information onch6mical risks Zhird in a three-part series
Vincent T.Cnvdn Columbia University New Yo&. NY 10032 The Emergency Planning and Commun i Right-to-Know ~ Act of 1986 requires companies to submit data to federal, state, and local agencies on routine and accidental releases of toxic chemicals into the environment. As this act and other community right-to-know laws and referenda such as California’s 1444
Environ. Sci. Technot., Vol. 23, No. 12, 1989
Proposition 65 go into effect, two events are likely to occur. First, the d e mand for information on chemical risks will increase sharply. Second, industry and government officials will increasingly be called upon to explain information on chemical risks and to put this information into perspective. Several authors have argued that risk comparisons provide this perspeaive (1-4). In a typical risk comparison, the risk of the chemical in question is compared w!$ the risks of other substances or activitles. Because comparisons are perceived to be more intuitively mean-
inghl than absolute probabilities, it is widely believed that they can be used effectively for communicating information about chemical risks and other hazards. A basic assumption of the a p proach is that risk comparisons provide a conceptual yardstick for measuring the relative size of a risk, especially when the risk is new and unfamiliar. Interest in risk comparisons has increased in part because of perceived difficulties in communicating quantitative information on chemical risks to the public (5-8). Government and industry officials often complain that lay-
0013-936W89/0923144~01.~/00 1989 American Chemical Sooiety
people lack the ability to understand quantitative information on chemical risks and that this leads to misperceptions and misunderstandings. Representatives of public groups and individual citizens are often equally frustrated. They perceive government and industry to be uninterested in the public’s concerns, unclear in communications about risks, unwilling to take actions to solve seemingly simple and obvious risk problems, and reluctant or unwilling to allow participation in decisions that affect their lives. Risk comparisons have several strengths that address important facets of this problem. They appear to be compatible with intuitive, natural thought processes such as analogies; they avoid the difficult and controversial task of converting diverse risks into a common unit (e.g., dollars per life lost or per day of pain and suffering); and they avoid direct numerical reference to small probabilities, which can be difficult to comprehend and evaluate in the abstract (9). Given the potential value of risk comparisons for informing people about chemical risks, this article has two main objectives: to identify the primary strengths and limitations of risk comparisons for communicating righttn-know information on chemical risks and to identify means by which risk comparisons can be improved. Approaches to risk comparisons The risk analysis literature contains two basic types of risk comparisons: comparisons of the risks of diverse activities and comparisons of the risks of similar or related activities. Each type is briefly described below. Comparisons of the risks of diverse substances and activities. In this type of comparison, the risks of a new or existing substance or activity are compared to the risks of a diverse set of substances and activities using a common scale or metric. For example, the risks of particular chemicals and chemical manufacturing processes have been compared to the risks of smoking, driving, flying, sunbathing, bicycling, eating peanut butter, drinking diet soda, and mountaineering (1-2). An underlying assumption is that the risks of the new or existing substance or activity can be more easily appreciated if compared to the risks of familiar activities. The validity of this assumption is discussed later in this paper. For comparing risks, analysts have used a variety of scales, including scales based on the annual probability of death, the risk per hour of exposure, and the overall loss in life expectancy (1-4, 10-13). Several different formats are used for presenting results. For ex-
TABLE 1
Risks that increase the chance of death by one part in one million Smokino 1.4cioarettes Drinkingone-h& liter 01 wine Spending 1 h in a coal mine
~.~~
Soendina 3 h in a.coal mine ...~~ ~. Livino 2 aavs in New York or Boston Traveiing G m i n by canoe Traveling 10 mi b bicycle navelina 3~ v car ~ - mi- l-, FGing~lOWmi by jet Flying Woo mi by jet Living 2 months in Denver on vacation from New York Living 2 months in average stone or ~~~~
~
-
~
~
brick buildina One chest X-ray taken in a good hospital Living 2 months with a cigarette smoker Eating 40 tablespoons 01 peanut butter Drinking Miami drinkina water for 1 vear Drinking 30 12-02 cans‘bf diet soda. Living 5 years at site boundary of a typical nuclear power plant in the open Drinking loo0 W o z soft drinks from recently banned plastic bottles Living M years near a polyvinyl chloride plant Livino 150 vears within 20 mi of a
