Tobacco Documents Online

TRAPS AND ERRORS IN RISK ANALYSIS Lecture at George Washington University Washington, DC Richard Wilson Department of Physics and .Energy and Environmental Policy Center Harvard University Cambridge, MA 02138 March 31, 1982 Risk analysis in its most general form is commonsense com- bined with arithmetic. The errors are errors of commonsense. Rarely do we find errors of arithmetic, only failures to do the arithmetic. The fact that traps and errors exist should not, in my view, lead us to abandon risk analysis, but on the contrary, emphasizes.the need for good risk analyses. 'The first, and most often discussed error, is the demand for zero ;risk which appears in various forms. But while it is the new wisdom to.regard the zero risk advocates as rabble rousers, or at best middle headed idealists, I want toemphasize that there has in the past been a role for an attempt to reduce the risk to zero. This was the case when the major risk of premature death was that of communicable disease. This is brought out most clearly in a paper by Sir Richard Doll„'L Doll_showed that by most simple measures, health is better now than it was 100 years ago and is improving. This is because there has been a steady reduction in disease. I can also 1Dol]., Sir;Richard, "The pattern of disease in the post-infection era: national trends," Proc. R. Soc. Lond. B205, 47-61, 1979.

-2- illustrate this by a-plot of death rates versus time (Figure 1). The death rates are decreasing in all age groups except the one (15-24) where teenagers kill themselves with automobiles. We can also go back over a longer period and look at life expectancy. In the year 800 A.D. life expectancy was about 28. It increased to 45 one hundred and fifty years ago, and is now 72 years for men and 76 for women (I personally am jealous of the women). The number of people surviving to a given age falls fairly slowly (Figure 2, but falls off sharply at age 70; an interesting age, !:)ecause according to the nineteenth psalm, "the days of our years are'three score years and ten" and nowadays compulsory retire- ment in the U.S. Does not begin until then. Death before age 70 can . be considered premature and formerly premature death was mostly from diseas- Diseases were removed as causes of death, not by medical advances, but by technical and scientific ones. An Oxford historian has pointed out the influence of the drainpipe (for main drainage) in the European history of the early 19th century. Chlorination of drinking water also played a major role. We must not, therefore, turn our backs on technology, but use it wisely. But there is an important corol- lary; the expense of reducing disease was moderate. In each case there were a large number of dead bodies obviously attritutable to a disease; and once the cause was recognized, a N massive and rapid technological effort was made to remove the hazar~T; completely. This led to a general view that once a risk of disease~o is recognized it is possible to complete eleminate it--and it shou* be completely eliminated. This, then, is a justification for the ~ ~.

FIGURE 1 DEAT1i RATES BY AGE: UNCTI:D STATES, SELECTED YEARS 1M-1977 65 rurs ma arer under t ye.r I P I mnt~ 111/0 41111 4 1/n...._ . . ...q1`411aqibbNb4nlo......_. K6{ yl8rf f~ `a- ~ " ..00 >. 11111///1/1111111/1/1/111/IIB01111`I ~ fOG I['[°°'~~_' '•~~a\!~r - 101 1910 1020 1930 1m 1fl50 1ow 1970 N®T[r 1977 ®su e.e er..44l,alu au fe, ati.ne.l feln aro fuyr y.~ta ywn w 1f00.1l2S. llS~ 1ri0 :~r q1 prws 1Sta r.I,l.rurl. Ins 1077 TeOURCE: Nmcarai Cent:er ta Mostm Stnissxa. Drvts+on ot Vivt Stnmies. C H EWp tWO 1980

-3- zero :eisk ;approach. 'The reason that the zero risk advocates are wrong is that the situation has changed dramatically. We have almost removed communi- cable disease in the Western world as a cause of death. The second important error in a discussion of risks is a fail- ure to recognize the changed nature of the risks which terminate our :L:i.ves. The earlier risks can be called historical; the risk could often be calculated from historical data, and once the idea was suggested, the proof of causality was straightforward. There are some people, and for politeness I will not refer directly to them,, who have called for only regulating on these risks; for re- gulating only known measureable effects on health. If taken literally, this would prevent our controlling a large number of risks which are well worth reducing. We must recognize that risk is an expression of uncertainty. The uncertainty can arise in two ways; we can for example believe that cars kill people who cross the road, yet be uncertain that an individual is killed; or we can be unsure whether a new technology will ever kill anyone or not. The new risks of society are for new technology, about which we must speculate. It is improper to say "we do not know if there is a risk of air pollu- tion at present levels." If we do not know--there is a risk, but the magnitude of the risk may be uncertain. These two errors are at the root of our present discussion ~ about the Clean Air Acts. On the one hand some people still want ~ zero risk, forgetting that air pollution is not a communicable ~ ~ ~ W N

-4- disease and zero risk makes no sense, and on the other hand others are unwilling to regulate until dead bodies can be counted and reliably attributed to the source. No risk analysis is perfect. In those I have reviewed, and Crouch and I discuss a number in our book on Risk/Benefit Analysis2, the most important is omission of an important risk. It is foolish to compare risks of nuclear power and coal without mentioning the r:i.sk that either may contribute to war; the one perhaps by making nuclear fuel more easily available, the other by making energy more costly. We must insist on putting numbers on each and every risk we can, but there are some risks that cannot be easily expressed in numerical terms. If the numerically expressed risks are properly discussed, these other risks will be highlighted; all too often they are ignored. For example, a 400 page report on LNG safety ignores questions of sabotage: Instead of saying so in the abstract, introduction and summary, the sentence is hidden in the raLddle'. Yet for LNG and LPG facilities it is my considered view that the sitina should be governed by the possibility of sabotage.3 Samuel Johnson once wrote that "Round numbers are always falseP' and a risk analysis without a discussion of uncertainty can be false. As I noted before, it is in the region of uncertain risks-risks of a consequence which we are not even sure exists-- 2E.A.C:. Crouch and Richard Wilson, Risk/Benefit Analysis, Ballinger Press, Cambridge, MA, 1982. 3Richard Wilson, Testimony to Energy Facilities Siting Council, Massac:husetts (1979).

-5- that a risk analysis is the most important. In this case, however, it is vital to state the assumptions--early and often, as it used to be said about instructions to Cook County voters. But it must not be: thought that an uncertain number is useless. Dr. Samuel Epstei'.n and Samuel C. Florman in two separate articles in the Technology'Review4' 5 argue that the estimates of risk of exposure to vinyl chloride at low doses vary by a factor of 100,000, and are therefore useless. They are wrong on both counts. Estimates vary by an infinite factor; some scientists argue that the risk is finite, others that it is zero. Any finite number divided by zero gives; infinity (not 100, 000) . Yet risk analysis can be useful; as of 1981, 82 cases of liver angiosarcoma had been attributed to occupational exposure of vinyl chloride world wide. These were due to past high exposures over a 20 year period. Liver angiosarcoma is a rare cancer, and few attributable cases would have been missed. Allowing for some cancers still in the latency period, cancers at other sites in the body, and cancers from exposures in the general environment, I find a maximum of 1000 over 20 years from past exposures. Exposures have now been reduced 1000 fold, and most scientists would agree that the number of cases will fall at least linearly to 1 every 20 years worldwide, and I believe no liver angiosarcoma has been attributed to exposures in the last 5 years. 4Samuel Epstein, "Cancer inflation and the failure to regulate," Technology Review, Dec./Jan., 1980. 5Samuel C. Florham, "Living with technology: tradeoffs in paradise," Technology Review, Aug./Sept., 1981.

-5a- A number of people argue that natural foods are better and .our p.roblems of cancer and so on arise from artificial chemicals. It is conventional wisdom to ridicule this. This ridicule is illuatrated in a New Yorker cartoon (Figure 3) showing that natural ingredients were used for witches' brews as well as for honest men's sustinence. Indeed, I like to point out to our local health food store that peanuts dried in the sun dry more slowly than those dried artificially and develop more molds and therefore have higher concentrations of aflatoxin B1. But it would be an error to ridi- cule the "health food nuts" without trying to see if there is a legii:Lmate point in their favor. There is. We have historical data on natural foods, and believe a major disast:er is unlikely, whereas if a new chemical (such as thalido- mide) is put on the market, thousands of people may be hurt before we know what the danger is because of a latent period. Etut this is no excuse for turning our back on technology. Naturadl phenomena may wipe out the human race, and got close to doind so in the black death. That particular risk is probably now eliminated by technology. It is important to understand a proper flow of information in reaching decisions about risk. Risk analysis is an aid to a decision and should not be a decision in itself. It is a grevious error to confuse the risk analysis with the decision itself.

"It's going to be great! All natural ingredients. "

-6- :1 illustrate the flow of a discussion of risks in Figure 4. The information flows from the scientist to the assessor; to a compai:ison with costs and benefits to a decision maker (evaluator) who may be a bureaucrat, a politician or an ordinary citizen. All interested parties, unions, public, churches, industry, fishermen, etc. must be identified and these will have their own value judg- ments,. Value judgments enter into the decision, but should not in- fluence the assessment and comparison. Of course the assessor has to know the questions for decision, and some idea of the alterna- tives that are open to the decision maker in order to write down his assessment in such a way that the decision is properly illumi- nated. For this reason I put a dotted line between questions for decision and assessment. The risk assessment is most useful when it is kept distinct in this way. The Carcinogen Assessment Group of EPA has been steadily improving in this regard; they usually quote the upper 95th pscentile of the risk, but state also that it might be zero. They nowadays avoid directly supporting a decision. I make here an analogy with the Anglo-Saxon legal system. The judge decides on questions of law; the jury--the assessors--decide questions of fact. The judge instructs the jury before they.assess the facts as to what is legally relevant. ~ :Hcwever the separation is sometimes deliberately broken as a ~ histo:rical example shows. in Cromwell's time, a law was passed ~ making it illegal to kiss in public. Perhaps the law was sensible, ~ ~r? ~ ~

RISK ANALYSIS ASSUMPTIONS Risk { m It HAssessfnNN ~ Numbers ± ~ Uncertainty ~~cientific Data f ~ Econom~c ~ Cost gineering sse{ D E22titTk!5ment { { ~ ~ { 1 I Interested I Z Value F'arties Questions for Decision Benefit Assessment Alternative Possible Decisions w 1 De Minimis Risk. STOP 10-s/yr. Occupational 10-e/yr General Total Societal Impact 10/yr Numbers ± ~I Numbers ± Results of decision Knighthoods , Insults, Anonymous letters etc. Risk/Benefit Risk / Risk Risk/Cost Comparisons