(Now with updates and minor corrections, for extra goodness.)
When I came to the University of Maryland College Park for grad school in the fall of 1998, I didn’t have to buy either The Strategy of Conflict or Arms and Influence, having inherited the battered copies my parents shared in college. That was a tiny advantage; my real fortune was that both titles were required reading in a class led by the author. Thomas Schelling’s course covered his own major ideas on strategy and social policy: tacit bargaining, tacit coordination, tipping, self-command, and so on. I also had the opportunity to study with Stansfield Turner and with John Steinbruner, who also oversaw my master’s thesis. But for the rest of strategic studies, I was left to my own devices.
It’s hard to believe how long it has been.
Not until about a decade later did I return to reading Schelling with care, and began to appreciate just how central his work remains, despite the specific, vanished circumstances of its origins. His work is not always easy to digest, especially those two landmark books of the 1960s; they’re rather lumpy reads. But read again I did, and came away with new understandings.
It’s with some ambivalence that I finally try to set down some of these thoughts now that he is gone and can no longer correct my interpretations. But never mind. Here goes.
(This is the first part of a two-part essay. See part two here.)
* * *
Nuclear deterrence is famously plagued by problems of credibility, leading to notions like “the rationality of irrationality.” But what does it mean for a deterrent threat not to be credible? Why is this an issue at all?
To a first approximation, credibly promising retaliation with strategic nuclear weapons to deter strategic nuclear attack just isn’t that hard. Vengeance alone is a powerful motivator. (Set aside any question of vulnerability to a first strike; I’m just addressing the willingness to act.) What’s harder is to credibly promise to use strategic nuclear weapons to deter conventional attack, especially a limited conventional attack on a peripheral interest.
And that was the prime security problem that faced the United States for most of the decade after the Korean War. It would be only a modest exaggeration to say that the classic American literature on deterrence (ca. late 1950s and early 1960s) was about nothing but the problem of defending West Berlin. Schelling’s work was no exception, and grasping that point makes his writings easier to understand. NATO forces along the Central Front, it was presumed, would be defeated in a full-scale Soviet invasion, and had no answer at all to a move on West Berlin; in either case, how credible would it be to threaten an American nuclear response if that entailed national suicide?
This particular security situation—more powerful Soviet armies in Europe overall, a totally indefensible position for NATO in Berlin, and what Americans believed to be a rapidly growing Soviet nuclear arsenal—is what gave rise to ideas like “the threat that leaves something to chance,” the title of the eighth chapter in The Strategy of Conflict (1960). There, Schelling observed that a sufficiently credible threat in these circumstances requires a partial relinquishment of control over events:
The key to these threats is that, though one may or may not carry them out if the threatened party fails to comply, the final decision is not altogether under the threatener’s control. The threat is not quite of the form ‘I may or may not, according as I choose,’ but, has an element of, ‘I may or may not, and even I can’t be altogether sure.’
These risks, Schelling observed, could involve “ ‘chance,’ accident, third-party influence, imperfection in the machinery of decision, or just processes that we do not entirely understand.” Viewed from a certain angle, much of the subsequent literature is an elaboration on this sentence, even if presented as critical of the policy of deterrence. The potential for loss of control isn’t a bug; it’s a feature.
The special salience of “the threat that leaves something to chance” to European security can be detected in some passages in The Strategy of Conflict and Arms and Influence (1966), but both books lean to the broadly conceptual over concrete applications. Nowhere in Schelling’s works was the European connection better spelled out, as far as I am aware, than in “Nuclear Strategy in Europe,” an article that appeared in spring 1962 in World Politics and was not anthologized. The gist is that the true value of forward-deployed nuclear weapons was not any contribution to victory in a land war, but their enhancement of the credible danger that a Soviet invasion would lead first to a local nuclear exchange and then to a strategic nuclear exchange. This point even found its way into official NATO thinking within a few years, although that was clearly a justification for deployments long since carried out, and the idea that nuclear dangers needed any enhancements soon went out of style.
Here’s a story that conveys the centrality of Berlin to American thinking in those years. A couple of years ago, I visited Tom at his home in Maryland, where we spoke for awhile. During our conversation, he mentioned his role in organizing and moderating a series of “crisis games” under Pentagon auspices—from 1957 or 1958 until late 1963, if his (and my) memory serves. [Update: It appears that the game series ran for just 25 months, from September 1961 until October 1963. See the end of this post for more details.] These three-day events took place from Friday through Sunday at Camp David; the highest-ranking official involved was Attorney General Robert Kennedy.
