estimate the probability (per reactor-year of operation) that accidents of varying severity would occur.12 Its stated findings are that the actuarial risks (sums of the probabilities of consequences multiplied by the severity of consequences) are very small, and that the chances of severe accidents that would cause large numbers of casualties are extremely small—so small as to be within the range of risks we hardly deign to consider. Nevertheless, these findings have been challenged on several grounds: that the statistical treatment is in some respects incorrect and in others misleadingly presented; that casualty figures for the most severe types of accidents are underestimated; and that accident frequencies may have been overestimated (industry analysts typically arguing the latter, and nuclear critics, the former).13

The Risk and Impact Panel of this study examined the controversy, but could not reach more than qualitative conclusions. These conclusions are, briefly, that the statistical inferences of the report should be corrected upward, owing to the report’s use of medians rather than means of certain probability distributions where the correct procedure would have been to use the mean values, and that in addition to this upward correction in the “best estimate” of the accident risk, the counterclaims of optimism and pessimism for accident frequencies and consequences ought at least to be interpreted as indicating that the uncertainties accompanying both probabilities and consequences are greater than the uncertainty factors stated in WASH-1400.

We would estimate higher average risks than WASH-1400—not so high as to be alarming, but with sufficient uncertainty that there remain legitimate grounds for controversy whether the risk of reactor accidents ought to be an important consideration in decisions about nuclear power. Thus on safety grounds alone, the expansion of nuclear power would be acceptable,* provided the rate of expansion were consistent with the rate of improvement of knowledge about accident risks, especially reductions in uncertainty.

The reactor accident at Three Mile Island occurred after most of CONAES’s deliberations had been completed. That fact and the fact that several investigations of the accident are still in progress make it inappropriate for CONAES to discuss its implications at length, and impossible to do so with authority. The information so far released about the accident (and interpreted by nuclear specialists on the committee) seems consistent with CONAES’s cautious, positive findings on reactor safety.

Another element of public concern is apprehension about the ability of

*

Statement 5–7, by J.P.Holdren: Decisions on what is “acceptable” are the business of the political process, not of this or any other NAS committee.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement