes are driven instead by constraints on behavior (Wagner, 2007, p. 30). Without getting into the compatibility of these two types of dilemmas, the article will focus its attention on a critique of security dilemma, which is the core concept underlying this logic of arming, rather than prisoner’s dilemma, which is simply the model used by some to illustrate this logic.

Security dilemma is based on two underlying assumptions (Wagner, 2007, p. 26): Because power is relative, an increase in security for one state decreases security for another and no state can be certain of the intentions of another. Coupled with the assumption of anarchy, realism and its variants conclude that states must respond—either through the use of force or, in this case, arming. Fear alone is enough to make one state arm against another, and because of uncertainty of preferences or intentions states must always assume the worst (Jervis, 1976).

As Wagner (2007, p. 26) so succinctly argues, however, it is not clear why we would conclude this outcome from these premises. To believe this outcome likely we must anticipate arming as probable even if no one benefits from an arms race and absent a history of disagreement. Such an assertion seems odd, particularly given the ubiquity of enmity, rivalry, or ongoing competition in the conceptual definitions of arms races. Indeed, the assumption that an arms race presupposes rivalry (Diehl & Crescenzi, 1998) is widely accepted and receives considerable support in the empirical literature (Gibler et al., 2005; Rider, 2013; Rider et al., 2011). Further, if rivalry is a filter for the context where arms races are likely then the security dilemma logic may be even more problematic. Security dilemma assumes states respond to arming because of uncertainty over the other state’s intentions. But if a rival state is rapidly arming, the decision to reciprocate may not be a result of uncertainty of intentions as security dilemma suggests but instead because history has demonstrated that the state in question is a threat. Indeed, that is precisely why the rivalry exists in the first place. That is not to say that uncertain plays no role in arming decisions. There are theoretical reasons discussed later in the article to suggest it likely does but potentially not in the way that security dilemma posits.

In general, it is likely that security dilemma overemphasizes the extent to which the problem it articulates actually plagues the international system. It assume that all states are Germany prior to World War II, and all leaders are Hitler—hostile, revisionist, and unappeasable. But not all states are potential threats, and thus not all acts of arms building will be viewed as demanding a response. The action-reaction process that produces an arms race—building by State A and reciprocation by State B—require context, generally preexisting enmity or disagreement over the status quo. Building arms is expensive, and triggering an arms race even more so both in resources and the added risk of escalation to war.

Governments generally need strong justification for reallocating resources toward increased defense spending. This can be provided by reference to an existing competitor or enemy. It is unlikely provided by reference to general and nonspecific fear of a state’s actions, where there is not existing or recent history of enmity (Gray, 1974, p. 218). In short, observing some other state arming is likely not sufficient to trigger reciprocation simply because that state might have revisionist or greedy intentions. Indeed, states likely have the ability to signal to each other information about their intentions that may help avoid these types of outcomes (Kydd, 1997). Finally, an adequate theory of arms races must at minimum answer the key inefficiency question—given the cost and risks of arms races, and that states would be better off to resolve disagreements short of these types of competitions, why arm (Kydd, 2000)? Security dilemma does not provide an answer to this question.

Second, empirical research on arms races typically pit spiral model and deterrence theory as alternative explanations for the potential consequences of arms races, and perform comparative theory tests to determine which best matches the empirical record (For examples, see Diehl, 1983; Horn, 1987; Sample, 1997, 1998, 2002; Wallace, 1979, 1982). Glaser (2000, p. 265) suggests that this comparative hypothesis test method is problematic because spiral and deterrence conditions are contextual and not mutually exclusive; in other works, he suggest both or neither may be right depending on circumstances and motivations of the particular states, an idea Kydd (1997) well explores. Additionally, one could argue that the key question that scholars pursue is not necessarily about arms races but why states make the decision to initiate war. Puzzlingly, and with few exceptions, a theory of how war comes is absent from the arms race literature (Morrow, 1989).