Cancer, heart disease Cirrhosis of the liver Black lung disease Accident Air pollution Accident Accident Accident Accident Cancer caused by cosmic radiation Cancer caused by cosmic radiation Cancer caused by natural radioactivity Cancer caused by radiation Cancer, heart disease Liver cancer caused by aflatoxin B Cancer caused by chloroform Cancer caused by saccharin Cancer caused by radiation
I
Cancer from acrylonitrile monomer Cancer caused by vinyl chloride (1976 standard) Cancer caused by radiation Cancer from oenzopyrene Cancer caused by radiation
Sourse: Reference 2
ample, Wilson (2) identified a set of activities that increase a person’s chance of death (during any year) by one in a million (Table 1). Comparisons of the risks of similar or related substances and activities. Several researchers have adopted a narrower approach to risk comparisons, comparing risks of substances or activities that are similar or closely related (14-19). For example, Ameset al. (17) compared the risks of eating foods that contain synthetic chemicals (e.g., food additives and pesticide residues) with the risks of eating natural foods. The study concluded that human dietary intake of natural carcinogens in food is likely to be at least 10,ooO times greater than the intake of potentially carcinogenic synthetic chemicals in fwd (although partial protection against the effects of natural carcinogens is provided by the many natural anticarcinogens that also appear in food). Another common procedure is to compare exposures to the same risk agent. Wilson and Crouch (I), for example, compared the risks associated with different sources of exposure to radiation (Table 2). A differentapproach to risk comparisons is reflected in a study by Doll and Pet0 of the causes of cancer (19). The study found that the combined effect of food additives, occupational exposures to toxic agents, air and water pollution, and industrial products account for only
about 7% of U.S.cancer deaths. These results suggest that removing all pollutants and additives from the air, water, food, and work place would result in only a small decrease in cancer mortaity (although even this small percentage represents a substantial number of lives). By contrast, the combined effects of alcohol, diet, and smoking are related to 70% of U S . cancer deaths. Consequently, even a modest change in personal habits would result in a significant decrease in cancer mortality. Risk Comparisons and risk acceptability. Many of the risk comparisons described above have been advanced not only for gaining perspective and understanding but also for setting priorities and determining which risks are acceptable (1-4, 10-13). More specifically, risk comparisons have been advocated as a means for determining which risks to ignore, which risks to be concerned about, and how much risk reduction to seek (3, 13, 20). Thus, Wilson (2) has argued that we should “try to measure our risks quantitatively. . , . Then we could compare risks and decide which to accept or reject.” Lord Rothschild Ql), an advocate of the comparative risk approach, has observed that there “is no point in getting into a panic about the risks of life until you have compared the risks which worry you with those that don’t, but perhaps should.” Based on such arguments, Wilson Environ. Sci. Technol., Vol. 23,No. 12, 1989
1445
I
I
Compatison of m m o n radlsflon risks
ldenl near Cher
Not relevant
tlnental round trip n 20 mi of nuclear
(20) constructed a scale ranking risks from acceptable to unacceptable. Wilson argued that activities falling in the upper zone of the scale, representing risks of death per year of exposure of less than one in a million, can be regarded as acceptable. According to Wilson, these activities have insignificant risks-insignificance beiig defined as the level of risk that individuals routinely accept in their personal and daily activities. For example, because individuals routinely accept the risk of being shuck by lightning-which poses a risk of death of one in a million per year of exposure-risks of this sue can be regarded as acceptable. Following the same logic, Wilson argued that &vities representing risks of death that are greale; than one in a thousand per year of exwsure can be regarded as unacceptabe. According to-Wilson, activities falling in the middle zone of the scale are the most prohlematic: The acceptability of these risks cannot be d e tennined a priori. Instead, they must be and subjected to closely ~c~tinized analysis and debate. As will be discussed later in this paper, serious questions can be raised a b u t the legitimacy of studies that use risk comparisons for determining which risks are acceptable. A basic criticism is that such efforts fail to recognize the importance and legitimacy of basing decisions a b u t the acceptability of a risk on factors other than the size of the risk. These factors include voluntariness, fairness, benefits, alternative~,and control.
