Engineering Ethics

risk.zip

Home >English homework help >Engineering Ethics

advancing+risk+assessment+brief.pdf

Science and Decisions Advancing Risk Assessment

Risk assessments evaluate potential adverse health effects posed by harmful chemicals found in the environment and inform a range of decisions from protecting air and water to ensuring food, drug, and consumer product safety. Unfortunately, the risk assessment process is bogged down by challenges to its timeliness and credibility, a lack of adequate resources, and discon- nects between the available scientific data and the information needs of decision-makers. This report recommends significant changes to advance the use of risk assessments, including greater attention to planning and problem formulation, improved stakeholder involvement, and a better match of the level of detail needed in a risk assessment to the questions that need to be addressed.

Virtually every aspect of life involves risk. How we deal with risk depends largely on how well we understand

it. The process of risk assessment is used by the U.S. Environmental Protection Agency (EPA), other federal and state agencies, industry, and others to evaluate potential health risks posed by harmful chemicals found in the environment. Risk assessments inform a wide range of regulatory and technology decisions from protecting air and water to ensuring the safety of food, drugs, and consumer products such as toys.

EPA uses the risk assessment process that was set forth in the 1983 National Research Council report known as “the Red Book” (see Box 1). Since that report’s publication, EPA has greatly advanced risk assessment through the establishment of guidelines and of intra- and cross-agency science-policy panels, and improvements in peer- review standards. Despite this progress, major risk assessments for some chemicals are taking more than 10 years. In the case of trichloroethylene, which has been linked to cancer, the assessment has been under development since the 1980’s.

There are several reasons the risk assessment process is bogged down. The credibility of risk assessment is often challenged because of the impacts of regulation, both nationally and internationally. When state and federal lawmakers move forward with risk management decisions in the absence of completed risk assessments, the value and credibility of risk assessments are further threatened. EPA is struggling to meet demands

for hazard and dose-response information but is challenged by a lack of resources, including funding and trained staff. Uncertainty, an inherent property of scientific information, continues to lead to multiple interpretations and contribute to decision-making gridlock.

Box 1. The Risk Assessment Process Risk assessment describes what research

findings do and do not tell us about threats to human health and to the environment. There are four steps in the process—hazard identification, exposure assessment, dose-response assess- ment, and risk characterization—which were defined in the 1983 National Research Council report Risk Assessment in the Federal Govern- ment: Managing the Process, known as the “Red Book.” After a risk assessment is com- plete, decision makers use it to determine how to reduce exposure to toxic substances.

In addition, the rapid development of large quantities of scientific data, stemming from advancements in fields such as genomics and biomarkers, are increasing the complexity of risk assessments and the decisions that these assessments support. These data have led to questions about how to address, for example, multiple chemical exposures, multiple risks, and susceptibility in sensitive populations. In addition, risk assessment is now being extended to broader environmental questions, such as the “life-cycle analysis” of chemicals from their manufacture through their many uses, and also to issues of costs, benefits, and risk-risk tradeoffs.

In light of these challenges, EPA asked the National Research Council to conduct an independent study on improvements that could be made in the short term (2-5 years) and in the longer term (10-20 years). The report concludes that EPA’s overall concept of risk assessment, which is generally based on the National Research Council’s 1983 “Red Book” should be retained. However, a number of significant changes are needed to make the process more useful to decision- making.

MAJOR RECOMMENDATIONS The report offers the following conclusions

and recommendations to enhance the credibility and usefulness of risk assessment.

Improving the Design of Risk Assessment This report defines “design” as the process of

planning a risk assessment and ensuring that its level and complexity are consistent with the needs to inform decision-making. Good design involves bringing risk managers, risk assessors, and various stakeholders together early in the process to determine the major factors to be considered, the decision-making context, and the timeline and depth needed to ensure that the right questions are being asked in the context of the assessment.

Increased emphasis on planning and scoping and on problem formulation has been shown to lead to risk assessments that are often more useful and better accepted by decision-makers. However, EPA’s incorporation of these stages in risk assessment has been inconsistent to date. The report recommends EPA focus greater attention on design in the formative stages of risk assessment, specifically on planning and scoping and problem formulation as articulated in EPA guidance for ecologic and cumulative risk assessment. An important element of planning and scoping is defining of a clear set of options for consideration in decision-making.

To that end, the report proposes that EPA adopt an expanded risk assessment framework that has the same core as the Red Book model but differs in its preliminary and final steps. The framework begins with a “signal” of potential harm, for example, a suspicious disease cluster, or findings of industrial contamination. Under the traditional paradigm, the question has been, “What are the probability and consequence of an adverse health (or ecologic) effect posed by the signal?” In contrast, the recommended framework asks, implicitly, “What options are there to reduce the hazards or exposures that have been identified, and how can risk assessment be used to evaluate the merits of the various options?” The focus therefore shifts to the risk decisions to be made, more clearly laying out the information needed from the risk assessment.

