Rational deterrence theory developed alongside the advent of nuclear weapons. The intersection of the two events yielded the concept of ‘nuclear deterrence’, an idea that has become popular. It asserts that a real or perceived threat of catastrophic nuclear attack from one state to another allows that first state to deter the second from certain causes of action, namely military confrontations. An extended version of this idea holds that the world is now somehow more peaceful as a result of interlocking looming threats of nuclear attack.
Nuclear deterrence is largely a deductive theory, and has been criticised for, among other things, being ahistorical and lacking in empirical validity. Deterrence proponents have responded by saying that empirical deterrence scholarship such as comparative case studies have failed to persuasively challenge the assumptions behind the theory (Achen & Snidal 1989). This paper will argue that the empirical work that has been done on deterrence theory, criticising its assumptions and conclusions, is largely correct. In particular it will be argued that the conceptual and methodological limitations of the empirical criticisms of nuclear deterrence theory are not too great to be overcome—that is, that it is perfectly valid to build social scientific theories and concepts from empirical observation.
I - Description of The Idea of Nuclear Deterrence
Nuclear deterrence will be described before discussing the validity of the various empirical criticisms that have been levelled against it. The reason why it must be described is because its use has entered common parlance, and as such has suffered from great variation and diffusion in meaning. A fairly rigorous approximation of nuclear deterrence will be set out in order to provide some terms of reference, so that the deterrence criticisms coming later will fit into some form of coherent dialogue.
Nuclear deterrence proceeds from rational choice analyses. When nuclear deterrence is invoked, the following assumptions are usually implicit:
Rational actor assumption. Actors have exogenously given preferences and choice options, and they seek to optimise preferences in light of other actors’ preferences and options.
Principle explanatory assumption. Variation in outcomes is to be explained by differences in actors’ opportunities. (Appeals to exogenous changes in preferences, or to norms, roles, or culture, are temporarily or analytically suspended.)
Principle substantive assumption. The state acts as if it were a unitary rational actor. (Changes in personnel, in decision-making patterns, or in bureaucratic politics are not the explanatory focus (Achen & Snidal 1989, p. 150).
It follows that a relationship of deterrence is established when a threat of retaliatory nuclear attack from one state alters the preference optimisation calculus of another that is attempting to alter the balance of power in a system. The relationship of deterrence is meant to be established in the following manner. Let us say that there are two states in a system and a state of affairs between them such that there is a stable power ‘status quo’ in the system before any kind of relationship of deterrence comes into being. One state is a defender of the status quo, and the other is an initiator of change of power relations. Pursuant to the assumptions above, both states exist in competition and seek to better their position in the system. Both states perform a calculation of the best decisions to make at any particular time, and let us say that, given the initiator is savvy to the preferences and options of the defender (this is important, preferences being close to ‘desires’ and options being close to real courses of action related to capabilities), it has the preference to attack the defender and alter the status quo. Whether or not it acts on that preference is based on a calculation that involves weighing up the costs and benefits of making that act. One of the costs involved in attacking the defending state is the future threat of it making a retaliatory attack, as ‘punishment’. If the cost of a retaliatory attack from a defending state is too great, an initiator will be deterred from challenging the status quo. Hence a relationship of deterrence is obtained.
A - Credibility
The problem-solving capacity of deterrence theory concerns the process by which a defending state actively and purposively influences the cost/benefit calculation of an initiator’s decision to attack and change the status quo. The focus which most have placed on the ability of deterrence theory to solve security problems comes from the fact that a significant portion of the initiator’s cost/benefit calculation revolves around an indeterminacy, something that the initiator cannot work out with certainty. Relaxing the assumption above that the initiator is fully aware of preferences and options of the defender, the analyst encounters the problem that initiator cannot know for sure whether the defender will have the ability and commitment to fight back after an attack (Achen & Snidal 1989, p. 151).