Limitations of risk comparisons Several important limitations of the risk comparison approach can be noted. These include failure to identify and emphasize uncertainties involved in the calculation of comparative risk estimates; failure to consider the broad set of quantitative dimensions that define and measure risk; and failure to consider the broad set of qualitative dimensions that underlie people's concerns
a b u t the acceptability of risks and technologies. Each is described below. Failure to identify and emphasize uncertainties involved in the calculation of comparative risk estimates. Despite their s m g t h s , even the best risk assessments cannot provide exact answers (22-26). Due to limitations in scientific understanding, data, models, and methods, virtually all risk assessments are characterized hy substantial uncertainties. The uncertainties inherent in risk assessment are especially evident in the assessment of chronic health effm due to low-level exposures to toxic chemicals (23-24). Parallel problems exist in engineering risk assessments designed to estimate the probability and severity of rare, high-consequence accidents in complex technological systems such as chemical production and chemical storage p h t . (25). Because of these uncertainties, policy considerations often enter into risk assessment through the selection of assumptions. The National Research Council, for example, identified over 50 critical points where assumptions are needed because of the current lack of scientific knowledge (26). At each point, the consequences of selecting one assumption over another are substantial (23). Given these uncertainties, a critical tlaw in many risk comparisons is the failure to provide information on the assumptions (conservative or liberal) underlying the calculation of comparative risk estimates. Because risk estimates typically are drawn from a variety of different data sources, tables of comparative risks may contain risk estimates based on conservative estimates together with noncomparahle estimates based on liberal assumptions. S i l y , tables of comparative risks may contain risk estimates based on actuarial statistics (e.g., deaths from motor vehicle accidents) together with estimates based on controversial models, assumptions, and judgments (cancer deaths from chronic exposure to pesti-
1446 Environ. Sci. Tschnoi., Vol. 23,No. 12,1989
cides or to chemical air pollutants). A related flaw in many risk comparisons is the failure to describe and characterize uncertainties. Risk comparison tables that report only single values for adverse health or environmental consequences ignore the range of possibilities and may provide an inaccurate picture to the public. At a mini", risk comparison tahles should include data describing the uncertainty in each risk estimate. Failure to consider the broad quantitative dimensions that de6ne and measure risk. Most lists of comparative risks are unidimensional. They present statistics for only one dimension of risk, such as expected annual mortality rates or reductions in life expectancy. The use of such narrow quantitative measures of risk of an activity or substance can obscure the importance of other significant quantitative dimensions, such as expected annual probability of injury or disab~ty,spatial extent, concentration, persistence, recurrence, population at risk, delay, maximum expected fatalities, transgenerational effects, expected environmental damage (e.g., ecological damage or adverse effects on endangered species), and maximum expected environmental damage (27). Significant distortions and misunderstandings also result frdm comparative analyses that fail to provide the full range of relevant quantitative risk information. Consider, for example, some of the problems involved in comparing the risks of airplane travel to the risks of automobile or train travel. Using a measure of risk to an individual based on the number of deaths per hundred million passenger miles, travel by airplane appears to pose slightly less risk to an individual (0.38 deaths per hundred million passenger miles) than does travel by automobile (0.55 deaths per hundred million passenger miles) and slightly more risk than travel by hain (0.23 deaths per hundred million passenger miles) (28). However, for airplane travel, the landing and take-off phases represent the period of highest risk; thus it can be argued that a better estimate of individual risk is the number of journeys made rather than the number of miles traveled. Using this measure, traveling as an airplane passenger poses a slightly greater risk (1.8 deaths per million passenger journeys) than traveling as an automobile passenger (0.027 deaths per million passenger journeys) or as a train passenger (0.59 deaths per million passenger journeys). As a result, if distance traveled is the selected measurement criterion, then airplane travel is marginally safer than automobile travel and marginally less safe than train tra-
