North Korea: Deterrence Roulette or Preventive War?

North Korea is a hostile power dashing to acquire the capability to launch a devastating nuclear attack upon the U.S.. This alarming state of affairs has so far been treated with astonishing equanimity by the foreign policy establishment on the grounds that the U.S. can easily and effectively eliminate the risk of such an attack by pursuing a policy of deterrence. Susan Rice aptly summarizes this near-universal establishment consensus in a New York Times op-ed, writing that “history shows that we can…tolerate nuclear weapons in North Korea — the same way we tolerated … Soviet nuclear weapons during the Cold War”.

How to estimate the risk of deterrence roulette

The risk of a catastrophic deterrence failure that leads to a nuclear attack arises from the very logic of deterrence itself. If two hostile countries are in a deterrence relationship, then each is committed to retaliating if it believes that the other is attacking or is about to attack. While this commitment does indeed limit the possibility that one country deliberately decides to launch an unprovoked nuclear attack on the other (by the standard logic of mutually assured destruction), it simultaneously creates the possibility of nuclear war by mistake. If an event such as political/military crisis, a military misunderstanding, or an early warning system failure leads to one country mistakenly believing that it is or is about to be under attack, that country may in turn mistakenly launch what it thinks is a retaliatory strike (as the logic of deterrence demands). If the probability of nuclear war by mistake is high, then deterrence cannot eliminate the risk of a nuclear attack.  

How can we estimate the risk of a nuclear war by mistake? We need to base our estimate of what might happen in the future on what could have happened during the Cold War. The simple, obvious, and (as it turns out) wrong way to proceed is to then take what did happen to us during the Cold War as a good representation of what could have happened. Going down this path, we find that there were, in fact, no U.S./Soviet nuclear wars (“we tolerated Soviet nuclear weapons during the Cold War”). We then infer that the risk of a deterrence failure is low and so happily conclude that, as Scott Sagan put it in Foreign Affairs, “the same approach that prevented nuclear catastrophe during the Cold War can deter Pyongyang.”

The flaw in this line of reasoning is thinking that what did happen to us captures the true range of possible outcomes. When we look back to what happened to us during the Cold War, the only things that could have happened to us are things that are consistent with us ending up where we are now (even though we could have ended up in a very different situation). It follows that our particular path through history may provide a very biased view of what could have happened in general (in statistics this phenomenon is referred to as observer selection bias). Consequently, what did happen to us may prove to be a very poor guide to what might happen in the future.

To make this pointless abstract, suppose that you have played some large number of rounds of Russian roulette with a gun that has an unknown chance of killing you in each round. Based upon your history, you wish to estimate the risk of playing one more round. If you approach this problem the way the foreign policy establishment approaches North Korea, you will: i) count up the number of times that you have killed yourself in previous rounds; ii) find that you did not kill yourself even once; and iii) happily conclude that Russian roulette is perfectly safe.     

But that would be wrong. Since you are alive to do the counting, it must be the case that your path through previous rounds is consistent with your being alive now. The only path that meets this requirement is one in which you did not kill yourself. Therefore, it is an inherent feature of your current situation that you did not kill yourself in the past. So, looking back from your current situation, you will necessarily find that you did not kill yourself—no matter how high the risk of killing yourself is in a given round. Since you observe the same history of non-death for any true probability of killing yourself, the fact that you are alive now tells you absolutely nothing about the risk of playing another round.

So, if estimating the risk of a catastrophic deterrence failure is like estimating the risk of death in a game of Russian roulette, then estimating the risk of a deterrence failure by counting up the number of U.S./Soviet nuclear exchanges is the wrong approach to take. Is it?

An Office of Technology Assessment report on the consequences of a U.S./Soviet nuclear war found that a full U.S./Soviet nuclear exchange would have destroyed every major city in the U.S., killed in excess of 100 million people, and completely demolished the economy. So, if such an exchange had occurred, many of the people now involved in the “what to do about North Korea” debate would be dead (or would never have existed), the U.S. would not exist in its present form, and, whatever people in this alternative post-apocalyptic U.S. of 2019 were worried about, it would not be “just how effectively can the U.S. deter North Korea?”. In short, there is no path from 1960 to anything close to our present situation that includes even a single U.S./Soviet nuclear exchange.