The concept that is used to measure the probability by which an initiator estimates the cost of a defender’s ability and commitment to have a war or some other kind of conflict over the alteration of the status quo is called deterrence credibility. Working backwards in the initiator’s decision-making process, suppose that a defender’s threat to retaliate is credible. This means that the initiator believes that the defender has the option to retaliate (not just the desire, but the capacity exists: the defender has the military means to retaliate after an attack and is politically free to do so), and that the defender would find retaliation in its interest if the prize is threatened. Then, if the threatened retaliatory punishment from the defender exceeds the gains of the initiator from attacking, the initiator will believe the attack will make it worse off than by practicing restraint, and will be deterrable, or deterred from attacking.
B - Reassurance
By dint of history there is a corresponding nuclear deterrence concept which is related to nuclear deterrence credibility but is somewhat different. The concept is called nuclear deterrence stability, and it comes from extending the idea of credibility by replacing the system of one initiator and one defender with one which features two actors that are simultaneous initiators and defenders. Because both actors exist in the same competitive environment as the credibility model, where both actors are seeking to satisfy their preferences as much as possible, and both powers have nuclear weapons, there exists the risk that one actor may pre-empt the other’s nuclear deterrence threats, and totally wipe it out, yielding a zero-sum win for the pre-empting actor. Stability is, in this context, a very narrow indicator. It is meant to measure the ‘second strike’ (or ‘strike second’) nuclear capacity of either actor in this system. If both actors have a second strike capacity, they still have the capacity to form the option of a credible deterrence threat after they have been attacked with a nuclear arsenal. The trouble with this—and this introduces some impurities into the rational choice description of nuclear deterrence—is that there is always the incentive for both actors strive to build nuclear weapons systems that deprive their competitor of a second strike capacity, eradicating the effect of nuclear deterrence entirely. Here the indeterminacy problem of deterrence credibility becomes transmuted into nuclear reassurance. Instead of there being some uncertainty to be explained as to the desire and capacity of a state to attack for the purpose of deterrence, the problem-solving function of the rational choice schema is set towards working out whether either actor believes that its competitor won’t attack. The active and purposive component of reassurance is that, instead of issuing threats, each side seeks to reassure the other that it won’t commit to plans that will see the other side wiped out, destroying the effectiveness of its deterrence. Reassurance is more or less the same as deterrence credibility because a state protects its own deterrence credibility by promising not to harm its competitor’s credibility.
Deterrence and reassurance here turn on the credibility of a retaliatory attack by a defender which alters the cost/benefit calculation of the initiator. It is important to note that credibility turns correspondingly on the perception either generated passively or created actively (and there is some controversy here to be sure) inside the initiator about (a) the nature of the defender, and (b) the defender’s interests and intentions. It should be noted that the empirical literature has much to say about this. For the sake of the current discussion however, it is important to point out how nuclear deterrence theory gets around the perception issues inherent in its operation. The way it gets around the issue of how a defender changes the behaviour of an initiator is that it tightly controls how either actor thinks. First, as per the first assumption above, it makes state preferences (the things they desire) exogenous to the rational choice games through which it puts them. This means that the things that states want are assumed beforehand, and held constant so that games can be constructed in a sensible manner. Second, it makes sure that the behaviour that states exhibit is related to the choice of those assumed preferences, and not anything else. Forces such as mistakes, the role of communication and language, or complex cognitive processes are not the explanatory focus of the theory.
