Wednesday, September 03, 2008

Risk and McCain

One of the key points in my recent discussion of climate change is the importance of unlikely but extreme disaster scenarios. Unfortunately, voters usually aren't responsive to the magnitude of such risks, even if an objective analysis demonstrates that they are critical to proper policymaking. This raises an interesting question: what other risks are unaddressed in the political mainstream?

One obvious answer comes to mind: nuclear war. Of course, we can't say with certainty that there is a nontrivial chance of nuclear war. The political and structural factors leading to nuclear conflict are far too complicated to model mathematically, and the very fact that we are around to discuss the question indicates that we have no empirical evidence to construct a probability distribution. We're left with vague subjective considerations, ill-suited for rigorous risk analysis.

But any eventuality that involves the destruction of the world as we know it demands our urgent attention, even if we can't estimate its likelihood with any precision. A principled refusal to consider risks so opaque that they lie "outside of a probability distribution" (as Jim Manzi puts it in another context) may cause us to ignore the most dangerous possibility of our time.

Consider the implications of a 1% chance of nuclear war between the United States and Russia within the next decade. If you take the position (as I do) that the near total destruction of humanity is thousands of times worse than practically any other policy-relevant scenario at hand, even a 1% chance of nuclear confrontation should be enough to dominate your decision-making. If there is any way to eliminate that 1% possibility, or even to clip it a little to 0.9%, you should pursue it at the expense of almost all ordinary political considerations.

Do I know that the probability is 1%, or even within an order of magnitude of 1%? Of course not -- again, there is no rigorous way to estimate the likelihood of an event as theoretically and empirically hard to quantify as nuclear war. But retreating to my intuition for a moment, I think that in many political environments, 1% is a very plausible estimate, perhaps even a lower bound. A foreign policy that aggressively confronts Russia within its sphere of influence, reviving the tension that existed during the Cold War, creates a very real chance of all-out conflict.

How? The possibilities are almost too numerous to list. Maybe a leader on either side will prove unpredictable and irrational, moving a situation beyond the point of no return. Maybe America's effort to install missile defense systems on Russia's periphery will make a paranoid Russian government suspect a first-strike attempt, and jump to retaliate after some faulty sensor readings. Maybe a deranged jihadist organization, seeking to destroy both America and Russia, will take advantage of existing tension by detonating a nuclear weapon and disguising it as an attack from the opposing nuclear superpower. Maybe an atmosphere of suspicion, coupled with thousands of massively destructive weapons on hair-trigger alert, will lead to a escalation that we never could have predicted.

A 1% chance of nuclear war, or anything close to it, is unconscionable. And while it is difficult to tell exactly which policies best prevent nuclear disaster, it is easy to determine what we should not do. We should not elect a president who advocates gratuitously insulting Russia by expelling it from the G8, alludes to religion while declaring his solidarity with a nation at war with Russia, and has established a pattern of making major decisions in an impulsive and poorly considered way.

To me, this is the most convincing case against McCain of them all...


ASP said...

Interesting premise - and "risky" does seem to be shaping up as an Obama campaign theme. Keep in mind, though, that the "1%" argument can be shaped to almost any policy. In fact, according to Ron Suskind, Cheney used precisely that formulation - we can't afford a 1% risk of nuclear terrorist attack -- to justify implementing a torture regime. Of course, many people think that torturing hundreds of terror suspects for years on end, many arrested on the flimsiest of grounds, increases the risk of major terrorist attack.

Matt Rognlie said...

Good points -- the Cheney reference is an important one, and it's important to distinguish between my "1 percent" argument and his. Specifically, I don't think that we should treat events with an estimated 1 percent probability as if they are certainties; we simply should simply say that they have a chance of 1 percent. In the case of all-out nuclear war, this still results in enormous expected damages that should be pivotal in our policymaking.

A nuclear terrorist attack, although hardly benign, would cause a much, much, much smaller amount of damage than all-out nuclear war. In fact, a 1% chance of a nuclear terrorist attack isn't nearly large enough to be dominant in our policy decisions: to take a crude estimate, if 1 million people will die in an attack that will occur with 1% probability, then it has an "expected cost" of 10,000 lives, which isn't so far from the costs of our overseas military adventures. (And much lower if we include the deaths on the other side...)

By the way, I also completely agree with the point about our policies increasing the risk of terrorism.