Amid China’s nuclear buildup and strengthening Chinese-Russian relations, an old idea is back in vogue in Washington: a larger nuclear arsenal will enhance U.S. security. But would nuclear superiority, as its adherents claim, help the United States prevail in conflicts and extract greater concessions during crises with nuclear-armed adversaries? Or, as new research suggests, might it instead be a disadvantage?
Nuclear superiority—or having more nuclear weapons than one’s opponent(s)—initially found favor in the 1950s and 1960s, when the U.S. nuclear arsenal dwarfed that of the Soviet Union. Eventually the idea faded from prominence, to be replaced by an acceptance of strategic stability enabled by U.S.-Soviet nuclear arms control. After the end of the Cold War, the United States retained relative parity with Russia—although it continued to maintain a much larger arsenal than any other nuclear-armed adversary, including China.
Today, however, China is rapidly expanding its arsenal. Between January 2022 and January 2023, nongovernmental estimates suggest that China’s arsenal increased in size by 17 percent, as Beijing added sixty new nuclear weapons. This trend is likely to continue alongside substantial efforts by China to modernize its delivery systems. Among other projects, China has been building hundreds of new silos, likely for nuclear-armed intercontinental ballistic missiles.
This development appears to represent a challenge for the United States. It must now confront what some have called the two-peer problem, in which the United States must simultaneously deter two nuclear-capable adversaries—Russia and China—and even, if it comes to it, fight nuclear wars against both of them.
Some scholars and experts have called for the United States to address this challenge by expanding the U.S. nuclear arsenal in a bid to secure nuclear superiority. A recent report drafted by a group of respected U.S. nuclear experts (many with extensive government and military experience) and published by Lawrence Livermore National Laboratory concludes that “the United States should get the numbers right” and strengthen its nuclear forces. Franklin Miller, a renowned nuclear expert with a thirty-one-year civil service career, recently advocated for the United States to increase the number of nuclear weapons in its arsenal to prepare for, among other things, threats from China. Georgetown University professor Matthew Kroenig has similarly argued the United States needs to reevaluate the size of its nuclear arsenal to ensure it can “maintain a favorable balance of power over China at each rung of the escalation ladder.” This October, the Strategic Posture Commission—selected by Congress in 2022 to complete a threat evaluation and provide recommendations for U.S. nuclear force posture—released a report embracing a significant U.S. nuclear build-up.
But are they right that the answer to the two-peer problem is a matter of numbers of nuclear weapons?
To answer this question, it is necessary to revisit scholarship about the nuclear balance of power and to look at the empirical record, where nuclear superiority has often failed to award states its promised advantages.
Does Nuclear Superiority Matter?
Some nuclear strategists have argued that nuclear superiority makes states “more likely to prevail in nuclear crises.” The logic runs as follows: Imagine a crisis between two nuclear-armed states. The less powerful state will be more likely to give in to the more powerful state’s wishes, or so the argument goes, because it knows it would fare worse if the crisis were to escalate all the way to nuclear war. The weaker state thus has greater incentive to deescalate a crisis than does the stronger state. Keir Lieber and Daryl Press explain that this thinking was critical to U.S. and Soviet strategy throughout the Cold War, writing that “both superpowers were well aware of the benefits of nuclear primacy, and neither was willing to risk falling behind.”
Some scholars have argued the empirical record shows that this strategy works. For example, Kroenig examines a dataset of crises between nuclear states and finds that states with more nuclear weapons than their opponents have achieved more favorable crisis outcomes. Mark Bell and Julia Macdonald analyze three historical crises between nuclear-armed states and come to a similar conclusion.
In contrast, other scholars, as well as various politicians, have disputed the idea that nuclear wars can be won. Former U.S. president Ronald Reagan first spoke the oft-echoed line that “a nuclear war cannot be won and must never be fought.” In 1982, McGeorge Bundy, George F. Kennan, Robert S. McNamara, and Gerard Smith famously wrote: “Any use of nuclear weapons . . . carries with it a high and inescapable risk of escalation into the general nuclear war which would bring ruin to all and victory to none.”