The reason why it is important to point out the way in which nuclear deterrence theory gets around the problems of state communication and interaction is that many of the empirical studies on nuclear deterrence have raised questions about the validity of these rational choice assumptions. The big problem with the empirical criticisms of nuclear deterrence theory, however, is that they proceed from a very different conceptual and methodological paradigm than that of rational choice analyses. The fundamental difference between nuclear deterrence theory and the empirical work that has been done on it is the order in which theory and evidence is used. This has a real effect on the empirical approach’s ability to verify the descriptive and prescriptive claims that nuclear deterrence theory makes. Whereas nuclear deterrence theory is deductive, which abstracts from reality for the purpose of building a theory which is meant to be conceptually coherent and simple in its explanatory import, empirical attempts to engage with deterrence are inductive: observations come first, and typologies and generalisations come second. Methodologically speaking, nuclear deterrence advocates are probably less concerned with individual case explanations than analysts who are attempting to empirically verify the theory. This is significant because the only way that an empirical study of the validity of nuclear deterrence can proceed is by conducting case studies. For instance Achen & Snidal’s complaint that empirical case studies don’t really address the concerns behind the method of deductive theories:
[f]or theoretical purposes, the difficulty with explaining individual cases is that there are so many details in every case that no single theory can reproduce them all, and some evidence can be found for too wide an array of variables and propositions (Achen & Snidal 1989, p. 157).
While this is probably an overstatement of the limits of conducting case studies, there are indeed difficulties with making generalisations with an empirical inquiry when looking at such a deductive and conceptually ‘pure’ theory such as nuclear deterrence. As Jervis (1989, p. 195) notes, when conducting empirical studies,
[i]diosyncrasies are many and powerful; similar outcomes can be reached by many different paths and apparently similar initial conditions do not always yield the same results. Partly for this reason and partly because of the limitations of existing research, what we have no is more a list than a tight theory.
So there is a real divide between the conceptual and methodological fundamentals of deductive and inductive theories. The typologies between empirical case studies of nuclear deterrence theory can be incredibly fluid and varied, and may not rigorously address every facet of the single deductive theory to which they might be aimed. But the disadvantages outlined above should come with a qualification. What we know from the evidence on nuclear deterrence
is not without structure and coherence. Many of the generalisations produced by the case studies are linked; the inferred processes can generate additional propositions; and the arguments apply to a wide range of problems in addition to deterrence and deterrence failures (Jervis 1989, p. 196).
It is important to discuss all this because treating the empirical criticisms of nuclear deterrence theory as if they apply directly to nuclear deterrence’s own analytical focus only tells half the story. As will be shown below, the evidence on nuclear deterrence takes issue with the way states interact and communicate to be sure, but it doesn’t do so in a way that starts from the same first principles as rational deterrence theory. The evidence proceeds in a way that holds a lot of other things constant, and not just the assumptions behind nuclear deterrence. For instance, the evidence might investigate the possibility of preference endogeneity with respect to deterrence, but in a way that sees states as the product of bureaucratic and not unitary decision-making processes. Despite the piecemeal and highly qualified progress of empirical work of nuclear deterrence, it should be seen as valid. The validity comes from its ability to make coherent and persuasive generalisations that have strong and convincing prescriptive and descriptive power.
II - Empirical Findings
Jervis (1979) provides a concrete summary of what the case study literature finds about the ‘reality’ of nuclear deterrence. The findings fall under several headings (or ‘themes’) and these are: (a) the reality of state risk-taking behaviour; (b) the role of rewards and concessions in deterrence; (c) mistakes, and the things that can go wrong; (d) the limitations of state information-processing and cognition; and (e) the nature of state interests and commitments—that is, the costs that states impose on themselves for the purpose of bringing about nuclear deterrence.
A - Risk-taking
The first of these themes, risk-taking, is relatively straightforward, and it relates to the ‘brinkmanship’ aspect of nuclear deterrence. Nuclear deterrence theory points towards the use of ‘credible threats of force’—a willingness and capacity to deploy threats of force in order to deter— that are meant to alter the cost/benefit analysis of an initiator’s thinking when that initiator wants to change the status quo power relations. What the empirical evidence says about this process is that states and state elites of a ‘defending’ state are often far more prudent and risk-averse than nuclear deterrence theory would have one believe. States are very often wary of foreclosing the policy options that they have available to them, reducing possible courses of action down to one single choice. The cognitive and information-processing aspect of this preference-selection phenomenon will obviously discussed later, but it is significant to note that the rational calculation aspect of nuclear deterrence suggests that defending state elites should have no trouble issuing grave threats and managing an incredibly risky relationship with an initiator, when in reality this does not seem to be the case at all. What the evidence suggests is that defending and initiating state elites bargain with each other: defending state elites prefer to ‘trade an increment in their chance of prevailing for an increment in the chances of maintaining peace’ (Jervis 1979, p. 303).