Uncertainty and Variability Addressing uncertainty and variability is critical

for the risk-assessment process. Uncertainty stems from lack of knowledge; it can be characterized and reduced by the use of more or better data but not eliminated. Variability is an inherent characteristic of a population, inasmuch as people vary substantially in their exposures and their susceptibility to potentially harmful effects of the exposures. Variability cannot be reduced, but it can be better characterized with better information.

Just as a risk assessment itself should be more closely tied to the questions to be answered, the level of detail for characterizing uncertainty is appropriate only to the extent that it is needed to inform specific risk-management decisions appropriately. The required extent and nature of uncertainty analysis should be decided in the planning and scoping phases of a risk assessment. EPA does not have a consistent approach to determine the level of sophistication or the extent

Emerging scientific advances hold great promise for improving risk assessment. For example, new toxicity-testing methods are being developed that will probably be quicker, less expensive, and more directly relevant to human exposures, as described in the National Research Council’s Toxicity Testing in the 21st Century: A Vision and a Strategy (2007). However, the real- ization of the promise is at least a decade away.

of uncertainty analysis needed to address a particular problem. Inconsistency in the treatment of uncertainty among components of a risk assessment can make the communication of overall uncertainty difficult and sometimes misleading.

The report recommends that in the short term, EPA should adopt a “tiered” approach for selecting the level of detail to be used in uncertainty and variability assessments, which should be made explicit in the planning stage. EPA should develop guidance to determine the appropriate level of detail needed to support decision-making.

A Unified Approach to Dose-Response Assessment

Historically, dose-response assessments at EPA have been conducted differently for cancer and noncancer effects. For cancer, it has generally been assumed that there is no dose threshold of effect—that is, that the smallest exposure has some health effect. For noncancer effects such as asthma or birth defects, risk assessments try to determine a dose threshold—a reference dose--below which effects are not expected to occur or are extremely unlikely in an exposed population. Consequently, noncancer effects have been underemphasized, especially in benefit-cost analyses.

This report recommends a significant departure from current practices to unify the approach to cancer and noncancer effects, which the report concludes is scientifically feasible and should be implemented. The approach for dose-response modeling should include formal, systematic assessment of background disease processes and exposures, possible vulnerable populations, and modes of action that may affect a chemical’s dose-response relationship in humans.

In this approach, the reference dose would be redefined as a risk-specific dose that provides

information on the percentage of the population that can be expected to be above or below a defined acceptable risk with a specific degree of confidence. The risk-specific dose will allow risk managers to weigh alternative risk options with respect to that percentage of the population and determine a quantitative estimate of benefits for different risk-management options. For example, a risk manager could consider various population risks associated with exposures resulting from different control strategies for a pollution source and the benefits associated with each strategy. The report acknowledges, however, the widespread application and public-health utility of the reference dose; the redefined reference dose can still be used as it has been to aid risk-management decisions.

Selection and Use of Defaults Much of the scientific controversy and delay in

completion of some risk assessments has stemmed from the long debates regarding the adequacy of the data to support the use of a default—an assumption made when chemical-specific data are not available—or an alternative approach that is used in place of a default. The 1983 Red Book recommended the development of guidelines to justify and select from among the available defaults to ensure consistency and to avoid manipulations in the risk-assessment process.

The report concludes that established defaults need to be maintained for the steps in risk assessment when chemical-specific data are not available. EPA, for the most part, has not yet published clear, general guidance on what level of evidence is needed to justify use of agent-specific data and not resort to a default. There are also a number of defaults that are implicitly engrained in EPA risk-assessment practice but are absent from its risk-assessment guidelines. For example, chemicals that have not been examined sufficiently in epidemiologic or toxicologic studies are often insufficiently considered or even excluded from risk assessments. This is a problem at Superfund and other risk assessment sites; a relatively short list of chemicals for which there are epidemiologic and toxicologic data tends to drive exposure and risk assessments.