vel; but if number of journeys is the criterion, then airplane travel is marginally less safe than both automobile travel and train travel. A related deficiency in most risk comparisons is the failure to estimate the total quantitative risk of the activities included in the risk comparison. Technological activities encompass a variety of different components; stages of development (e.g., extraction of raw materials, production, consumption, and disposal); and relationships (direct and indirect) with other technological and societal activities (14,291. Detailed examination of the risks of these different components, stages of development, and relationships may significantly alter the overall ranking of a technology or activity. Consequently, any risk comparison that claims to be comprehensive must either present risk data for each of these aspects or explicitly acknowledge those aspects that have been excluded. Even when data are provided on the total quantitative risk of an activity, the comparison can nonetheless be misleading if it fails to provide risk data for sensitive, susceptible, or high-risk groups. These include children, pregnant women, fetuses, the elderly, and individuals who are particularly vulnerable or susceptible because of illness or disease. Most lists of comparative risks present only population averages. Important distinctions can be masked in other ways. For example, it is seldom clear from risk comparison tables what is included in the specific-risk entries. For example, many risk comparison tables contain data on the risks of smoking but do not specify whether the data include deaths from cardiovascular disease as well as from lung cancers. Nor is it clear if the risk estimate is based on the entire population or only on the population that is exposed. Even if the analyst carefully and accurately reports risk data, misunderstandings can develop if important situational qualifiers are left out. For example, the risk calculation for driving includes many different driving situations. Yet speeding home after a party just before dawn is orders of magnitude more risky than other driving situations. Similarly, the risk of being hit by lightning for people who remain on a golf course during a thunderstorm is much higher than the average risk for the U.S.population. A related deficiency stems from the failure to recognize the importance of framing effects on risk comparisons (30-31). Different impressions are created by different presentation formats. Each format for presenting quantitative risk information, such as deaths per million people, deaths per unit of con-
Severity of
Consequences Probability of occurrence Catastrophic potential Reversibiiity
condltcm assoddled
C d i s assodated
Many fatalities or injuries per event (airplanecrashes) High probability (heart and lung diseases among heavy smokers) Fatalities or injuries grouped in time or space (large industrial explosions) ineversible consequences
Few fatalities or injuries per event (deaths from fails) Low probability (rare diseases)
(AIDS)
Delayed effects Impact on
Chronic effects that are deiaved . (cancer) . Risks borne equally or more by future futuregenerations (ozone aenerations deoletionl impact ~1 Children specifically at risk children (birth defects) Victim identity idantlfiable victims (sailor lost at sea) Familiarity Unfamiliarrisks (ozone depletion) Understanding Lack of personal understanding of mechanisms or processes involved (nuclear power piant accidents) Scientific Risks aDDaar to be unclear or uncertainty uncertdin to scientists (disagreementsamong scientists about tne risks of nuclear oowerl Dread Risks evoke fear (toxic chemicals in an aDandoned hazardous waste sitei Wtuntariness Exposures are involuntarty (air pollution) Controllability Little personal control over risk (traveling as a passenger in a car or airplane) Benefits from or need for clarity of benefits activity generatlng risk are questioned (nuclear power) Equity Those at risk from the activitv do not directly gain its benefits (people living near an abandoned hazardous waste site) Institutional Lack of trust in institutions tNSt responsible for risk management (regulatory agencies with perceived close ties to industry) Psrsonai Individual personally at risk stake (living near an abandoned hazardous waste site) Risk caused by human failure Atlnbution of blame (explosionat an industrial plant caused by negligence) Media Much media coverage (airline attention crashes; industrial accidents) DExamplesIn $arentheSsa
centratiou, or deaths per activity, is likely to have a different impact on the audience. A final deficiency is the failure in most risk comparisons to acknowledge deficiencies in the quality of the data. Most risk comparisons draw on diverse data sources that vary considerably in quality. Because of the high cost and difficultyof collecting original data, re-
Fatalities distributed randomly in time and space (automobile accident deaths) Consequences appear reversible (gonorrhea) All effects immediately realized iburnsl
hsks borne primarily by current generation (sunbathing) Risks tnreaten adults only (occupationalIISKS) Statistical victims (highway fatality estimates) Familiar nsks (household accidents) Mechanisms or processes involved are personally understood (fires: slipping on ice) Risks appear to be relatively well known to scientists (auto accidents) Risks not dreaded (food poisoning) Risks are taken at one's own choice (skiing) Some personal control over risk (driving an automobile)