Furthermore, there appear to be different risk-management behaviours for different kinds of defenders and initiators. If the above were taken to be a general rule, it would seem that states care less about long-run strategic considerations than one would expect. This is not the case. When it comes to great power (say USA versus USSR) deterrence interactions, long-run considerations feature much more prominently in state decision-making. This means that either side is much more likely to retreat and give concessions instead of issuing new threats. This evidence contradicts nuclear deterrence theory in two ways: first, not all defenders and initiators are the same, and in certain cases, defenders will more than not opt to retreat, trusting initiators to not exploit their concessions.
B - Rewards
The second theme looks at the role of rewards, instead of threats and disincentives, in a relationship of nuclear deterrence. The evidence on this category refers to the way the defender’s communication with the initiator is meant to change that initiator’s cost/benefit calculation when it wants to attack. What nuclear deterrence theory says about the deterrence relationship is that an initiator’s decision calculus is geared in such a way that only threats of violence from a status quo defender will influence its operation, and cause it to back down. What the evidence says on this matter is that threats of disincentives do not necessarily, or even in most cases, alter an initiator’s decision calculus in a manner that establishes a relationship of deterrence. It may be that a state has no other option—or thinks it has no other option—but to be bellicose, and threats of nuclear violence might only steel its resolve. In instances where things are no so dire, threats of nuclear violence risk ‘humiliating’ an initiator state; on this point the evidence suggests that frequently it is hard for a state to admit total defeat. Threats of nuclear force may in these circumstances cause initiator feelings of resentment, and may make deterrence far more difficult in the future if not couples with some kind of reward or concession—that is if total deterrence is possible at all.
C - Mistakes and Things That Can Go Wrong
This third theme is a very substantive one, and deals directly with the ‘perception issue’ inherent in nuclear deterrence’s rational choice paradigm. Mistakes, and things that can go wrong when attempting to purposively create a relationship of deterrence fall under another four headings. Deterrence may fail where it is supposed to succeed because the defender: (i) may fail to understand the other’s values; (ii) may misunderstand the way the initiator sees the world; (iii) may have incorrect beliefs about the initiator’s strength and options; and (iv) may face extreme difficulty in correctly determining the initiator’s basic intentions.
Each of these four mistakes in judgment on the part of the defender speaks to the second assumption behind rational deterrence theory. This assumption causes states behaviour to be explained through their ‘opportunities’ only—that is, the opportunity that they have to act on their exogenously given preferences—and not anything else. Not only does this radically simplify the kinds of behaviours states can have, but it also says something about the way states are meant to detect and interpret the behaviours of their counterparts (or ‘competitors’). If the behaviour of states is meant to be tightly linked to their preferences, per a rational choice paradigm, a defending state should be able to safely estimate the preferences of an initiator by looking at its behaviour. However all four of these different types of mistakes and errors in judgment found in the empirical evidence suggest the opposite. They suggest that in practice, states have a very difficult time divining the desires and capabilities of their counterparts.
The explanation in the empirical literature for these difficulties usually falls within a different analytical paradigm. For instance, Jervis (1982) stresses the role of the concept of perceptions in explaining the difficulties states seem to have in correctly working out information about other states. Perceptions represent a significant increase in complexity over exogenous preferences and behaviour in terms of opportunity, because it adds an extra layer of cognitive grey matter over state interaction. Perceptions depend on the attitudes and values of the state that is doing the perceiving, and so when interacting with other states, the perceiving state is heavily conditioned by its own conscious and unconscious psychological elements. This says something about the way states perform their supposed rational calculation under rational choice analyses—this will be discussed next.
