The best studies confirm our impact---err on the side of a consensus of empirical research---our evidence assumes every skeptic.

Consistency with influential relevant theories lends credence to the expectation that US security commitments actually can shape the strategic environment as deep engagement presupposes. But it is far from conclusive. Not all analysts endorse the theories we discussed in chapter 5. These theories make strong assumptions that states generally act rationally and focus primarily on security. Allowing misperceptions, emotions, domestic politics, desire for status, or concern for honor into the picture might alter the verdict on the strategy’s net expected effects. And to model the strategy’s expected effects we had to simplify things by selecting two mechanisms— assurance and deterrence— and examining their effects independently, thus missing potentially powerful positive interactions between them.

This chapter moves beyond theory to examine patterns of evidence. If the theoretical arguments about the security effects of deep engagement are right, what sort of evidence should we see? Two major bodies of evidence are most important: general empirical findings concerning the strategy’s key mechanisms and regionally focused research.

General Patterns of Evidence Three key questions about US security provision have received the most extensive analysis. First, do alliances such as those sustained by the United States actually deter war and increase security? Second, does such security provision actually hinder nuclear proliferation? And third, does limiting proliferation actually increase security?

Deterrence Effectiveness The determinants of deterrence success and failure have attracted scores of quantitative and case study tests. Much of the case study work yields a cautionary finding: that deterrence is much harder in practice than in theory, because standard models assume away the complexities of human psychology and domestic politics that tend to make some states hard to deter and might cause deterrence policies to backfire. 1 Many quantitative findings, mean- while, are mutually contradictory or are clearly not relevant to extended deterrence. But some relevant results receive broad support:

Alliances generally do have a deterrent effect. In a study spanning nearly two centuries, Johnson and Leeds found “support for the hypothesis that defensive alliances deter the initiation of disputes.” They conclude that “defensive alliances lower the probability of international conflict and are thus a good policy option for states seeking to maintain peace in the world.” Sechser and Fuhrmann similarly find that formal defense pacts with nuclear states have significant deterrence benefits. 2 3

The overall balance of military forces (including nuclear) between states does not appear to influence deterrence; the local balance of military forces in the specific theater in which deterrence is actually practiced, however, is key. 4

Forward- deployed troops enhance the deterrent effect of alliances with overseas allies. 5

Strong mutual interests and ties enhance deterrence. 6

Case studies strongly ratify the theoretical expectation that it is easier to defend a given status quo than to challenge it forcefully: compellence (sometimes termed “coercion” or “coercive diplomacy”) is extremely hard.

The most important finding to emerge from this voluminous research is that alliances— especially with nuclear- armed allies like the United States— actually work in deterring conflict. This is all the more striking in view of the fact that what scholars call “selection bias” probably works against it. The United States is more inclined to offer— and protégés to seek— alliance rela- tionships in settings where the probability of military conflicts is higher than average. The fact that alliances work to deter conflict in precisely the situations where deterrence is likely to be especially hard is noteworthy.

More specifically, these findings buttress the key theoretical implication that if the United States is interested in deterring military challenges to the status quo in key regions, relying only on latent military capabilities in the US homeland is likely to be far less effective than having an overseas military posture. Similarly, they lend support to the general proposition that a forward deterrence posture is strongly appealing to a status quo power, because defending a given status quo is far cheaper than overturning it, and, once a favorable status quo is successfully overturned, restoring the status quo ante can be expected to be fearsomely costly. Recognizing the significance of these findings clearly casts doubt on the “wait on the sidelines and decide whether to intervene later” approach that is so strongly favored by retrenchment proponents.

The Causes of Nuclear Proliferation Matthew Kroenig highlights a number of reasons why US policymakers seek to limit the spread of nuclear weapons: “Fear that nuclear proliferation might deter [US leaders] from using military intervention to pursue their interests, reduce the effectiveness of their coercive diplomacy, trigger regional instability, undermine their alliance structures, dissipate their strategic attention, and set off further nuclear proliferation within their sphere of influence.” These are not the only reasons for concern about nuclear proliferation; also notable are the enhanced prospects of nuclear accidents and the greater risk of leakage of nuclear material to terrorists. 9 8

Do deep engagement’s security ties serve to contain the spread of nuclear weapons? The literature on the causes of proliferation is massive and faces challenges as great as any in international relations. With few cases to study, severe challenges in gathering evidence about inevitably secretive nuclear programs, and a large number of factors in play on both the demand and the supply sides, findings are decidedly mixed. Alliance relationships are just one piece of this complex puzzle, one that is hard to isolate from all the other factors in play. And empirical studies face the same selection bias problem just discussed: Nuclear powers are more likely to offer security guarantees to states confronting a serious threat and thus facing above- average incentives to acquire nuclear weapons. Indeed, alliance guarantees might be offered to states actively considering the nuclear option precisely in order to try to forestall that decision. Like a strong drug given only to very sick patients, alliances thus may have a powerful effect even if they sometimes fail to work as hoped. 10

Bearing these challenges in mind, the most relevant findings that emerge from this literature are:

The most recent statistical analysis of the precise question at issue concludes that “security guarantees significantly reduce proliferation proclivity among their recipients.” In addition, states with such guarantees are less likely to export sensitive nuclear material and technology to other nonnuclear states. 12 11

Case study research underscores that the complexity of motivations for acquiring nuclear weapons cannot be reduced to security: domestic politics, economic interests, and prestige all matter. 13

Multiple independently conceived and executed recent case studies nonetheless reveal that security alliances help explain numerous allied decisions not to proliferate even when security is not always the main driver of leaders’ interest in a nuclear program. As Nuno Monteiro and Alexandre Debs stress, “States whose security goals are subsumed by their sponsors’ own aims have never acquired the bomb. … This finding highlights the role of U.S. security commitments in stymieing nuclear proliferation: U.S. protégés will only seek the bomb if they doubt U.S. protection of their core security goals.” 15 14

Multiple independently conceived and executed recent case research projects further unpack the conditions that decrease the likelihood of allied proliferation, centering on the credibility of the alliance commitment. In addition, in some cases of prevention failure, the alliances allow the patron to influence the ally’s nuclear program subsequently, decreasing further proliferation risks. 17

Security alliances lower the likelihood of proliferation cascades. To be sure, many predicted cascades did not occur. But security provision, mainly by the United States, is a key reason why. The most comprehensive statistical analysis finds that states are more likely to proliferate in response to neighbors when three conditions are met: (1) there is an intense security rivalry between the two countries; (2) the prospective proliferating state does not have a security guarantee from a nuclear- armed patron; and (3) the potential proliferator has the industrial and technical capacity to launch an indigenous nuclear program. 18 19 16

In sum, as Monteiro and Debs note, “Despite grave concerns that more states would seek a nuclear deterrent to counter U.S. power preponderance,” in fact “the spread of nuclear weapons decelerated with the end of the Cold War in 1989.” Their research, as well as that of scores of scholars using multiple methods and representing many contrasting theoretical perspectives, shows that US security guarantees and the counter- proliferation policy deep engagement allows are a big part of the reason why. 20

The Costs of Nuclear Proliferation General empirical findings thus lend support to the proposition that security alliances impede nuclear proliferation. But is this a net contributor to global security? Most practitioners and policy analysts would probably not even bring this up as a question and would automatically answer yes if it were raised. Yet a small but very prominent group of theorists within the academy reach a different answer: some of the same realist precepts that generate the theoretical prediction that retrenchment would increase demand for nuclear weapons also suggest that proliferation might increase security such that the net effect of retrenchment could be neutral. Most notably, “nuclear optimists” like Kenneth Waltz contend that deterrence essentially solves the security problem for all nuclear- armed states, largely eliminating the direct use of force among them. It follows that US retrenchment might generate an initial decrease in security followed by an increase as insecure states acquire nuclear capabilities, ultimately leaving no net effect on international security. 21

This perspective is countered by “nuclear pessimists” such as Scott Sagan. Reaching outside realism to organization theory and other bodies of social science research, they see major security downsides from new nuclear states. Copious research produced by Sagan and others casts doubt on the expectation that governments can be relied upon to create secure and controlled nuclear forces. The more nuclear states there are, the higher the probability that the organizational, psychological, and civil- military pathologies Sagan identifies will turn an episode like one of the numerous “near misses” he uncovers into actual nuclear use. As Campbell Craig warns, “One day a warning system will fail, or an official will panic, or a terrorist attack will be misconstrued, and the missiles will fly.” 22 23

Looking beyond these kinds of factors, it is notable that powerful reasons to question the assessment of proliferation optimists also emerge even if one assumes, as they do, that states are rational and seek only to maximize their security. First, nuclear deterrence can only work by raising the risk of nuclear war. For deterrence to be credible, there has to be a nonzero chance of nuclear use. If nuclear use is impossible, deterrence cannot be credible. It follows that every nuclear deterrence relationship depends on some probability of 24 nuclear use. The more such relationships there are, the greater the risk of nuclear war. Proliferation therefore increases the chances of nuclear war even in a perfectly rationalist world. Proliferation optimists cannot logically deny that nuclear spread increases the risk of nuclear war. Their argument must be that the security gains of nuclear spread outweigh this enhanced risk.

Estimating that risk is not simply a matter of pondering the conditions under which leaders will choose to unleash nuclear war. Rather, as Schelling established, the question is whether states will run the risk of using nuclear weapons. Nuclear crisis bargaining is about a “competition in risk taking.” Kroenig counts some twenty cases in which states—including prominently the United States—ran real risks of nuclear war in order to prevail in crises. As Kroenig notes, “By asking whether states can be deterred or not … proliferation optimists are asking the wrong question. The right question to ask is: what risk of nuclear war is a specific state willing to run against a particular opponent in a given crisis?” The more nuclear- armed states there are, the more the opportunities for such risk- taking and the greater the probability of nuclear use. 27 26 25