Choosing to deter North Korea is to engage in a gamble: you avoid the costs of a preventive war today when North Korea is relatively weak, but you run the risk of an accidental nuclear war later when North Korea is vastly more powerful. Using plausible estimates of the probability of accidental nuclear war derived from the U.S.-Soviet experience during the Cold War, I find that gambling on deterrence will lead to 7.5 million U.S.-South Korean-Japanese deaths on average (under optimistic assumptions) while a preventive war now will lead to 1.4 million deaths (under pessimistic assumptions). So, not only is deterrence a gamble, it is a reckless and foolish one. Preventive war is the wise and prudent response to North Korea's nuclear threat.
Dealing with North Korea's Nuclear Threat: The Options
The North Korean regime is completely committed to its nuclear program, and no diplomatic option will lead to North Korea freezing or disbanding it. This being the case, North Korea is and will be capable of launching a nuclear strike at the U.S. and its allies (either on purpose or by accident). To deal with this risk, the U.S. can either: i) limit North Korea's incentives to launch a nuclear strike through deterrence; or ii) eliminate North Korea's ability to launch a nuclear strike by means of a preventive war that eradicates the Kim regime and North Korea's WMD program.
Deterrence can (let us suppose) eliminate North Korea's incentive to launch a deliberate attack, but deterrence cannot eliminate the non-trivial chance that North Korea launches a nuclear strike by accident. Preventive war now can eliminate the risk of both a deliberate and an accidental nuclear attack in the future. But, obviously, a war aimed at eradicating the Kim regime will inevitably lead to North Korea launching (or at least attempting to launch) a nuclear strike at South Korea and Japan. So, by choosing deterrence rather than preventive war, the U.S. would be taking a gamble that the benefit of avoiding a war now is greater than the risk posed by the chance of accidental nuclear war in the future. Is this a wise or a foolish gamble?
Recommended: 8 Million People Could Die in a War with North Korea
Let us measure the cost to the U.S. and its allies of the preventive war and deterrence options by the number of by civilian deaths that follow from each choice in South Korea, Japan, and the U.S. I implicitly assume here that: 1) economic and other costs are proportional to lives lost; 2) military deaths are small in proportion to civilian deaths. I assume that North Korea will pose a threat for 30 years (until 2048), after which time the probability of nuclear war falls to zero.
The expected number of deaths that results from each option is a function of the deaths resulting from a war if one happens at a given point in time and the probability of a war at that point in time. Consider each factor in turn.
North Korea's Nuclear Arsenal
Following Congressional testimony by David Albright of the Institute For Science and International Security, I assume that North Korea's 2018 nuclear weapon inventory consists of 25 weapons each with a yield of 20 kt and that North Korea will be able to build four additional weapons per year. In light of North Korea's now demonstrated ability to construct thermonuclear weapons and North Korea's rapid advances in missile design, I further assume that all new weapons will have a 250 kt yield and that these weapons can be launched at the U.S.
On targeting, I assume that: 1) the 20 kt weapons will be targeted at South Korea and Japan with a 50%-50% split; 2) the 250 kt weapons will be targeted at South Korea, Japan, and the U.S. with a 25%-25%-50% split; and that: 3) only 50% of weapons successfully detonate over their target (due to a combination of mechanical defects, guidance problems, and missile defense). These targeting assumptions do not make a material difference to the overall outcome of the analysis. And while the actual proportion of successful strikes is of course highly uncertain and will affect the absolute number of deaths resulting from each option, any constant success rate used will not affect the ranking of the deterrence and preventive war options.
I calculate the deaths a successful detonation will cause by using the NukeMapwebsite. Since not every weapon can detonate in the same place, I set a weapon's impact equal to the average of a detonation over a range of cities. A weapon that detonates in Japan will cause deaths equal to the average of an air detonation over Tokyo and Kyoto. A weapon that detonates in South Korea will cause deaths equal to the average of an air detonation over Seoul and Busan. A weapon that detonates in the U.S. will cause deaths equal to the average of an air detonation over Los Angeles, San Francisco, Chicago, Washington D.C., and New York. I weigh South Korean, Japanese, and American lives equally. So, a successful detonation of a 20 kt weapon causes expected deaths of 112 thousand and the successful detonation of a 250 kt weapon causes expected deaths of 546 thousand.
Given these assumptions, a war in 2018 will cause 1.4 million Japanese-South Korean-U.S. deaths (25 weapons (20 kt) * 50% chance of successful detonation * 112,000 deaths per successful detonation). The number of deaths resulting from war will increase by 1.1 million per year (4 additional weapons (250 kt) * 50% chance of a successful detonation * 546 deaths per successful detonation). It follows that a war in 2019 will cause expected deaths of 2.5 million, a war in 2020 will cause expected deaths of 3.6 million, and so on. A war in 2048 will cause expected deaths of 34.2 million.
The Risk of Accidental Nuclear War
While both the Americans and the Soviets were rational actors during the Cold War (at least as rational as Kim Jong Un, anyway) and while both the Americans and the Soviets knew that nuclear war would have disastrous consequences, there were nonetheless a number of occasions when the Americans and the Soviets came very close to starting a nuclear war by accident. These near misses arose from a combination of military misunderstandings, technical malfunctions in early warning systems, and human error. To illustrate:
Military Misunderstandings: During the Cuban Missile Crisis, the Soviets sent a squadron of 4 submarines armed with nuclear torpedoes towards Cuba with the intention of establishing a base on the island (Operation Anadyr). The subs ran into the U.S. naval blockade. In the tense circumstances of the Crisis and under an enormous amount of stress, the captain of one of the subs came under the mistaken impression that his sub was under attack by a U.S. task force led by the carrier USS Randolph and that war may have already broken out. Thinking that they were doomed, the captain cried out "We will die, but we will sink them all..." using the nuclear torpedo. It is highly likely that a Soviet attack that destroyed a U.S. carrier with a nuclear weapon at the height of the Cuban Missile Crisis would have led to a full-scale war between the U.S. and the Soviet Union, and that this war, in turn, would have led to a nuclear exchange. Fortunately, the squadron chief of staff was also on board, and he was able to dissuade the captain from pursuing that course of action.
Reflecting upon the Crisis, Graham Allison wrote that while President Kennedy thought that the chance of nuclear war was between 33% and 50%, "what we have learned in later decades has done nothing to lengthen those odds."
Technical Malfunctions: In 1983–a time of high U.S.-Soviet tension–the Soviet Union's early warning system detected a signal that the U.S. had launched a first strike. The system indicated that the signal was of the highest possible level of reliability. In this situation, it was the duty officer's responsibility to report the alert and its reliability level to his superiors. Soviet nuclear protocols then required that the Soviets launch a counter-strike before the U.S. missiles hit. Fortunately, the duty officer on the night was Stanislav Petrov. He had a skeptical streak due to his scientific background, and he thought that the attack signal might be too reliable to be real. Torn between reporting and not reporting the alarm, he finally decided to disobey his explicit orders and to instead notify higher command that the system had malfunctioned. He believed that if any of his colleagues (all of whom had a military education) had been on duty that night, they would have reported the alarm as an attack. If the alarm had been passed on, it is possible and perhaps even likely that the Soviets would have followed protocols and launched an accidental first strike at the U.S.