Spiral model predicts that war should come from an arms race but the precise mechanism is unclear. Presumably states launch war not out of preference or material gain but instead fear (Jervis, 1976). But fear alone seems an inadequate answer for why states, better off solving problems short of war, are unable to find an acceptable solution short of war (Fearon, 1995). The spiral and Richardson models assume that fear and hostility increase over the course of the arms race, triggering decisions to go to war. But Morrow (1989, p. 501) argues that this is implausible. Decisions to go to war are made based on how much a state values some good, relative to the cost of war, and the likelihood of victory. Fear and hostility will have no effect on how much a state values the prize of competition or prospects for success, and thus it is unlikely that arms races influence decisions to go to war through accumulation of fear.

Morrow (1989) argues that instead the arms race must change a state view of the outcome of a war, through how it affects relative power, on the probability of winning the disputed good. Whether Morrow’s model of war and arms races is accurate is an empirical question yet to be fully explored. But the more salient point, that the arms race literature, and particularly spiral model, lacks a theory of war is well made and shared by other scholars (see also Intriligator, 1975; Intriligator & Brito, 1984; Kydd, 1997, 2000).

Deterrence theory, as depicted in the arms race literature, assumes that competitive buildups should lead to peace. But this prediction is a vast oversimplification, and fails to test many of the expectations of deterrence theory (Diehl & Crescenzi, 1998). Deterrence occurs when one state is adequately able to signal that it has the capabilities and will to stop the other state. It works by raising the cost of winning a disputed good to a level that is unacceptable for at least one of the parties. We would only expect arms races to have a deterrent effect to the extent that the race accomplishes this, and existing theoretical treatments and empirical tests in the arms race literature rarely fully engage this question (Intriligator & Brito, 1984). Much like with spiral theory, we lack a concrete connection between the arms race and war. War is a product of decision making. If we are to causally link arms races to war, we need theory that captures how the arms race affects preferences over and the incentives for war, and bridges directly to the logic of deterrence.

Recent Theoretical Developments on the Causes of Arms Races

Recent theoretical and empirical research on arms races has shifted the focus from trying to understand the consequences of arms races to theorizing about their causes. This is motivated by both the inadequacy of existing theory on the arms race-war question but also by the simple idea that understanding of the cause could very well better inform our understanding of the consequences. To this end, both Kydd (2000) and Rider (2009, 2013) present theories of arms race onset, and Rider (2009, 2013) preliminary empirical evidence. Taking a page from Morrow’s (1989) suggestion that arms race-war research needs a theory of war, these scholars have brought to bear theories of bargaining and war to explaining the causes of arms races.

War is an inefficient outcome in that fighting the war burns resources and, thus, shrinks the size of the spoils split by the adversaries. Given a choice, states would prefer to divide the larger pre-war pool of resources rather than the smaller post-war pool. Which begs the question, why are states unable to take advantage of that pre-war bargain that would make them both better off? This is the classic inefficiency question posed by Fearon (1995). As an answer, Fearon advances that there are obstacles to pre-war bargains states are sometimes unable to overcome. One of those obstacles, uncertainty, is borrowed to answer a similar question applied to arms races. Kydd (2000, p. 229) proposes that because rapid increases in military spending are “inefficient, states have an incentive to avoid them” but that “arms races arise when there is uncertainty.” For Kydd, that uncertainty is over the ability of the adversary to compete economically. If the two states are uncertain about their adversary’s ability to carry a large defense burden needed to sustain an arms competition, an arms race might commence. And in reducing this uncertainty through the competition, the states may find that mutually acceptable negotiated settlement needed to avoid war.

Rider (2009, 2013) builds on the initial insights of Kydd, but in addition to capabilities emphasizes an additional source of uncertainty—resolve. Rider (2009) agrees with Kydd that, given the costs, states will wish to avoid them if possible. But he builds on this idea with additional theoretical expectations. First, the cost of the arms race will make justifying difficult for all but the most salient issues over which states disagree. Given the salience of territorial issues (Vasquez, 1993), Rider (2009) maintains that these are the issues most likely to produce arms competitions.

