Risky Business

A book tries, and fails, to quantify catastrophic risks.

Richard Posner’s Catastrophe: Risk and Response was inspired, he says, by Margaret Atwood’s 2003 novel Oryx and Crake, set in the near future, which imagines the extinction of the human race in a world menaced by bioterrorism and uncontrolled technological advance.” He hastens to assure his readers that he hasn’t become “an apocalyptic visionary,” but a peculiar unreality suffuses his book, which proves unexpectedly blind to the real threats we face.

By “catastrophic risk,” Posner means risks that have a low probability of materializing but are likely to create nearly unimaginable harms if they do. He believes they are real and growing. His examples include falling asteroids that could wipe out a quarter of the earth’s population within 24 hours; global warming, which could cause floods (“Harvard gone the way of Atlantis”) followed by “snowball earth”; nanotechnology that could envelop the earth in “gray goo”; bioterrorism; and “superintelligent robots” that “may kill us, put us in zoos, or enslave us, using mind-control technologies to extinguish any possibility of revolt, as in the movie The Matrix.”

Posner’s thesis when discussing these emotionally laden subjects is as deadpan as his prose: “[T]he tools of economic analysis—in particular cost-benefit analysis—are indispensable to evaluating the possible responses to the catastrophic risks.” Unfortunately, assigning precise numerical weights to the costs and benefits of preventing catastrophic risks is a daunting challenge, and Posner’s attempts to do so are numbingly technical and ultimately unsatisfying. In the end, balancing liberty and security involves disputed questions of value rather than precisely quantifiable facts—questions that must be resolved not by experts but by politics.   

Consider the possibility that atomic particles, colliding in a powerful accelerator such as Brookhaven Lab’s Relativistic Heavy Ion Collider, could reassemble themselves into a compressed object called a stranglet that would destroy the world. Posner sets out to “monetize” the costs and benefits of this “extremely unlikely” disaster. He estimates “the cost of extinction of the human race” at $600 trillion and the annual probability of such a disaster at 1 in 10 million. These figures are “arbitrary,” he acknowledges, because it is impossible to calculate the real probability of a stranglet disaster, and scientists who have attempted to do so have been attacked for politicizing the question. His attempts to calculate the benefits of developing the RHIC are also arbitrary because there’s no impartial way to calculate the value the public assigns to research in particle physics. After elaborate calculations based on arbitrary figures, he suggests that perhaps the costs and benefits can’t be precisely monetized: Congress, or the public, could be told instead that “there is one chance in 10 million of a world destroying accelerator accident that could be avoided by closing down RHIC at a cost in benefits forgone estimated at $2.1.” What then, is the point of the elaborate calculations?

Posner argues that even  arbitrary figures could promote reasoned decision-making that might close down some of the most dangerous research in RHIC. But if the decision is made by democratic bodies accountable to the public, this may be too optimistic. Behavioral psychologists have found that the public tends to make judgments about risk based on emotional feelings about whether something is good or bad, safe or dangerous, rather than on a dispassionate calculation of costs and benefits. Paul Slovic of the University of Oregon, for example, has argued that risk is a subjective concept that has different meanings to different citizens—some focus more on the low probability of particular threats, others on the potential severity, still others on the possibility that children might be harmed.

These differences in risk perception can only be resolved through political negotiation.  But democratic politics is an enterprise for which Posner has contempt. He is addicted to the rule of experts, and he proposes a series of arid and (for a self-styled pragmatist) surprisingly impractical policy solutions for applying cost benefit analysis to risk calculation: a “science court” of experts that would review dangerous government  research projects; the creation of an international environmental protection agency to enforce a modified Kyoto Protocol under the auspices of the United Nations; a federal review board that would forbid any scientific research that poses an “undue risk” to human survival. Few of these proposals have any realistic chance of being adopted in America. And even if they were adopted, public emotionalism would continue to demand irrational (or as the behavioral psychologists say, “quasi-rational”) allocations of resources that would thwart the experts’ recommendations. Although Posner promises to monetize the costs of these psychological and political impediments, he fails to do so.

Even if Posner’s proposals could be imposed by judicial fiat, which they can’t and shouldn’t, they seem underwhelming on their own terms. In a surprising hole at the end of the book, Posner declines to offer practical examples of how cost-benefit analysis could cast precise light on the very real terrorist threats that menace us. Consider the possibility of biological terrorism. Posner argues plausibly that the government should balance the costs of abridging civil liberties against the benefits of preventing terrorist catastrophies. He correctly criticizes some civil libertarians for failing to calculate these costs and benefits. But he then proves unable to calculate them himself—and dismisses those, including me, who have argued that the public tends to overestimate the likelihood that they will be personally harmed by especially frightening forms of terrorism that are easy to visualize.

After 9/11, in fact, respondents in a poll perceived a 20 percent chance that they would be personally hurt in a terrorist attack within the next year. These predictions could have come true only if an attack of similar magnitude to 9/11 occurred nearly every day for the following year. Although the actual probability of terrorist attacks is impossible to measure, nothing in al-Qaida’s history suggests anything like the capacity to produce 9/11-scale attacks on a daily basis.

Posner also insists that we should calculate the costs to liberty and privacy of extreme police and military measures, such as torture, and the likelihood that these extreme methods would, in fact, increase security. But he then proves unable to calculate these costs and benefits as well. “I have no idea whether [torture] is necessary,” he says after a long digression on the hypothetical benefits of torture, and the effort to monetize the benefits of privacy similarly defeats him.

It may be possible, in fact, to attempt to calculate the costs and benefits of some security technologies with more precision than Posner offers. Consider the government’s original proposal, called Total Information Awareness, to use data-mining at airports to determine whether individual travelers had consumer and travel patterns that resembled the 19 hijackers of 9/11. After the system was proposed, libertarian critics, using cost-benefit analysis, pointed out the great danger of false positives: Even if the system were 99 percent accurate in identifying terrorists, a 1 percent error rate applied to 300 million travelers would mean that 3,000,000 (that’s .01 x 300 million) of those identified as potential terrorists would be wrongly identified. But if we assume that the next attack will look nothing like the last one, a data-mining system that looked for passengers who took flying lessons in Florida, for example, is more likely to have something closer to a 1 percent accuracy rate. Such a system would falsely accuse nearly all innocent travelers of being terrorists and correctly identify only a fraction of terrorists while missing nearly all of the real terrorists. No rational evaluation of costs and benefits would support such a system, which is why the government correctly abandoned Total Information Awareness and replaced it with a system designed to verify a traveler’s identity rather than model suspicious behavior.

Posner does not describe the successful attempt by civil libertarians to lobby against badly conceived security technologies by applying the methods of cost-benefit analysis because actual political debates have no place in his elaborate models of catastrophic  risks. He wants to reorient legal education to produce polymaths like himself, requiring law students to demonstrate “basic competence” in math, statistics, and science so that they could replicate his Herculean feats of interdisciplinary synthesis. Specialists in various disciplines may benefit from the collaborative research projects that Posner usefully outlines. But the greatest challenges that menace us cannot be precisely quantified by science; they are psychological and political.