Consequently, starting from our present situation, there are—by definition—no U.S./Soviet nuclear exchanges in our history. It is, therefore, no surprise that we don’t find one when we look. It follows that the lack of U.S./Soviet nuclear exchanges during the Cold War tells us nothing at all about the risk of catastrophic deterrence failure. We cannot tell (from that information) if we made it through the Cold War because the risk of a deterrence failure was low or if we made it through because the risk was high and we were lucky.

To figure out how to correctly estimate the risk of a deterrence failure, we need to find some way of dealing with observer selection bias.

To do so, we can think of the Cold War as consisting of a series of tranquil rounds and crisis rounds. A tranquil round is one with no risk of a nuclear war; a crisis round consists of an event that could lead to a nuclear war and ends in either a nuclear war near miss or a nuclear war full stop. So, the risk of a nuclear war by mistake increases as the probability of a crisis increases. 

Our Cold War experience consists of a specific path through history consisting of tranquil rounds and crisis rounds. The set of all possible paths through the Cold War include those with a crisis that ends in nuclear war. Of course, given that we are in our present situation, we know that we did not take such a path.

But—crucially—our path could include nuclear war near misses because a near miss would not have destroyed the U.S. as we know it today. Since a nuclear war near miss indicates a crisis that could have led to nuclear war if circumstances had been ever so slightly different, we can estimate the probability of crisis by counting up (and analyzing) these near-miss events. 

So, we can correctly gauge the risk of a nuclear war by mistake by examining our experience with deterrence during the Cold War. We have to look for the right things. And the right things to look for are nuclear war near misses rather than nuclear exchanges.

The risks of deterrence roulette with North Korea

There were a terrifyingly high number of nuclear war near misses during the Cold War. It follows that deterrence roulette is extremely risky and that (as uncomfortable as this is to hear) good luck played a crucial role in getting us through the Cold War without a nuclear exchange.   

Near misses arose from political/military crises such as the Cuban Missile Crisis, military misunderstandings such as the Able Archer exercise of 1983, and early warning system failures. Examining nuclear war near miss events suggests that a U.S./North Korean deterrence relationship will be far more prone to catastrophic failure than was the U.S./Soviet deterrence relationship.

Two episodes, in particular, illustrate just how fragile any U.S./North Korean deterrence relationship will be.

On 9 November 1979, a software simulation of a full Soviet nuclear attack was somehow (it is still not clear how) transferred onto the operational warning system displays at NORAD, SAC, and the National Military Command Center. All three commands reacted as if the Soviets had genuinely launched an attack. National Security Advisor Brzezinski was—literally—a minute away from waking President Carter to determine the U.S. response when he was informed that it was a false alarm.

The military was able to determine that this alarm was false because they had (barely) enough time to check the raw satellite data to see if it was consistent with the attack reports on their displays. They had this time because Soviet launch sites are a long way from the U.S..

North Korea is a small country bordered by a U.S. ally on one side and oceans open to the U.S. Navy on two others. Any U.S. attack will begin with submarine-launched missiles and strikes by stealth aircraft, and North Korean command and control systems will surely be among the first targets. Consequently, staff operating the North Korean early warning system will have very little time indeed to sort out true alarms from false alarms before having to decide what to do.    

Systematically analyzing the publicly available data on early warning system performance and false attack alarm rates, Barrett, Baum, and Hostetler put the probability of a U.S./Soviet nuclear war by mistake during the Cold War at about 2% per year (mean estimate)—or a 40% chance over 25 years. Given that North Korean early warning system staff will be operating under far greater time pressure than the already extreme time pressure faced by U.S. and Soviet staffs, it would be rash to assume that the chance of a U.S./North Korea nuclear war by mistake is lower than the chance of a U.S./Soviet nuclear exchange during the Cold War. 

In addition to time, accurately vetting attack alarms requires that early warning system officers exercise (in the words of a recent Chatham House report) independent and “prudent judgment, which might involve disobeying previous orders.” The Stanislav Petrov incident illustrates just how important judgment can be.

On 26 September 1983, Stanislav Petrov was the duty officer at the Soviet early warning system command. Soviet satellites detected the launch of a U.S. first strike, and the reliability of the report was rated as “Highest.” In this situation, Petrov’s standing orders required him to report the alert up the chain of command. In 1983, the Soviet leadership was extremely worried about the possibility of a U.S. first strike. It is therefore likely that, if Petrov had reported the alarm (of the highest reliability) to his superiors, the Soviets would have followed their doctrine and retaliated. Petrov’s gut instinct was telling him that the alarm was false. Fortunately for the world, Petrov ignored his orders and went with his gut instinct.     

To the best of my knowledge, there is no publicly available study on how early warning system officers deal with the uncertain, complex, and stressful situation of a nuclear attack alarm. However, studies of human error in airplane crashes (also uncertain, complex, and stressful events) show that people from uncertainty-averse cultures tend to deal with complex and stressful situations by following rules and avoiding independent judgments. Consequently, pilots from such cultures are more likely to crash planes.

South Korean culture is among the most uncertainty-averse in the world. Unsurprisingly, the South Korean airline KAL had a disastrous safety record until the airline explicitly set out to transform its operating culture. It is safe to assume that North Korean military culture will be even more uncertainty-averse than South Korean culture, and it seems unlikely that Kim Jong-Un will actively seek to instill an ethos of independent judgment and initiative in the North Korean officer corps. Consequently, the North Korean officer corps will be spectacularly ill-suited to exercise the prudent judgment that avoiding a nuclear war by mistake requires.

To recap: any U.S./North Korea deterrence relationship will inevitably be subject to political/military crises and early warning system failures that create the risk of a nuclear war by mistake. Plausible estimates put the risk of a U.S./Soviet nuclear war arising from an early warning system failure at 2% per year and examining the factors that drive this risk to suggest that the risk of accidental nuclear war with North Korea could be higher still.   

So, deterrence will work against North Korea. It just won’t work very well. 

Starting From Here

Kim Jong-Un vowed in his 2018 New Year’s address to “mass-produce nuclear warheads and ballistic missiles,” and this is one vow that he is keeping. If North Korea continues along its present path, it will ultimately be in a position to launch a nuclear strike that will kill tens of millions of people in the U.S., South Korea, and Japan. We are therefore starting from a poor position. Our goal has to be to find a way of dealing with the North Korean nuclear threat that minimizes the expected number of civilian deaths in the U.S., South Korea, and Japan.

The U.S. has three options: i) convincing North Korea to denuclearize via diplomacy; ii) limiting the probability of an attack with deterrence; and iii) eliminating the risk of an attack (and the North Korean regime) by means of preventive war.

Diplomacy is, of course, the ideal solution as it solves the problem in a way that does not involve any civilian deaths (and substantially advances North Korean economic development to boot). However, it is apparent that North Korean participation in the diplomatic process kicked-off by the Singapore Summit is just a charade to reduce the chance of a war (which was a definite possibility pre-Singapore) while North Korea strengthens its nuclear arsenal. The real options on the table are imperfect deterrence and preventive war. 

The U.S. could accept North Korea as a nuclear state and limit the risk of a nuclear war with deterrence. The deterrence option, therefore, enables the U.S. to avoid a preventive war now with a relatively weak North Korea, but only at the cost of a substantial risk of a nuclear war by mistake with a far more powerful North Korea in the future.

The only way that the U.S. can eliminate this risk of a nuclear war with a vastly more powerful North Korea in the future is to eliminate that threat now. Since eliminating North Korea’s nuclear threat will require eradicating the North Korean regime and destroying the North Koran military, this war could lead to hundreds of thousands of American, South Korean, and Japanese civilian casualties.

So, deterrence roulette or preventive war? A nuclear war with North Korea in the future (if it happens) will kill a far, far greater number of American, South Korean, and Japanese civilians than would a preventive war now. So, to conclude that deterrence with a risk of war later is preferable to a preventive war now, one would have to be very certain that the probability of a nuclear war by mistake in the future is very low. It is, of course, easy to reach this conclusion if one estimates the risk of a catastrophic deterrence failure the wrong way. However, if one estimates this risk correctly by examining the evidence on nuclear war near misses, I at least do not see how it is possible to be very certain that the probability of a nuclear war by mistake is very low.

North Korea is commonly referred to as “the land of no good options.” This expression does not mean that there are no good options. After all, an option is neither good nor bad in any absolute sense, but only relative to other options. And, logically, it is impossible for every option to be worse than every other option. Instead, this expression is best understood as a lament that we are starting from here with the options we have rather than starting from somewhere else with an entirely different set of options. I wish that we were starting from somewhere else too. But, due to the utter fecklessness of the Obama administration, we are where we are.

And starting from here, a preventive war now is a far better option than recklessly betting the lives of millions of Americans, South Koreans, and Japanese on deterrence roulette.

Kevin R. James is an economist in the Systemic Risk Centre at the London School of Economics (@kevinrogerjames).