Second, a rapid increase in military power is not something states can do on short notice. Indeed, the process of approving budget or personnel increases, placing and fulfilling orders for armaments, and realizing the benefits of this increase in power can take years. This means that arms races are only useful for certainty types of foreign policy action, namely, as part of grand strategy (Fearon, 1997, p. 83). This means that we are most likely to see arms races during enduring competitions like rivalries rather than short-term crises. Finally, Rider (2013) advances that arms races are a product not just of uncertainty over capabilities but also resolve. Traditionally, arms races are conceived as competitions over capabilities. But, building on the work of Slantchev (2005), Rider argues that states may initiate military-ups to demonstrate not just capabilities but also that they are willing to pay high costs (i.e., resolve) in order to win the disputed good. Preliminary empirical evidence is provided that demonstrates a correlation between the emergence of arms competitions and periods of heightened uncertainty.

This research, with a renewed focus on the causes of arms races, should be viewed as a starting point rather than an end point. It is promising but certainly more theoretical and empirical research is needed before definitive conclusions may be reached. The next, and final, section discusses the road forward.

The Road Forward

Shifting focus from the consequences to the causes of arms races has been a fruitful enterprise that should continue. This research has successfully integrated theories of bargaining failure (Fearon, 1995), which have revolutionized our understanding of the cause of war, into understanding why arms races occur. Arms races, much like war, are a product of failure of states to reach mutually satisfying resolutions to status quo disagreements. Using these new insights from bargaining theory and war was a natural next progression, especially given suspected links between arms races and war. Although war is viewed as a product of bargaining failure, scholars also recognize the war fighting process as a continuation of bargaining (Filson & Werner, 2002) that must overcome the problems that led to bargaining failure (e.g., war onset) in the first place (Slantchev, 2003). Advances in applying theories of the causes of war to understanding war duration, termination, and settlement have generated great insights (Fearon, 2004; Slantchev, 2004; Werner, 1998). Such should be the case with arms races as well.

If we conceptualize arms races as a product of failed bargaining, and yet still a continuation of the bargaining process, we might expect that the arms race will end when the obstacle to resolution producing the arms race is finally resolved. Indeed, Kydd (2000) suggests this explicitly in his model of arms race onset. He offers that if arms races are a product of uncertainty over an adversaries’ ability to compete, we might expect arms races to end when sufficient information is revealed to determine the distribution of settlement. One implication from Kydd’s model is that if information is revealed through the arms race, it makes fighting a war unnecessary; thus, arms races might be associated with a lower probability of escalation. Such a conclusion is inconsistent with much of the statistical models (see Gibler et al., 2005; Sample, 2002), which conclude that there is a modest but robust positive correlation between arms races and war. But until empirical models explicitly integrate these theoretical models on the causes of arms races, we cannot reasonably draw conclusions. Additionally, the idea that information revelation through the arms race could affect the competition might also be applied to understand the duration of arms races and rivalry, how arms races end, and the types of settlements that emerge.

It is possible that arms races, although triggered by uncertainty, affect important factors other than uncertainty that might also influence decisions for war. Once an arms race commences, states have multiple options for how to proceed, key among them: continue the arms race, acquiesce the issue, or launch a war. But how does the arms race dynamic influence which of these decisions are chosen? Take, for example, the decision to go to war. Morrow (1989) posited that arms races might cause war through changes in distribution of power, which affects states’ expectations of winning a disputed good should war occur. An implication from Kydd’s model is that information revelation from the arms race will make war less likely. But it is also possible that arms buildups could make war more likely not through fear (Jervis, 1976) or even failed deterrence but by shifting of preferences. Over time, the cost accumulation of the arms race could shift preferences such that launching a war is preferred to continuing the race (Slantchev, 2005). War may also come when the defense burden, the cost of the military buildup relative to the country’s total budget or economy, is unsustainable (Sample, 2002). Without more explicit theorizing, which integrates theories of bargaining, arms races, and war, we will continue to struggle with answering these types of questions.