By this way of thinking, even small nuclear arsenals—so long as they can survive a first strike—should be enough for deterrence to kick in. After all, what does it matter if an adversary can do more damage in a nuclear war that would be an all-out, unthinkably devastating cataclysm? The theorized advantages of nuclear superiority should be lost once both sides possess arsenals large enough to inflict unacceptable damage on each other.
In support of this view, some scholars have argued there is no empirical support for the purported benefits of nuclear superiority. For example, Todd Sechser and Matthew Fuhrmann find that “nuclear weapons are uniquely poor instruments of compellence.” Having collected data on compellent threats made by nuclear states, they conclude that nuclear superiority has not consistently enabled states to compel their adversaries to make political concessions. Studying nuclear and nonnuclear crises from 1900 to 1980, Paul Huth and Bruce Russett determine that “a quest for strategic nuclear superiority is unlikely to be the most effective means for providing security to America’s friends and allies in a crisis, or to America itself.”
Scholarship on nuclear superiority, therefore, has so far has found mixed evidence for these two contrasting arguments.
A New Measurement of Nuclear Superiority
A first step to resolving this disagreement is to note that proponents of nuclear superiority typically argue that the benefits of nuclear superiority come from what it allows states to accomplish strategically—that is, impose more damage on their adversaries while limiting the amount of damage their adversaries can impose on them. But this ability does not come from having one more weapon than the opponent; instead, it is due to a combination of many factors, including the number and yield of its nuclear weapons as well as its delivery capabilities, posture, targeting, and other elements that are critical to a nuclear arsenal’s functionality.
Unfortunately, scholars have tended to measure superiority as a binary question of whether one state has a greater number of nuclear weapons than another or not. This flat approach, called “numerical superiority,” misses many of the crucial components that make nuclear weapons strategically valuable.1
To address these challenges, Abby Fanlo and I have developed a new statistical approach that distinguishes between symmetric and asymmetric nuclear crises. Symmetric crises feature states with similarly powerful nuclear capabilities; asymmetric crises involve one state that possesses a nuclear arsenal vastly larger and more sophisticated than that of its opponent.
In a symmetric crisis, one state might have a handful more nuclear weapons than other, but this information, on its own, is not enough to truly determine which side has the superior arsenal. In an asymmetric crisis, though—when one side has, say, 1,000 times as many nuclear weapons as the other—then there can be no doubt that the state with the larger arsenal is more capable.
This approach raises an obvious question, however: What is the threshold between a symmetric and asymmetric crisis? Our answer is to try every possible threshold to identify exactly when any benefits or drawbacks of superiority kick in.2
Revisiting Nuclear Crises
To implement this technique, we use crises captured by the International Crisis Behavior (ICB) dataset, which has been the basis for previous scholarly quantitative work on nuclear superiority.3 For example, when the Soviet Union invaded Afghanistan in 1979, it had roughly the same number of nuclear weapons as the United States. During the Cuban Missile Crisis in 1962, its nuclear arsenal was about eight times smaller.4
For each nuclear crisis, ICB data determines which state “won,” or achieved the goals it held going into the crisis.5 In the 1956 Suez crisis, for example, the United States was able to stave off the threatened Soviet intervention in Egypt. But neither India nor Pakistan won outright in the 1990 Kashmir crisis, which cooled after then U.S. deputy national security adviser Robert Gates visited South Asia to encourage restraint.
For each possible definition of nuclear asymmetry, we used ICB data to test whether asymmetrically superior states perform better in crises than asymmetrically inferior states (leaving out data from the crises classified as symmetric by each definition being tested).6
Figure 1 shows the results of our analysis. The X axis in the figure indicates possible definitions of nuclear asymmetry (for example, a value of seven indicates that nuclear asymmetry kicks in when one side has seven times as many nuclear weapons as the other). At each threshold, the Y axis shows the probability that an asymmetrically superior state wins, minus the probability that an asymmetrically inferior state wins. If that value is positive, nuclear superiority is correlated with crisis victory; if it is negative, nuclear superiority is linked to failure.