Clear benefits (travelingby oar) Distribution of risks and benefits appears to be equitable (vaccinations) Responsible institutions wellt r d s f e d (management of recombinant DNA research bv the National Institutesof Health and universities) Individual not personally at risk (disposal of hazardous waste at a remote site) Risk caused by nature (lightning) Little media attention (on.thP job injuries)
searchers seldom have access to data developed exclusively for the comparison. Instead, a variety of existing data sources are used, each varying in quality and completeness. As a result, risk comparison tables often contain highquality data along with data of questionable scientific validity. Failure to consider the broad qualitative dimensions that underlie peoEnviron. Sci. Technol., Vol. 23,No. 12. 1989 1447
d e l b e s for improving the effectiveness of risk compariso arison to a specific audience's lev81of knowledge. 2. Be specific about the intent of the comoarison an
unwarranted conclusions. y acknowledge and explain es in the calculation of risk estimates. risk estimates for the worst case. best ely case. oid comparisons that ignore factors that infiuen ns of risk and acceptability, such as voluntarin its, alternatives, and control. knowledge limitations in comparisons that i fluencepublic perceptions of risk and acceptabilitv. 7,FoCus the comparison on classes of substances, products, proc.esses, or activities that are similar or related In their characteris. ..tics. such as activities that serve the same function and whose nefits tend to be similar. rmulate the comparison nmental consequences, inc
ts.
rovide information on the socl
ple's concerns a b u t the acceptability of risks and technologies. A common argument in many risk comparisons is that risks that are small or comparable to already accepted risks should themselves be accepted. Such claim cannot, however, be defended (32). Althongh carefully prepared lists of comparative risk statistics can provide insight and perspective, they provide only a small part of the information needed for setting priorities or for determining which risks are acceptable. Judgments of acceptability are related not only to annual mortality rates-the focus of most risk comparisons-but also to a multiplicity of qualitative dimensions or factors (32-36). As -s ' in Table 3, a broad and diverse set of factors influences public perceptions of risk and acceptability. These factors-catastrophic potential, familiarity, and so forth-explain, in part, public concerns about the risks of routine and accidental releases of toxic chemicals. Because of the importance of these factors, comparisons showing that the risks associated with chemical releases are lower than the risks of other activities or technologies may have no effect whatsoever on pnblic perceptions and attitudes. For example, comparing the risk of living near a chemical manufacturing plant with the risk of driving X number of hours, eating X number of tablespoons of peanut butter, smoking X number of cigarettes a day, or sunbathing X number of hours may provide perspective but may also be highly inappropriate. Because such risks differ on a variety of qualitative dimensions-perceived benefits, extent of personal control, voluntariness, cata1448 Environ. Sci. Technoi., Voi. 23, NO. 12, 1989
strophic potential, familiarity, fairness, origin, and scientific uncertainty-it is likely that people will perceive the comparison as meaningless (7). The fundamental argument against such comparisons is that it is seldom relevant or appropriate to compare risks with differentqualities for risk acceptability purposes, even if the comparison is technically accurate. Several reasons underlie this argument. First, as noted abve, there are important psychological and social differences among risks with different qualities. Risks that are involuntary and result from lifestyle choices, for example, are more liely to be accepted than risks that are perceived to be involuntary and imposed. Second, people recognize that risks are cumulative and that each additional risk adds to their overall risk burden. The fact that a person is exposed to risks resulting from voluntary lifestyle choices does not lessen the impact of risks that m perceived to be involuntary and imposed. Third,people perceive many types of risk in an absolute sense. An involuntary exposure that increases the risk of cancer or birth defects is perceived as a physical and moral insult regadless of whether the increase is small or whether the increase is smaller than risks from other exposures. Finally, judgments about the acceptability of a risk can seldom be separated from judgments about the risk decision process (37). Public responses to risk are shaped both by the characteristics of the risky activity and by the perceived adequacy of the decision-making process. Risk comparisons play