These games are mentioned in Fred Kaplan’s The Wizards of Armageddon. Schelling also discussed gaming in a 1961 World Politics article, an internal RAND Corporation essay in 1964, and a chapter in the 1987 book Managing Nuclear Operations, edited by Ashton Carter (whatever became of him?), John Steinbruner, and Charles Zraket. [See the updates below for more.] Here’s a tip for FOIA filers: to the best of my knowledge, the official reports from the games still lurk in the bowels of the Pentagon, but have never been released.
Each of the first four games, Tom told me, involved a Berlin crisis. After that, they moved on to various scenarios in what was then called the Third World; I assume the intervention of the Cuban Missile Crisis drove the change. It is interesting to speculate about how involvement in the games may have influenced the thinking of EXCOMM members.
As an aside: Tom also mentioned that last game took place in the fall of 1963—in late October, I think—and that he sold Bobby Kennedy on adapting the game format to the investigation of social problems, e.g., desegregation, with a promise to pitch the idea to his brother, the President. Did that conversation ever take place? The series did not continue into the Johnson years.
Back to Berlin and all that. Extrapolating a general theory of nuclear-armed deterrence from the particular circumstances that NATO faced ca. 1961 would be a mistake, but it’s not hard to see how to expand upon it. A couple of different basic combinations can be quickly identified. If a nuclear-armed country enjoys conventional superiority over its main, nuclear-armed enemy and is utterly determined to defend an interest, even at the risk of nuclear war, well, that’s boring! The interesting problem arises when one side has the conventional superiority and the other side has the determination to defend an interest. (There are also various in-between conditions, but let’s keep it simple.) Under this “split” condition, each side has a distinct set of problems and a distinct attitude toward stability—meaning, in this instance, the likelihood of avoiding nuclear use in the event of crisis or war.
Schelling nodded to ambivalence in the West toward stability in The Strategy of Conflict, and mentioned it again in his foreword to the 2013 book Strategic Stability: Contending Interpretations, edited by Mike Gerson and Bridge Colby, to explain why it never became central to the discussion:
‘Too much stability’ was recognized by some analysts as possibly immunizing a Soviet attack on Western Europe from a U.S. nuclear response, but after the 1962 Cuban escapade, that issue seemed to disappear.
Whether or not stability was to be favored or avoided in the abstract, the experience of such a hair-raising close call showed both superpowers the true extent of their common interest in survival. It’s not a coincidence that arms control really got going afterward, starting with the Hotline Agreement.
In its outlines, what I’m saying here is not so new; Glenn Snyder parsed out the various potential conditions all the way back in 1965 in a classic book chapter, “The Balance of Power and the Balance of Terror.” (See Paul Seabury, ed., Balance of Power.) Looking around the world today, we can see that the United States and NATO now occupy the “Soviet” role in Europe; it is Russia that must explore doctrinal commitments to the limited use of nuclear weapons in response to what it sees as a conventional threat from NATO. Pakistan is in the “Russian” position as it looks at India, as is North Korea looking at the American-South Korean alliance. (For one take on this idea, especially how this strategy enables less powerful states to harass their opponents, see Jerry Meyerle’s 2014 essay on “coercive escalation.” In Vipin Narang’s book, also from 2014, the idea is termed “asymmetric escalation.”) Evidently, none of the actors in these dyads have lately become scared enough to start finding cooperative ways to turn down the heat; they are all still “pre-Cuban.”
China is interesting for not having adopted a version of this strategy. Instead, its leaders have committed to enhancing both their conventional forces and their public commitment to their “core interests,” as they define them. Perhaps these choices can be linked in part to the fundamentally maritime nature of Chinese security concerns; there is no real land-warfare threat today to China or to any state that China is committed to defend, since China has no such commitments.* Contrast this with how Russia, Pakistan, and North Korea understand their own circumstances. Regardless of Beijing’s exact reasons, I think Schelling would understand.
(* Yeah, I know about the treaty with North Korea. It’s a dead letter.)
Oh, and if you are looking for a contemporary application of this discussion of the relationship of conventional power and nuclear danger, here it is. The country with the most powerful military in the world—today, kids, that’s the United States—is the one with the most to lose from instability and unpredictability.
Update 1: I now see that Schelling contributed an account of the crisis games to a 2016 book, Harrigan and Kirschenbaum, eds., Zones of Control: Perspectives on Wargaming. Based on what’s available in Google Books, it sounds like I may have misunderstood or misremembered what he told me about timeframe; he makes it sound as if the Berlin games began in 1961, inspired by the rules he developed for an Iran game played at MIT in 1960.