We find that, if we define nuclear superiority as previous scholars did—by saying that having one more weapon than the other side is enough to constitute a meaningful advantage—it looks like nuclear superiority matters. The effect of nuclear superiority on the likelihood of winning a crisis is positive and significant under this definition.
But if we only look at crises that can reasonably be considered asymmetric—crises where, for example, one state has at least twice as many nuclear weapons as the other side—then nuclear superiority is no longer advantageous. As the gap becomes larger, we find that nuclear superiority becomes correlated with failure during crises. In crises where one side had at least eighteen times as many nuclear weapons as the other—the current difference between the U.S. and Chinese nuclear arsenals—the less-powerful state deterred its adversary 82 percent of the time. If one state has a nuclear arsenal at least fifty times larger than its opponent’s, that state always fails.7
What this figure shows, then, is that the advantage that some scholars thought came from nuclear superiority may not be real. When states face far weaker adversaries, nuclear superiority either does not help or makes things worse.8 We argue that this is because weaker adversaries can show more resolve. Indeed, they often have to, in order to guarantee their survival.
The theorized advantage of nuclear superiority only appears when states are facing adversaries with similar capabilities—circumstances in which it is difficult to know whether superiority is really the primary factor in determining crisis outcomes. After all, when one state has an arsenal 1.1 or 1.2 times larger than the other side, can we really conclude that constitutes meaningful nuclear superiority?
Lessons From Nuclear History
These historical crises can provide lessons for understanding and managing the global nuclear order today, as states like China and North Korea expand their nuclear arsenals.
The first lesson is that small differences between nuclear arsenals are not enough to meaningfully indicate nuclear superiority. When it comes to the two-peer problem, this should be reassuring. A strengthening Russia-China alliance could pose many complications for U.S. grand strategy, but even if Russia and China were acting fully in combination with each other, their nuclear balance of power vis-à-vis the United States would still be symmetric. A combined Russia-China arsenal would only be 1.2 times larger than the U.S. nuclear arsenal. Russia has approximately 12 percent more nuclear weapons than the United States, but the U.S. nuclear arsenal is currently almost eighteen times larger than the Chinese stockpile. China could continue to build up its nuclear arsenal, but it would take a long time to reach parity with the United States.
Nevertheless, should China’s arsenal ever expand that far, the challenges that the United States faces in the Indo-Pacific today would be made worse. A China able to match U.S. capabilities could, for example, gain more latitude to operate aggressively in Taiwan or the South and East China Seas. This would be a legitimate concern—although our research suggests it cannot be redressed by simply adding more nuclear weapons to the U.S. arsenal in a bid to maintain superiority. Instead, deterring Chinese aggression will require the United States to show its commitment to the region through improving interoperability within its alliances and investing in conventional deterrence capabilities.
The second lesson—that nuclear superiority can prove disadvantageous—may be more of a mixed blessing. States struggle to win crises against far weaker adversaries. This finding helps explain the difficulties the United States has faced in attempting to influence North Korean behavior. This finding should be reassuring to countries like France and the United Kingdom, whose nuclear arsenals may be able to deter Russia, despite its far-more-powerful nuclear forces. However, this idea may be disquieting to the United States, which often approaches weaker adversaries as if it had a clear deterrent advantage over them. In reality, however, U.S. nuclear superiority may actually prove to be a disadvantage, and a U.S. nuclear buildup could actually backfire.
U.S. leaders have been and presumably will remain committed to possessing a nuclear arsenal that can counterbalance threats from a wide range of adversaries. However, expanding that arsenal in order to achieve clear nuclear superiority is unlikely to be helpful in crises against smaller nuclear adversaries—like North Korea—and it is unnecessary, at least for now, for counterbalancing threats from China and Russia.