only a limited role in such determinations of acceptability. Other limitations. Other limitations can substantially diminish the usefulness of a risk comparison, especially when the purpose of the comparison is to set priorities or to determine which risks are acceptable. These include: failing to consider legal constraints, such as which options can legally be considered in reducing or mitigating risks; failing to consider the full range, costs, risks, and benefits of feasible and available technological alternatives; and failing to consider opportunities for risk reduction, risk redistribution, or compensation. Another limitation of risk comparisons is that they are not adapted to the needs and concerns of people according to their education, income, employment status, occupation, gender, age, marital status, geographic region, number of children, voluntary group memberships, ethnicity, political preferences, environmental beliefs, sociotechnological world views, and other relevant socicdemographic characteristics (38-45).
Improving risk comparisons Despite the limitations reviewed above, risk comparisons that are wellconstructed and welldocumented can help put risks in perspective and communicate risk information (7). They can provide a bench mark against which the magnitude of new or unfamiliar risks can be compared. They also can help inform people about the range and magnitude of risks to which they are exposed. For risk comparisons to achieve these goals, however, the l i t a t i o n s of the approach must be addressed specifically. The guidelines in the box may be helpful for this purpose. Ideally, different types of comparisons should be presented (7). Preference should be given to comparisons that people find helpful and that do not preempt judgments about the acceptability of a risk. Alternative types of comparison that might be considered include comparisons of similar risks, comparisons of risks with benefits, comparisons of alternatives, comparisons of the same risk at different times, comparisons to natural background levels, and comparisons with a regulatory standard.
Research needs and eonelusions In response to the Emergency Planning and Community Right-to-Know Act and related laws, risk communicators are seeking better ways to explain information about chemical risks accurately, responsibly, and ethically. Risk comparisons, although only a part of the answer, have an important role to
play in this process of explanation. Although risk comparisons have significant limitations, these limitations are balanced by important strengths. Foremost among these are the compatibility of risk comparisons with natural thought processes and the ability of comparisons to help people understand and appreciate new or unfamiliar risks. Given these strengths, a critical need exists to better understand factors that enhance or diminish the effectiveness of different risk comparisons. On the one hand, theoretical and Conceptual research is needed to develop improved classification schemes, scales, and Of risk Ihat and sure the multidimensional character of risk (27.45). On the other hand, empirical research is needed that addresses a diverse set of questions, including: What methods are most effective for placing risks incomparative penpective? What dimensions of risk need to be taken into account in structuring and presenting risk comparisons? How can uncertainties in risk estimates be most effectively presented in a risk comparison? How should important assumptions and cav a t s for risk comparisons be communicated? What comparative statistics (mortality rates, years of life expectancy lost) are most meaningful IO people? To what extent should risks be categorized for risk comparisons? What factors need to be considered in choosing refprence risks for a risk comparison? How can basic risk assessment term-such as "parts per billion"-be more effectively presented and explained through comparisons? In what context can risk comparisons be used warning labels? What means of display are most effective for risk comparisons? Is it possible to develop the Richter COnCeptual scale for scales use inanalogous risk comparto isons? To what extent should characteristics of the different nudiences for risk information influence the design and selection of a risk comoarison? Which individuals or groups ire perceived to be most credible and trusnvorthy as sourcesof information on risk comparisons? What evnluation methods are available for determining the effectiveness of risk comparisons? Research focused on these questions can help make risk comparisons more useful. However. risk comparisons cannot replace action when action is warranted. Nor can they take the place of credibility; trust; and long-term, sensitive interaction with the local community. Risk comparisons are helpful only when presented in the context of a continuing, sound, community relations program. In summary, the simplicity and intui-
tive appeal of the risk comparison a p proach is often deceptive. Many factors play a role in determining the legitimacy and effectiveness of risk comparisons. The success of the comparison will depend on the degree to which these factors have been adequately recognized. considered, and addressed
(22) ~owrancc.w.
Acknowledgment 1 am indebted to the following individuals for their helpful commentSon earlier drafts of this paper: Michael B ~ Timothy ~ ~ Earle, Rachelle Hollander, Roger Kasperson, Lester Lave. Brian Lehrer, Granger Morgan, lerYl Mumpower- Peter Sandman. John Slavick. Paul Slovic, and Detlof von Winterfeldt. Preparation of this paper was supported in part by a grant from the Chemical Manufacturers Associalion to the Center for Risk Communication at Columbia University.
References ( I ) Wilson. R.:
Crouch. E.A.C. Science 1987,236.267-70. (2) w,lron, R. Tpchno/. R ~ 1979, ~ , 81, 4046. (3) Cohen. B. L.: Lee. I. Healrh Phys. 1979. 36. 707-22. (4) Crouch. E.; wilson, R, InASleSSme(lr rind ,&fanaRPmcnrof chemica) Risks; Rodricks. I . V.; Tardiff, R. G.. Edr.: ~~~~i~~~ Chemical sKiety: washing. ton. DC. 1984. ( 5 ) Sandman, p. fipplaining ~nvi,onmento~ Risk. U.S. Environmental Protection Agency. Office of Toxic Substances. U S . ~~~~~~~~~t printing office: washing. ton. DC. 1986. (6) Covello. v. T.: winterfeldt, D.; Slovic. F! In Risk Communicoiion: Davie6, J. c.: cove~io, v. T.: A I I ~F.~ w., . Eds.; The Conservation Foundation: Washington, DC. 1986. (7) Covello. V. T.: Sandman, p; slovic, p. Rirk Communication. Risk Stmisrics. and ~ i . $ kco,npo,iznns: Chemical ~ ~
of
w. Acceptable Risk: Science and the Deierminarion of Safety:
Kaufman: Lor Altos. CA, 1976. (23) Ri,~kA s s e ~ ~ m ~ " l Conr,ol: conservation Foundation: Washington, DC, 1985. (24) UnfinishedBusincrs:A sessment of Environmrntol Problems. W. 1. Overview Report. US. Environmental Protection Agency. Office of Policy Analysis. US. Government Printing Office: Washington, DC, February, 1987. (25) Covetlo. v.: Merkhofer, M. In hsuring and MOnOging Horordous Risks: From Sewso IO Bhopol and Beyond: Kleindor-
~fer. F!:, Kunrcuther, H.. Eds.: SpringerVerlag: Berlin. 1986. (26) National Research Council. Risk Analysis in rhc Federal Cowmmem: MmORinR the P m r m : National Academy Press: Washington. DC. 1983. (27) Hohenemser, c . ; K ~ I ~ R. S . w.: Slovic, F! Sricnct 1983.220. 378-84. (28) D, In Aecep,obiliry of Risk, Council on Science and Society, Ed.: Barry Rose. Lid.: London. 1977. (29) Lave. L. B. Science 1987,236.291-95. (30) Kahneman D.; Tversky. A. Am. Psychol. 1984,39(4). 341-50. (31) Kahneman. D.; Slovic. I?;Tversky A,. Eds.: Judgment Under Uncrnointy: Heuristics and Biases: Cambridge University