Achen, C. H., & Snidal, D. (1989). Rational deterrence theory and comparative case studies. World Politics, 41(2), 143–169.Find this resource:

Anderton, C. H. (1989). Arms race modeling: Problems and prospects. Journal of Conflict Resolution, 33(2), 346–367.Find this resource:

Buzan, B., & Herring, E. (1998). The arms dynamic in world politics. Boulder, CO: Lynne Rienner.Find this resource:

Cusak, T. R., & Ward, M. D. (1981). Military spending in the United States, Soviet Union, and the People’s Republic of China. Journal of Conflict Resolution, 25(3), 429–469.Find this resource:

Diehl, P. F. (1983). Arms races and escalation—a closer look. Journal of Peace Research, 20(3), 205–212.Find this resource:

Diehl, P. F., & Crescenzi, M. J. C. (1998). Reconfiguring the arms race war debate. Journal of Peace Research, 35(1), 111–118.Find this resource:

Fearon, J. D. (1995). Rationalist explanations for war. International Organization, 49(3), 379–414.Find this resource:

Fearon, J. D. (1997). Signaling foreign policy interests—tying hands versus sinking costs. Journal of Conflict Resolution, 41(1), 68–90.Find this resource:

Fearon, J. D. (2004). Why do some civil wars last so much longer than others? Journal of Peace Research, 41(3), 275–301.Find this resource:

Filson, D., & Werner, S. (2002). A bargaining model of war and peace. American Journal of Political Science, 46, 819–838.Find this resource:

Freeman, J. R. (1983). Granger causality and the time series analysis of political relationships. American Journal of Political Science, 27(2), 327–358.Find this resource:

Gibler, D. M., Rider, T. J., & Hutchison, M. L. (2005). Taking arms against a sea of troubles: Conventional arms races during periods of rivalry. Journal of Peace Research, 42(2), 131–147.Find this resource:

Glaser, G. L. (2000). The causes and consequences of arms races. Annual Review of Political Science, 3, 251–276.Find this resource:

Gray, C. S. (1971). The arms race phenomenon. World Politics, 24(1), 39–79.Find this resource:

Gray, C. S. (1974). The urge to compete: Rationales for arms racing. World Politics, 26(2), 207–233.Find this resource:

Herz, J. H. (1950). Idealist internationalism and the security dilemma. World Politics, 2(2), 157–180.Find this resource:

Horn, M. D. (1987). Arms races and the international system. PhD dissertation, University of Rochester.Find this resource:

Huntington, S. P. (1958). Arms races: Prerequisites and results. Public Policy, 8(1), 41–86.Find this resource:

Intriligator, M. D. (1975). Strategic considerations in the Richardson model of arms races. Journal of Political Economy, 83(2), 339–354.Find this resource:

Intriligator, M. D., & Brito, D. L. (1984). Can arms races lead to the outbreak of war? Journal of Conflict Resolution, 28(1), 63–84.Find this resource:

Jervis, R. (1976). Perceptions and misperceptions in international politics. Princeton, NJ: Princeton University Press.Find this resource:

Kydd, A. (1997). Game theory and the spiral model. World Politics, 49, 371–400.Find this resource:

Kydd, A. (2000). Arms races and arms control: Modeling the Hawk perspective. American Journal of Political Science, 44(2), 222–238.Find this resource:

Levy, J. S. (1984). The offensive/defensive balance of military technology: A theoretical and historical analysis. International Studies Quarterly, 28(2), 219–238.Find this resource:

Majeski, S. J. (1985). Expectations and the arms race. American Journal of Political Science, 29(2), 217–245.Find this resource:

Majeski, S. J., & Jones, D. L. (1981). Arms race modeling: Causality analysis and model specification. Journal of Conflict Resolution, 25(2), 259–288.Find this resource:

McCubbins, M. D. (1983). Policy components of arms competition. American Journal of Political Science, 27(3), 385–406.Find this resource:

Mearsheimer, J. J. (2001). The tragedy of great power politics. New York: W. W. Norton and Company.Find this resource:

Morgenthau, H. J. (2006). Politics among nations: The struggle for power and peace (7th ed., Rev. by K. W. Thompson & W. D. Clinton). Boston: McGraw Hill.Find this resource:

Morrow, J. D. (1989). A twist of truth: A reexamination of the effects of arms races on the occurrence of war. Journal of Conflict Resolution, 33(3), 500–529.Find this resource:

Richardson, L. F. (1960). Statistics of deadly quarrels. Pacific Grove, CA: Boxwood Press.Find this resource:

Rider, T. J. (2009). Understanding arms race onset: Rivalry, threat, and territorial competition. Journal of Politics, 71(2), 693–703.Find this resource:

Rider, T. J. (2013). Uncertainty, salient stakes and the causes of conventional arms races. International Studies Quarterly, 57, 580–591.Find this resource:

Rider, T. J., Findley, M. G., & Diehl, P. F. (2011). Just part of the game? Arms races, rivalry and competition. Journal of Peace Research, 48(1), 85–100.Find this resource:

Sample, S. G. (1997). Arms races and dispute escalation: Resolving the debate. Journal of Peace Research, 34(1), 7–22.Find this resource:

Sample, S. G. (1998). Military buildups, war, and realpolitik: A multivariate model. Journal of Conflict Resolution, 42(2), 156–175.Find this resource:

Sample, S. G. (2002). The outcomes of military buildups: Minor states vs. major powers. Journal of Peace Research, 39(6), 669–691.Find this resource:

Senese, P. D., & Vasquez, J. A. (2008). The steps to war. Princeton, NJ: Princeton University Press.Find this resource:

Slantchev, B. L. (2003). The principle of convergence in wartime negotiations. American Political Science Review, 97(4), 621–632.Find this resource:

Slantchev, B. L. (2004). How initiators end their wars: The duration of warfare and the terms of peace. American Journal of Political Science, 48(4), 813–829.Find this resource:

Slantchev, B. L. (2005). Military coercion in interstate crises. American Political Science Review, 99(4), 533–547.Find this resource:

Smith, R., Sola, M., & Spagnolo, F. (2000). The prisoner’s dilemma and regime-switching in the Greek-Turkish arms race. Journal of Peace Research, 37(6), 737–750.Find this resource:

Snyder, G. H. (1971). Prisoner’s dilemma and chicken models in international politics. International Studies Quarterly, 15(1), 66–103.Find this resource:

Snyder, G. H., & Diesing, P. (1977). Conflict among nations: Decision making and system structure in international crises. Princeton, NJ: Princeton University Press.Find this resource:

Stoll, R. J. (1982). Let the researcher beware: The use of the Richardson equations to estimate the parameters of a dyadic arms acquisition process. American Journal of Political Science, 26(1), 77–89.Find this resource:

Wagner, R. H. (2007). War and the state. Ann Arbor: University of Michigan Press.Find this resource:

Wallace, M. D. (1979). Arms races and escalation: Some new evidence. Journal of Conflict Resolution, 23, 3–16.Find this resource:

Wallace, M. D. (1982). Armaments and escalation: Two competing hypotheses. International Studies Quarterly, 26, 37–56.Find this resource:

Waltz, K. (1979). Theory of international politics. Reading, MA: Addison-Wesley.Find this resource:

Ward, M. D. (1984). Differential paths to parity: A study of the contemporary arms race. American Political Science Review, 78(2), 297–317.Find this resource:

Weede, E. (1980). Arms races and escalation: Some persisting doubts. Journal of Conflict Resolution, 24(2), 285–287.Find this resource:

Werner, S. (1998). Negotiating the terms of settlement: War aims and bargaining leverage. Journal of Conflict Resolution, 42(3), 321–343.Find this resource:

Vasquez, J. A. (1993). The war puzzle. Cambridge, U.K.: Cambridge University Press.Find this resource:


(1.) For simplicity’s sake, the examples are drawn from one particular operational definition of an arms race provided by Gibler, Rider, and Hutchison (2005). Other operational measures demonstrate similar patterns relative to the complicated process of determining bilateral versus multilateral arming.

(2.) They key point made by both Weede and Diehl is that disaggregation leads to potential overestimation of the effects of arms races on war. Rather than a single arms race leading to a single war, disaggregation of these observations results in multiple arms races producing multiple wars.