Nuclear credibility in theory and in practice

Executive summary

Over its more than 80-year history as a nuclear power, the United States has sought to convince adversaries that it can wage, will wage, and will win a nuclear war. It has sought to convey the credibility of these threats through different combinations of the types, sizes, and locations of its nuclear weapons. No nation has yet exploded a nuclear weapon on the territory, people, or allies of the United States. That nonuse, however, has remained constant even as U.S. nuclear strategy has varied, making it impossible to conclude that strategy, rather than other factors, did the causal work. Nuclear attack, however, is not the only empirically observable event that can provide information about U.S. nuclear credibility: the United States has made and has been the recipient of nuclear threats, both implicit and explicit. These are moments of acute nuclear danger; each incident has the potential to turn into nuclear disaster. Responsible policy, therefore, will seek to minimize how often such crises arise.

This study examines whether and how different U.S. nuclear strategies are associated with few nuclear crises, or with many. It uses an original dataset of 36 U.S.-involved nuclear crises between 1945 and 2026—instances in which the United States or a nuclear-armed adversary made an explicit or implicit nuclear threat—and tests whether their frequency is associated with the features of U.S. nuclear strategy and the broader military and political context.

The single most robust finding is that nuclear crises were less frequent when the United States and its nuclear adversary approached rough parity in nuclear capability and markedly more frequent when one side held a substantial advantage. The policy levers the U.S. national security establishment has spent eight decades pulling in the name of credibility—doctrinal refinements, arsenal size, the forward deployment of nuclear forces—show no consistent, independent association with how often nuclear crises occurred.

These findings speak directly to live policy debates. Some officials and analysts warn that the United States faces a potential “credibility deficit”: that U.S. nuclear strategy today is inadequate to hold at risk enough Russian and Chinese targets to deter both simultaneously, and so it must prepare to expand its arsenal’s size, variety, and posture to close the gap. The historical record instead suggests that a Chinese arsenal organized around a secure second-strike capability is unlikely to increase the frequency of nuclear crises and might reduce it. Historically, it is an assured second-strike capability—not nuclear advantage—that has been stabilizing. Seeking to widen the U.S. lead over China, or to match a combined Russia-China target set, would therefore be to implement an enormously costly solution to a problem that the United States does not have.

The findings also have implications for arms control. If parity is associated with stability—if the condition most associated with fewer crises is mutual assured retaliation and not U.S. superiority—then the relevant policy question is not whether, but rather at what level, to seek parity. The historical patterns offer no evidence that larger arsenals on both sides are more stabilizing than smaller ones; what appears to matter is the balance, not the count. This should make small stockpiles far more appealing than large ones: negotiated reductions that preserve mutual second-strike capability while shrinking the overall nuclear inventory save taxpayer dollars without sacrificing security.