Andrew Davies and Rod Lyon have had an interesting debate here recently about nuclear weapons. Andrew has argued that we need to take the risk of nuclear war more seriously and that we should therefore work much harder to eliminate nuclear weapons. Rod sees the risk of nuclear war as very low, and argues that eliminating nuclear weapons would make the world more dangerous by making major conventional war more likely.

I’d like to triangulate their debate by supporting Andrew’s pessimism over Rod’s optimism about the risk of nuclear war, but offering a rather different view from Andrew’s of what we should do about it.

Rod argues that nuclear war is unlikely, and that nuclear weapons prevent major conventional wars, by extrapolating from the experience of the 73 years since nuclear weapons appeared. There’s something in what he says. It’s plainly true that since 1945 nuclear weapons have not been used, and that there have been no full-scale major-power wars, and it’s very probable that the existence of nuclear forces has helped prevent major wars from breaking out.

The question is how far we can credibly extrapolate from that. My problem with Rod’s analysis is that when he concludes on this basis that nuclear weapons prevent war, his sample size is too small. The most you can say is that they have prevented war so far, but almost all our data comes from one particular historical episode—the Cold War, the only full-scale hegemonic rivalry we’ve seen since 1945.

And as Rod himself says in his response to Andrew, every strategic contest is different and unique. The Cold War certainly was: its progress and outcome reflected the very specific strategic, ideological, geographic, economic, military and internal political characteristics of the rivals and their rivalry, and the skills and experience of those who made the big decisions on both sides.

In those specific circumstances, nuclear weapons appear to have prevented major war. In particular, it seems to me that that was because the circumstances of the Cold War created a situation in which each side was convinced that the other was resolved to fight a nuclear war, despite the catastrophic consequences, in order to prevent any significant shift in the other’s favour. Deterrence worked in the Cold War because both sides believed—or, to be more precise, believed enough—that the other wasn’t bluffing.

I’d share Rod’s optimism that nuclear weapons will keep on keeping the peace for as long as that is true. But how sure can we be that this will be so in future large-scale strategic contests between nuclear powers? And even more urgently, how sure can we be that this is true in the contests we see today?

The risk of nuclear war between, say, the US and China is not that either side will decide that it’s worth fighting one to achieve their strategic goals. The risk is that they will convince themselves that the other side is not willing to fight one to stop them, and they will turn out to be wrong.

How likely is that? How likely is it, for example, that China would come to believe that America would not fight a nuclear war to defend Taiwan? How likely is it that Washington today can display the statecraft required to convince Beijing that it would, or even understand what is required to convince China of its resolve?

I think these likelihoods are very low, in part because of the almost universal assumption in Washington that China could not possibly doubt America’s resolve in Asia. This is dangerously complacent, when the US has done so little to demonstrate that resolve, and provided so many good reasons for Beijing to doubt it. That makes the risk of nuclear war between America and China really quite high.

So what should we do about it? Andrew thinks we should redouble our efforts to eliminate nuclear weapons. I certainly agree that that would be a great thing to do, but it seems a very inadequate response to the risks we face today. After all, there’s no reason for confidence that nuclear weapons will be eliminated anytime soon.

That being so, we should focus on the more modest but more achievable task of trying to reduce the risk that nuclear weapons are used. And how could we do that? There seem to be only two sure ways.

The first would be to establish the clarity of resolve that underpinned successful nuclear deterrence in the Cold War. That’s harder for America than for China, but it wouldn’t be impossible. It must, however, be real, because bluffing is almost bound to fail. That means Americans must first debate and decide whether they really are willing to fight a nuclear war and risk losing cities to save Taiwan. The answer is most probably ‘no’.

The second would be to step back from the strategic contest that is now escalating between Washington and Beijing. Those who share Rod’s confidence that nuclear weapons prevent major war and will therefore never be used in combat aren’t much worried about increasing competition between nuclear powers, because they are sure it won’t lead to war, or that if war comes it won’t become nuclear. They are therefore less uneasy about escalating strategic rivalry, and more inclined to oppose policies that aim to prevent or limit it.

Those who, like me, see a real risk that rivalry between great powers leads to war, and that war between nuclear-armed powers can easily go nuclear, tend to place a higher priority on limiting rivalry, even at the price of making concessions that we would otherwise be very reluctant to make. Because, after all, we have to ask what is so important that we would be willing to fight a nuclear war to avoid it.

That’s a hard question, and it’s understandable that so many people prefer to ignore it in favour of debating the much easier one about whether nuclear weapons should be abolished.