So I'm not persuaded that nuclear deterrence is that reliable. I've published on this in "The Myth of Nuclear Deterrence" in the Nonproliferation Review, an essay that won a $10,000 prize. I know that most people are pretty sure that nuclear deterrence works well most of the time. I have some doubts about that based on the record. Let's talk.

Hi Ward -

Thanks for being willing to discuss deterrence. It's a subject that sometimes makes my head spin with some of the more esoteric game theory papers.

One of the important points you make in your book is that deterrence is a broader concept than nuclear deterrence. It's said to be what keeps countries from attacking each other. Or in civil society, it's argued that prison and the death penalty deter crime. But the connections among those things seem more tenuous to me than just deterrence.

For example, countries don't attack each other because it's costly in terms of money and lives. Or people don't steal from their neighbors because they're happy with what they have or they like their neighbors.

So there's more to these things than deterrence by threat.

So let's look at nuclear deterrence. The big example of that is the US - USSR rivalry during the Cold War. The two countries just kept building more and more nuclear weapons until about the 1980s. And they never went to nuclear war with each other. In your book, Ward, you give some examples of where one or both countries pressed pretty hard and it looks like nuclear war was imminent. But somehow, someone managed to avoid actually dropping bombs when push came to shove. There are a number of ways to look at all this: an overall big view says that we avoided nuclear war, so deterrence must have worked. It seems to me, as in my homlier examples above, a number of factors could have been in play besides the horror of using nuclear weapons.

That's background. The Cold War situation of mutually assured destruction (MAD) may have offered a certain stability: two players who believed they were equally matched. Even if we assume that deterrence worked in that environment, how can we extrapolate to today's world? The US and Russia (as inheritor of the USSR's status and nuclear weapons) still hold the vast majority of the world's nuclear weapons, which they are taking out of service slowly. The political situation between them, while not entirely friendly, is far from the dangerous rivalry of the Cold War, and it appears that accident is more likely than intention to cause a nuclear exchange between them.

Then there are the other nuclear weapon states. Britain and France seem unlikely to nuke anyone. China has fairly explicitly said that its nuclear force is a deterrent (that word again!), not an offensive force. Israel isn't saying whether they have nuclear weapons, but we all know that they do, and they have been very willing to threaten attacks on other countries in their area with words that don't seem to eliminate nuclear weapons from consideration. India and Pakistan are explicitly in a nuclear arms race with each other, with the occasional terrorist attack (one today) that often leads to threats from one side or the other. North Korea may or may not have deliverable nuclear weapons. Their history has been one of threatening and occasionally attacking their neighbors, although there are some signs that their new leader Kim Jung Un may be less bellicose than his father.

Now, even if deterrence works, how does one calculate the probabilities of trouble in this kind of world? In the Cold War scenarios, there were only two players, and game theory works for two players. What now?

The problem of many countries versus two countries was the problem that Shultz et al. talked about in their first Wall Street Journal op-ed. It was the factor they pointed to that made nuclear deterrence unreliable. I think the problems run much deeper.

First, we know from crime studies that ordinary deterrence works only a certain percentage of the time. We're not sure exactly what that percentage is--it might be 40%, it might be 60%. But we know from extensive studies looking at whether the death penalty deters murder that ordinary deterrence fails some relatively large percentage of the time.

The claim has always been that nuclear deterrence is so frightening that it must work a much higher percentage of the time. Perhaps something like 96%. Or higher. Some even argue that it's perfect. I don't think the evidence shows that. It is certainly true that the prospect of nuclear war is frightening. It's not clear that people are restrained by frightening things. Even nuclear war.

Recall that much of the work done initially on nuclear deterrence was done using logic and game theory. It seems to me, that a far more fruitful approach would be to examine the psychology of death penalty deterrence and then try to draw comparisons with nuclear deterrence. In any case, we should be doing far more psychological work, far more work on neuroscience, and far more thinking about the frailties of human cognition than about game theory.

There is no question that deterrence with two players is more stable than deterrence with multiple players. What I worry about is not the possibility of nuclear war in the world we have today. It seems fairly clear, from a historical perspective, that the current period of peace and stability is an anomaly. With the coming of global warming, it seems likely that the future into for which we are sailing will be one of flood, drought, famine, and considerably increased instability. The prospect of a nuclear war today seems quite unlikely. But today's circumstances will not last forever. And one of the hallmarks of crisis is that it creeps up on people without warning.