Press: New York, 1982. (32) Slovic, F! Sciencc 1987.236. 280-85. (33) Starr. c. Science 1969. 165. 1232-38. (34) Covello. V. T. Techtiological Forecming and Sociol Chongr 1983.23.285-97, (35) FirchhoR. B. I r s u ~ sin Science end Ttchnology 1985,2(1). 83-96. (36) Covello. V.: Menker. 1.: Mumpower 1.. Eds.: Risk Evoluorion and hfonogemenl: Plenum: New York. 1986. (37) Kasperson. R. Risk Analysis 1986, 6. 275-82. (38) Baram. M. ColumbiaJournolofEnvimnmmralLow1987, 13(1). 1-51. (39) Douglas. M.: Wildavsky. A. Risk a d Culrurr. University of California Press: Berkeley, 1982. (40)Otway. H. 1.: Yon Winterfeldt. D. Policy Scicnccs 1982.8. 127-52. (41) ~Kleindorfer, ~ E: Kunreulher, f ~ H.,Eds.; ~ In-
swing and Managing Hmordous Risks: turers Association: Washington. DC. From Sewso IO Bhopal and Beyond; 1988. Springer-Verlag:Berlin. 1986. (8) Covello. V.: MeCallum. D.: Pavlova. M., ~ ~ p eRisk r i ~~ ~ ~ ~ ~ (42) ~ Lave, ~ L.iRisk Analysis ~ ~ 1984, l 4.i 79-81. ~ ~ SOrite Role and Rcspon,vibi/iry of Govern. (43) Johnson. B.; COVCllO, V.. Ed%: merit and Non*.ove"nral Omgmiror i d ond Culturol Construerim of Risk: Essays on Risk Selection and Perception: tiom: Plenum: New York. 1989. (9) Firchhoff. B. Ct al. Acceprable Risk: Reidel: Boston, 1987. cambridgeuniversitypress: N~~ york. (44) Von Winterfeldl. D.: Edwards, W. Risk 1981. Analysis 1984,4(1). 55-68. c. (10) Snwby. F. D. Health Phys. 1965. 11. (45) FirchhoffB.: Watson. s.; icy Sciences 1984. 17. 123-39. 879-87. (11) Pochin, E. E. Community Heolrh 1974. 7.8,~ . _ . -~-,. _
A,,,
Pochin, E. E.BE Med. Bull. 1975,31(3). 184-90. Crouch, E.: Wilson, R. RisklBcnejir Annlwir: Ballinger: Cambridge, MA. 1982. Inhaber. H. Science 1979,203.718-23. Cohen. A. V.: Pritehard. D. K. Comporariw Risks of Elecrriciry Production Systcmr: A Criticol Survey oJrhc Litermure: U.K. Health and Safety~Executive. Research Report No. I I : Her MajertyS Stationerv Office: London. 1980. Eisen6ud. M. Envimnmmr 1984,26(10). 6-11.30-33. Amer. B. N.: Magaw, R.; Gold, L. S. Science 1987.236. 267-70. Tierney. J . Hippocrates 1988, January1
Viicent I Covello is a professor in the School of Public Health, Columbia Universiiy, New ibrk. H e is also director of Co lumbin UniversityS Center for Risk ComEnhrllrr" ' 0 . 2 P ..~.. ""..,.*, Doll. R : Pel". R Journalofth~Notion~l munication. Prior to his currenl positions. Covello was director of the Risk AssessCcmrrrlnriirutr 198I.66 1191-1308. ment Progmm at the National Science Wilu,n. R . Snennrr. Tpclmoloer. ondHu. Foundation nndpresident of the Socieiy for mati voluts 1984, %2). I 1-22: Rothschild. N. WaIlSt. 3.. May 13. 1979. Risk Annlysis.
.....
Envlron.Sd.Technol.,Vol. 23, No. 12, 1989 1448
-
:
