In the wake of the attempted Christmas bombing of an American airliner en route to
The economic question is the optimal expenditure on preventing terrorist attacks on airlines. The question is bafflingly difficult because of the uncertainty associated with such attacks. Cost-benefit analysis of precautions is a reliable tool of economic decision making only (in general—I will suggest an exception below) when not only the cost of the precautions and the loss (cost) that will occur if the event sought to be prevented is allowed to happen can be calculated, but also the probability of the event if the precautions are not taken. For that second calculation is also necessary in order to estimate the expected loss if the precautions are not taken. If the loss if the event occurs is $x, and the probability that the event will occur unless precautions are taken to prevent it is .01, then, as a first approximation, one should spend up to $.01x to prevent the event from occurring, but not more.
But what is the incremental probability of a successful terrorist attack on an airline if the precautions instituted after the Christmas bombing attempt are withdrawn? Moreover, although that is the most difficult question, there is also uncertainty about the loss should such an attempt succeed. There are pretty good value of life estimates, which would suggest for example that an airline bombing that killed 200 people would inflict a loss of $1.4 billion (200 x $7 million). But that leaves out the terrible fear that these people would experience unless the bombing caused instant death (which it would not), plus the fear of other airline passengers and crew after the bombing, plus added time and safety costs of passengers diverted by fear to other means of transportation—other means that may be more serious, depending on how common successful terrorist attacks on aircraft become; for the death rate per mile is, at present anyway, markedly higher for automobile transportation than for air transportation. There is an instinctual fear of flying (easy to explain in terms of the ancestral environment in which the human brain developed, for in that environment heights were exceptionally dangerous), and as a result the prospect of being killed in an airline crash fills many people with particular dread; that prospect is a cost, like any other. For many people, it exceeds the expected accident cost of driving relative to flying.
But I want to focus on the uncertainty of the occurrence of a terrorist attack. Some statisticians, and some famous economists of yore such as Frank Knight and John Maynard Keynes, distinguish between calculable risk, as in my .01 example, and uncertainty, in the sense of risk that cannot be calculated with any confidence. No one can say with any real confidence what the probability of a successful attack in the next year on a
But it makes a big difference to the optimal investment in precaution whether the probability of such an attack is .0001 or .1 (as it could easily be if the attack was the first in a planned series), and whether the cost of the attack if it is successful would be $1.4 billion or $5 billion--or much more, as it could be, if one thinks that the Iraq and Afghanistan wars, which together have already cost at least $1 trillion, were a consequence of the 9/11 terrorist attacks. There is thus uncertainty in size of loss as well as in probability of loss. As a result of these dual uncertainties, the realistic expected-loss range could, on my assumptions, easily extend from $140,000 to $500 million.
How to pick a point in that range? The question may seem unanswerable, but it is not. The reason is discontinuity in the range of available precautions. Greater vigilance and more screening equipment and screeners are costly but there are sharply declining marginal returns. One might decide to place two or three guards on every international flight, but it wouldn’t make any sense to place 10 guards on every flight; the incremental benefit would be negligible. Similarly, maybe every security line in every international airport should be equipped with a body scanner, but it wouldn’t make sense to equip every security line with two body scanners. The inability of our intelligence agencies to pool information effectively will be costly to correct, but these are costs that have to be incurred anyway—to protect the nation from a range of terrorist and other threats, not just threats to airline safety.
There is probably a bias among security personnel in favor of doing more than can be justified by the kind of analysis that I have just offered: that is, a bias in favor of adopting some precautions that have little or even no efficacy in preventing attacks, such as subjecting children who have already been patted down in the security line to a second pat-down at the airline gate. The reason for the bias is bureaucratic, or in other words careerist: from the career perspective of a security officer, the worst thing that can happen is an exact repetition of a successful attack. For then no excuses, even if reasonable, for having failed to prevent the new attack will be accepted. So security agencies will tend to overinvest in preventing the repetition of previous attacks. It has been argued persuasively that the nation has overinvested in airline security since 9/11 relative to security against other attacks, for example on trains or subways, because we have thus far been spared such attacks, though other nations, notably