Update 2: Thanks to Bill Burr of the George Washington University’s National Security Archive, I can share two documents I’d never seen or heard about before. (Both are from the Digital National Security Archive.) The first is a partially redacted September 1961 memorandum addressed to “General Taylor,” presumably Maxwell Taylor, then President Kennedy’s personal military adviser. It describes a Berlin game held at Camp David on September 8-11, evidently the first of the series. The second is a transcript of a November 1988 talk at Harvard by Schelling and Alan Ferguson, who had worked with Schelling to develop the first Berlin game over the summer of 1961 at RAND in Santa Monica. In the Q&A, Schelling dates his abovementioned exchange with Bobby Kennedy to the end of a game held on Halloween, 1963. The transcript includes an anecdote that Tom also shared with me, but I couldn’t remember well enough to reproduce. Here it is:
Schelling: Someone even said during the Cuban missile crisis: “This crisis sure demonstrates how realistic Tom Schelling’s games were.” Somebody else replied: “No, Schelling’s games only demonstrate how unrealistic this Cuban stuff is.” [Laughter]
Update 3: Still more! Thanks to Ellie Bartels of RAND.
I wish there were a message or lesson written at the end of this piece that would target Trump’s plans and intentions, and the associated threats. Richard Honour
Isn’t there?
I feel it in there, but I am not the expert, so I need a more direct statement that i can repeat from a relaible source. Richard
Here it is more directly. The USA has an extremely powerful military. It’s unparalleled in its excellence. Despite some local problems (the defense of Seoul or the Baltic states, let’s say), it can be counted on to prevail in any serious, head-to-head contest with another major military establishment. The President-elect seems to see virtue in unpredictability and perhaps even some instability, as I used the term in this essay, but those are actually the tools that Russia and especially North Korea feel they must use to keep us at arm’s length. Nuclear danger works to their advantage, not to ours; without it, we could not be challenged.
That is clear, thank you, and that is the core of the issue as I see it, i.e., foolish unpredictability coupled with poorly-informed, misinformed or dis-informed instability, most especially as manifested by his selected ‘generals,’ who have no driving force or motivation other than war at any cost; that is exactly why civilian rule is the required mandate in a sound democracy. Richard Honour
Schellings’s work is important, beyond a doubt, but I do see some limitations. Credibility of a threat may stem from rationality, but it can also stem from plausible irrationality. How was it rational to risk up to a 50% chance of nuclear war (in Kennedy’s view) to insist that the Soviet Union remove nuclear missiles from Cuba? How was it rational for both sides to refuse to back down, despite such high risk of nuclear conflagration, until both had agreed to a compromise?
The “threat that leaves something to chance” is only a rationalization, not a rational explanation, for nuclear threats that leave so much to chance. Because of a too-high U.S. willingness to risk nuclear war, it was far too credible that a Soviet invasion of Western Europe would lead to nuclear war. Because of a too-high Soviet willingness to risk nuclear war, the threat of nuclear retaliation was insufficiently likely to deter, if the Soviet Union were ever determined to invade.
Hence, it is far from obvious that NATO’s nuclear strategy was rational, given the risks (and abysmal consequences) of deterrence.failure. Similar concerns exist for the risky nuclear strategies of Russia, Pakistan, and North Korea today. Schelling never provided a cost-benefit analysis to show that the expected deterrence benefits from nuclear threats exceed the expected costs (including risk of nuclear war).
Maybe the difference lies between what is rational in the short term and what is rational in the long term. I discuss this in a brief postscript to tomorrow’s installment.
In your second installment, I see a discussion of the role of chance, but not a distinction between short run and long run. The risk of nuclear war is inherent in every move to the brink; it is not a short-run vs. long-run distinction.
With one proviso, I agree with your conclusion that if brinkmanship is resorted to repeatedly, nuclear war is all but inevitable. To derive this conclusion, the size of the risk also matters. If each crisis has 50% risk of nuclear war, then ten crises produce a nearly inevitable 99.9% chance of nuclear war. If each crisis has only 1% risk, then ten crises produce a not-so-inevitable 10% chance of nuclear war.
A cost-benefit analysis would need to show that the expected benefits of nuclear risk-taking exceeds the expected costs. Because the expected downside of nuclear war is extremely bad, the expected upside must be really good to justify such risks.
Yes. That’s what I’m getting at. Take a chance once or twice and maybe you’ll get away with it. Do it too many times and you won’t.
Beyond the short-term/long-term distinction, what seems rational at one point in time may not seem rational in another.
John Kennedy’s presidency was less than 20 years after the end of World War II, not a decade after the Korean War. Nuclear weapons were newer then, and approaches to using them were still developing.