D - Information processing and Cognition
The fourth theme is not directly related to the three assumptions outlined above, but nonetheless relates to them closely. Nuclear deterrence theory asserts that states should be able to perform a careful and precise rational appraisal of the costs and benefits of choices when considering bringing about, or reacting to relationships of deterrence. Typically, there should be very few limits in making mistakes or errors of judgment when calculating the kinds of choices involved in executing or reacting to a relationship of deterrence. Note that this is different from the heading above. The theme above describes the errors states makes about their counterparts, this theme covers the expectations that states have about the probabilities of certain things occurring in general.
The Jervis outlines three errors in rationality that states exhibit in practice (Jervis 1979, pp. 310-311). The first of these is that states usually focus on the payoffs involved in the decision-making process, rather than what should be their true focus of calculation: changes in event probabilities. The second is very straight forward, and that is that there is evidence to indicate that even if states have a firm grasp of the probabilities involved in attempting to engineer a relationship of deterrence, they don’t calculate as precisely as they should. The third impairment of rationality is that states, even when they have good estimates of event probabilities, fail to make proper explicit value trade-offs.
Each of these three errors in cognition are again related to the first assumption behind rational deterrence theory, that of preference exogeneity. The theory assumes that states rank and order preferences in a systematic and unbiased manner. What the evidences suggests about state rationality is that states don’t seem to draw up and scrutinise policy preferences in such a clear and ordered manner. Some preferences are more valued than others for cognitive, yet—if one was talking in a strict sense—irrational reasons, whereas others may not suffer from being under- or over-valued, but may not be compared and analysed properly.
E - State Interests and Commitments
The fifth and final theme that Jervis describes as emerging from the body of empirical literature about nuclear deterrence is that of the nature of state interests and commitments when attempting to establish a relationship of deterrence. Very briefly, this theme looks at the way states formulate their perceptions of the costs and benefits that are involved in establishing a relationship of nuclear deterrence, be they defender or initiator. The empirical literature suggests that there exist two factors that influence state rationality when they deal with nuclear deterrence relationships, and these are either (i) a state’s interest in something; or (ii) a state’s commitment with respect to something. What is important about these two factors is that they are measures of costs and benefits that states self-impose upon themselves. This is of course against the grain when it comes to the precepts of rational deterrence theory, which, again, returning to the matter of preference exogeneity, says that states consider all options equally and without bias.
III - Conclusion
This paper has made a very brief effort at outlining the operation of rational deterrence theory, one of the many bases that exist for substantiating nuclear deterrence theories, and in turn relating rational deterrence theory to the criticisms that have been made of it in empirical literature. The paper’s attention has chiefly been concerned with Robert Jervis’s summary of the empirical work that has been done on rational deterrence theory.
Achen, CH & Snidal, D 1989, ‘Rational Deterrence Theory and Comparative Case Studies’, World Politics, vol 41, no. 2, pp. 143-169.
Jervis, R 1979, ‘Deterrence Theory Revisited’, World Politics, vol 31, no. 2, pp. 289-324.
Jervis, R 1982, ‘Deterrence and Perception’, International Security, vol 7, no. 3, pp. 3-30.
Jervis, R 1989, ‘Rational Deterrence: Theory and Evidence’, World Politics, vol 41, no. 2, pp. 183-207.
MccGwire, M 1986, ‘Deterrence: the problem–not the solution’, International Affairs, pp. 55-70.
MccGwire, M 1994, ‘Is There a Future for Nuclear Weapons?’, International Affairs, vol 70, no. 2, pp. 211-228.
MccGwire, M 2006, ‘Appendix 2: Nuclear deterrence’, International Affairs, vol 82, no. 4, pp. 771-784.
Ritchie, N 2009, ‘Deterrence dogma? Challenging the relevance of British nuclear weapons’, International Affairs, vol 85, no. 1, pp. 81-98.