The US government insurance scheme for nuclear power plant accidents no longer makes sense

The Japan Center for Economic Research, a source sympathetic to nuclear power, recently put the long-term costs of the 2011 Fukushima accident as about $750 billion. Contrast that with the maximum of $13 billion that could be available after a catastrophic US nuclear accident under the plant owners’ self-insurance scheme defined by the Price-Anderson Act. The Act will have to be renewed before 2025; Congress should seize the opportunity not only to reflect on the lack of insurance in the event of a catastrophic accident, but also to reconsider our approach to nuclear power plant safety altogether.

Price-Anderson frees nuclear plant operators and all firms involved in nuclear construction and maintenance of any liability for offsite accident damage. The only chance for additional compensation lies in the act’s declaration that if accident damages exceed the legal limit “Congress will thoroughly review the particular incident” and will “take whatever action is determined to be necessary” to provide full compensation to the public. In short, a Fukushima-level accident would toss the costs of compensation and cleanup unto the lap of Congress.

Is such an accident possible in the United States? The nuclear community argues that it is essentially an impossibility. It explains away the major accidents that have occurred (and were similarly regarded as essentially impossible before they occurred): The 1979 Three Mile Island accident, during which 60 tons of the uranium core melted but did not release significant radioactivity; the 1986 Chernobyl accident, which did release huge amounts of radioactivity, but the Soviet plant had vulnerabilities not present in US-type plants; and the 2011 Fukushima accident, which did release large amounts of radioactivity and did involve US technology but was triggered by an earthquake-tsunami combination that the United States has not experienced.

The US nuclear safety guardian, the Nuclear Regulatory Commission, prides itself on making “risk-informed” safety decisions using “probabilistic risk assessment.” To decide whether additional protection is needed, the NRC performs a cost-benefit calculation that weighs the benefit of reducing the risk of an accident—radiation-caused deaths and damage, expressed in dollars—against the cost of additional protection, perhaps safety equipment, to accomplish risk reduction. The cost of, say, additional equipment, is a comparatively firm figure; but the dollar benefit of risk reduction is a very soft number, which means the cost-benefit balance is suspect.

There is a more fundamental problem. Consider the NRC’s metric for risk, which underlies its approach to nuclear safety. For an agency so devoted to “risk-informed” decision making, the NRC is strangely vague about its a definition of risk. It says merely that risk has to do with three questions: “What can go wrong? How likely is it? What are the consequences?” The key question is how to combine the probability of an accident and the consequences that would ensue. In practice, the NRC analysts take risk to be the probable, or average, loss per year—that is, the product of the probability of a particular accident and the consequences (computer-estimated radiologically caused deaths and contaminated land, expressed in dollars). (This is by no means the only possible way to express a risk goal. One may, for example, choose a figure of merit that puts more emphasis on reducing consequences, which the NRC leaves open-ended.)

Average risk can be a useful measure in situations where there is plenty of data (for example, accidents at an intersection, used in deciding whether to put a traffic light or stop sign there). It is not, however a realistic figure of merit where data are lacking. Nor is the average risk the number that expresses what most people would think represents safety.

The main public risk of nuclear power plants comes from rare but devastating nuclear accidents. Because data on such accidents is sparse, the probability of their occurrence has to be calculated on the basis of a model, rather than obtained from experience. Moreover, the extent of an accident and its monetary consequences are postulated on the basis of models that are limited by analysts’ imagination. Who would have imagined, for example, that the Fukushima accident would involve several reactors? Or that Japan would subsequently shut down all its other nuclear power plants?

To see how the NRC approach works in detail, take a case the commissioners considered a couple of years ago: whether the possibility of a severe fire in a reactor spent fuel pool (where highly radioactive used fuel is stored) was reason enough to encourage plant owners to move spent fuel out of fuel pools into dry cask storage containers as soon as possible.