That’s not a justification of risks taken, but an explanation. People just thought differently about the risks, and that thinking was still rational within its context.
Schelling’s Nobel Prize acceptance lecture ended with something like this:
“People who don’t like us very much are likely to become nuclear powers.”
The implied hope was that these “people” would learn about & adopt some of the same convention(s) that the US/USSR had created.
And so, achieve the same outcome –no use of nuclear weapons in war since 1945.
Shelling was a bit too optimistic on human learning, or perhaps he was hoping to coax more people to learn the lessons. In 1962 the world escaped a hair-raising nuclear crisis, but less than 10 years later Nixon thought it would be useful to play the part of a nuclear “madman.” Pakistan, North Korea, probably Putin, and maybe Trump have all decided that playing nuclear “madman” is a low-risk strategy. It is high risk.
The madman strategy works something like this: Bring your opponent to a cliff and threaten to punch him in the nose. Surely, if your opponent is sensible, he will walk away from the cliff and let you win without a fight. (Never mind that your opponent might punch you first, or that it might be sensible for you to walk away from the cliff and let your opponent win.)
Come to think of it, did Schelling advocate the madman strategy, advocate against it, or merely describe it? It would be nice to know what lesson Schelling wanted us to learn.
Jonah, Schelling never advocated or even described what you call the “madman” strategy. Not sure what source you are drawing this from. Thanks.
At a minimum, Schelling did describe the madman strategy. If memory serves me, Schelling described versions of the strategy (e.g., Kremlin politburo under Khrushchev) in Strategy and Conflict. In later editions, a preface specifically references the madman strategy and indicates that his description of a strategy is scientific only (as in people might do this), and should not be construed as endorsement.
Talking or behaving like a “madman” is a species of threat that leaves something to chance, at least in the eyes of opponents who cannot determine whether a man is truly mad. Leaving nuclear war to chance is not clearly wise or sane. Metaphorically, the cold-war NATO setup was “mad” to risk nuclear war, even if no particular individual within NATO was mad.
For an interesting diversion, see “Did Thomas C. Schelling Invent the Madman Theory?” http://historynewsnetwork.org/article/17183
Michael, that’s not exactly correct. There is a section of Arms and Influence that discusses the potential advantages of being seen as impetuous or willing to harm oneself. I don’t have my copy at hand, but it opens with a discussion of a notional anarchist with a bomb who can intimidate the police through his sheer willingness to die, and then goes on to discuss the difficulties of actually trying to implement such a policy at a national level.
I’m the newbee to your site, but I will take exception to a few points here. The idea behind “plausible irrationality” may be misplaced. I sat in a submarine off that point on land in southeast Cuba, east of Bay of Pigs, near Guantanamo, in 1962, waiting for the nose of any Russian ship to round that point of land, at which point we would have fired a few torpedos, so that “50% chance of nuclear war,” in my and our view back then at that location, was likely more at about 99%. The ship noses did appear, and then turned around at the last moment, to change the course of history. We do not need that again, which is exactly why Trump and his dating game with Putin is of grave concern. The game is the same, only the fools are different. Richard
Thanks for sharing that story, Richard.
Which ship appeared, Richard?
How do you see defense of Seoul as a “local” problem? It means that unlike other deterrents, the NK nuclear force (and continued development–even of ICBMs) is shielded behind the threat of utter destruction of Seoul by conventional artillery. We couldn’t and can’t engage in a strike to disable nuclear forces without inevitably provoking that as a result. Or as I put this over on Twitter, the NK regime has what in chess would be called a fork, or a defense that doesn’t rely upon the piece being defended.
Defending Seoul is a local problem in the geographic sense. It’s difficult for basically the reasons you say. The outcome of an all-out war would not be in doubt, but it wouldn’t be good for Seoul.
? Is this a spelling correction?
I don’t follow…
Oh, I think I know what you mean. I originally posted just the word “Seoul” in reply to comment – fumbled my phone, that’s all. I later edited it into a proper response. Try reloading the page.
Thank you. Do you think US will regard Seoul as expendable in the event of a real ICBM threat from NK?
No.
“The potential for loss of control isn’t a bug; it’s a feature.”
Bingo.
MK
It’s a feature when one can be sure that those “in control” think of it as a bug. Otherwise, pretty much a bug.
I think that this is a good example of Brinkmanship:
https://www.youtube.com/watch?v=3xhBl0R0qBY
Joshua writes: “Nuclear danger works to their advantage, not to ours; without it, we could not be challenged.”
This is an excellent point, I think.
Do I understand you to mean by “their advantage” deterring the US from any direct assault on their territories?
Sure. Or deterring retaliation for actions that they themselves might take.