Notes
1 More gradated measurements—for example, the ratio of one state’s nuclear capabilities to another’s—are also not suitable for systematic analysis because there have been only a small number of crises between nuclear states, and conventional statistical methods can’t draw reliable conclusions from small amounts of data using such a fine-grained measure.
2 Importantly, our strategy also takes into account other features of crises, like their location, history, and level of violence, as well as various aspects of the states involved, like their regime type, population size, and military capabilities. We also factor in whether states have a nuclear second-strike capability—that all-important ability to retaliate even after being attacked—by identifying whether the state in question has submarine-launched ballistic missiles, mobile missiles, or nuclear-armed aircraft on continuous airborne alert.
3 The ICB data concludes in 2010; earlier studies used ICB data that concluded in 2001. Some crises involve multiple nuclear powers; in these cases, our main analyses treat each adversarial relationship in the crisis as distinct. However, the research paper includes robustness tests, described in note 3, with different sets of crises and outcome codings, including an iteration that groups allied countries together. Additionally, we have not included nonnuclear players in our main analyses, as theories of nuclear superiority and previous quantitative work on nuclear superiority primarily focus on interactions between nuclear-armed powers. For more information on our methodology, please see our research article, which contains appendices with various treatments of nonnuclear states involved in crises with nuclear states.
4Data on the approximate sizes of nuclear arsenals is taken from research by Hans Kristensen and Robert Norris.
5 While the ICB dataset determines whether a crisis ends in victory for each side, there are numerous historical disagreements about individual crises. As a result, our research uses six different approaches to coding crisis outcomes. Our first and second methodologies respectively include and exclude France in the Berlin Wall Crisis, as France did not have any nuclear weapons at the time but was nuclear-capable. In our third approach, we collapse alliances, rather than treating members of alliances as engaging in separate dyadic crises with an adversary. Our fourth methodology drops all cases that did not have a visible, significant nuclear threat component. In our fifth analysis, we recode the outcomes of the Korean War and the Berlin Wall Crisis in line with our reading of the historical record as opposed to ICB’s interpretation. In the sixth approach, we recode the Korean War and the Ussuri River Crisis in order to replicate Kroenig’s approach. We consistently find that high degrees of nuclear superiority do not provide states with an advantage in achieving their crisis goals.
6 For each possible definition of nuclear asymmetry, we use logit models with cluster-robust standard errors to show whether asymmetrically superior states perform better in crises than asymmetrically inferior states, leaving out data from the crises classified as symmetric by any given definition.
7 If we include the USSR-France dyad in the Berlin Wall crisis—in which France did not have any usable nuclear weapons, although it was capable of building them—this is an arguable exception to the rule. The Soviet Union succeeded in dividing East and West Berlin, so the ICB data codes the event as a Soviet victory. However, this crisis might more properly be considered a stalemate, since the Soviet Union did not follow through on its demand that the United States and its allies withdraw from West Berlin.
8 The goal of our scholarship here is to improve the reliability of the statistical analysis that previous work has been able to provide. To this end, we have rerun all of our tests using a wide variety of different specifications. Across these iterations, we show that nuclear superiority does not dependably help states win crises. We implemented a non-parametric approach to improve inference, given the small number of cases in the data. We reestimated our models with amended determinations for the outcomes of crises to reflect various scholarly disagreements; these are described more fully in note 3. Additionally, we tested new models to reflect the joint capabilities of alliances. Further robustness tests we conducted involved recalculating our results to test whether the effect of asymmetric inferiority, rather than superiority, is related to crisis victory. This is important because victory is not zero-sum in our dataset, so it can be the case that both sides win or both sides lose a crisis. We also used a binary measure that defines nuclear superiority as the condition in which a state has at least three times more weapons than its opponent as well as a secure second-strike capability. This measurement reflects arguments by nuclear strategists about what capabilities would be necessary for a nuclear warfighting advantage, including significant damage limitation and credible ability to retaliate to nuclear first use. Finally, we show that measures previously used by scholars, including numerical superiority and the nuclear ratio, produce inconsistent statistical results.