The NRC staff studied the possibility of a fire in US nuclear plants similar to the ones that suffered the Fukushima accident. (The United States operates about two dozen such reactors, all over 40 years old.) The staff told the commissioners that such a fire could produce a radiological release 25 times as great as the release during the Fukushima accident (and this was not the absolutely worst case). But the staff also estimated the probability of such an occurrence to be so infinitesimally small that, even multiplied by the dollar value of the harm and damage caused by a release 25 times greater than at Fukushima, the annual risk was far below the added cost of removing spent fuel from cooling pools earlier than is now required. In other words, there was no need to do anything, which was music to the ears of the NRC commissioners, who resist imposing any additional costs on the financially shaky nuclear industry.

What happened was that the NRC staff multiplied a very large number by a very small number—each a more-or-less educated guess—to get a highly uncertain result, which the commissioners then took at face value. One of the probability figures on the staff table of results (the chance of a cancer fatality within 10 miles of the plant) for a particular configuration is presented as 2 trillionths per year (2×10^-12). Or, to put it another way, if a plant kept operating forever in that configuration, the accident might happen about once in 500 billion years. That’s once in 30 lifetimes of the universe. Do such numbers mean anything?

Curiously, from the chairman on down, the NRC misstates the legal standard for its safety decisions. The NRC and its staff claim their job is to provide “reasonable assurance of adequate protection,” whereas the standard in the Atomic Energy Act is “adequate protection.” Under the law, their job is to provide adequate protection, period. Do the commissioners think the extra cushion of “reasonable assurance” justifies weaker regulation?

To return to the Price-Anderson Act: As we’ve seen, a catastrophic accident would render the US self-insurance scheme for nuclear power plants pretty much irrelevant. But the indemnification of all industry participants would remain highly relevant: The industry would be free of any liability for offsite death or damage, whereas the victims would have to go hat in hand to Congress for restitution. This is an enormous subsidy—consider, again, the $750 billion and counting tab for Fukushima—that the federal government provides the nuclear industry, one without which not a single US nuclear power plant would or could operate. Freedom from liability also has had a perverse effect on nuclear safety. Without the liability protection of Price-Anderson, industry incentives to develop nuclear designs safer than light water reactors would surely have been higher.

Freedom from liability was put into law in the 1950s to get the US commercial nuclear power industry off the ground. It was meant to be temporary, until industry and insurers got some experience with the new technology. But even as time went on, industrial organizations like General Electric and Westinghouse would not participate in the civilian nuclear program if they risked responsibility for offsite damage from a nuclear plant accident.

This government guarantee was an understandable demand at the outset of the nuclear age, when the risks of nuclear power were unclear and maybe a little scary. But it’s now half a century later, and the NRC commissioners tell us that probabilistic risk assessment is a highly developed tool. They urge their staff to make greater use of it in making regulatory decisions. Yet the vendors continue to maintain their insistence on freedom from liability for offsite consequences.

If you accept the NRC accident estimates, the risk the vendors would run without an exemption from liability would be very small, and likely a lot smaller than other corporate risks they routinely run. What is clear is that the nuclear firms—the largest of which possess an understanding of nuclear safety far beyond that of the public—do not believe the NRC safety conclusions that the risk of a catastrophic nuclear accident is infinitesmal. Nor do they accept that probable risk—probability of an accident times the consequences, were one to occur—as the right measure of risk to their companies. They don’t want to risk their companies, period.

If they don’t believe the NRC numbers, why should the rest of us accept them? Why shouldn’t we have the same protection from physical harm that the nuclear industry has from financial liability? And just as the nuclear vendors will not participate on terms that do not include indemnification from the overwhelming cost of a severe accident, so should the public have the analogous power to only accept future nuclear designs that can demonstrate that they preclude offsite harm. And the designs should demonstrate that level of safety in a clear way, based on physical principles, not on complicated probabilistic calculations put forward by interested parties.

Such new designs would eliminate the current dilemma of a federal nuclear self-insurance scheme that cannot, as a practical matter, cover the financial consequences to the public of catastrophic nuclear power plant accidents. But how to get there? One of the disincentives is the Price-Anderson Act’s limitations on industry liability for offsite accident consequences. That should get phased out.