EPA should continue and expand use of the best, most current science to support and revise default assumptions. EPA should work toward the development of explicitly stated defaults to take the place of implicit defaults. EPA should develop clear, general standards for the level of evidence needed to justify the use of alternative assumptions in place of defaults. When EPA elects to depart from a default assumption, it should quantify the implications of using an alternative

Committee on Improving Risk Analysis Approaches Used by the U.S. Environmental Protection Agency: Thomas A. Burke (Chair), Johns Hopkins Bloomberg School of Public Health; A. John Bailer, Miami University; John M. Balbus, Environmental Defense; Joshua T. Cohen, Tufts Medical Center; Adam M. Finkel, University of Medicine and Dentistry of New Jersey; Gary Ginsberg, Connecticut Department of Public Health; Bruce K. Hope, Oregon Department of Environmental Quality; Jonathan I. Levy, Harvard School of Public Health; Thomas E. McKone, University of California, Berkeley, CA; Gregory M. Paoli, Risk Sciences International, Ottawa, ON, Canada; Charles Poole, University of North Carolina School of Public Health; Joseph V. Rodricks, ENVIRON International Corporation; Bailus Walker Jr., Howard University Medical Center; Terry F. Yosie, World Environment Center; Lauren Zeise, California Environmental Protection Agency; Eileen N. Abt, (Study Director)

National Research Council.

The National Academies appointed this committee of experts, who volunteered their time to address this specific task and to produce this report. The report is peer-reviewed and the final product signed off by both the committee members and the National Academies.

This report brief was prepared by the National Research Council based on the committee’s report. For more information, contact the Board on Environmental Studies and Toxicology at (202) 334-3060 or

visit http://nationalacademies.org/best. Copies of Science and Decisions: Advancing Risk Assessment are available from the National Academies Press, 500 Fifth Street, NW, Washington, D.C. 20001; (800) 624-6242; www.nap.edu.

Permission granted to reproduce this brief in its entirety with no additions or alterations.

assumption, including how use of the default and the selected alternative influences the risk estimate for risk management options under consideration.

Cumulative Risk Assessment There is a need for cumulative risk assessments

as defined by EPA—assessments that include combined risks posed by aggregate exposure to multiple agents or stressors; aggregate exposure includes all routes, pathways, and sources of exposure to a given agent or stressor. Although EPA has used cumulative risk assessment in various contexts, the process should be expanded to include consideration of nonchemical stressors (for example, smoking, diet, and alcohol consumption), vulnerability, and background risk factors.

Because of the complexity of considering so many factors simultaneously, there is a need for simplified risk-assessment tools such as databases, software packages, and other modeling resources, that would allow screening-level risk assessments and could allow communities and stakeholders to conduct assessments and thus increase stakeholder participation. Cumulative human health risk assessment should draw greater insights from ecologic risk assessment and social epidemiology, which have had to grapple with similar issues. Cumulative risk assessment is addressed in the National Research Council’s Phthalates and Cumulative Risk Assessment: The Tasks Ahead (December 2008).

Stakeholder Involvement Greater stakeholder involvement is necessary

to ensure that the risk assessment process is transparent and that decision-making proceeds effectively, efficiently, and credibly. Although EPA has made great progress in creating programs and guidance documents related to stakeholder involvement, it is important that it adhere to its own guidance. EPA should establish a formal process for stakeholder involvement in the framework for risk- based decision-making. The process should include time limits to ensure that decision-making schedules are met and with incentives to allow for balanced participation of stakeholders, including impacted communities and less advantaged stakeholders.

Capacity-Building EPA’s current structure and insufficient

resources may pose a challenge to implementing the recommendations in this report, which are tantamount to “change-the-culture” transformations in risk assessment and decision-making in the agency. Moving forward will require a commitment to leadership, cross-program coordination and communication, and training to ensure the requisite expertise. EPA should initiate a senior-level strategic examination of its risk-related structures and processes to make sure it has the institutional capacity to carry out these recommendations and develop a capacity building plan that includes budget estimates.

__MACOSX/._advancing+risk+assessment+brief.pdf

advancing+risk+assessment+full+report.pdf

Science and Decisions: Advancing Risk Assessment (Free Executive Summary) http://www.nap.edu/catalog/12209.html

Free Executive Summary

ISBN: 978-0-309-12046-3, 478 pages, 6 x 9, paperback (2008)

This executive summary plus thousands more available at www.nap.edu.

Science and Decisions: Advancing Risk Assessment

Committee on Improving Risk Analysis Approaches Used by the U.S. EPA, National Research Council

This free executive summary is provided by the National Academies as part of our mission to educate the world on issues of science, engineering, and health. If you are interested in reading the full book, please visit us online at http://www.nap.edu/catalog/12209.html . You may browse and search the full, authoritative version for free; you may also purchase a print or electronic version of the book. If you have questions or just want more information about the books published by the National Academies Press, please contact our customer service department toll-free at 888-624-8373.

Risk assessment has become a dominant public policy tool for making choices, based on limited resources, to protect public health and the environment. It has been instrumental to the mission of the U.S. Environmental Protection Agency (EPA) as well as other federal agencies in evaluating public health concerns, informing regulatory and technological decisions, prioritizing research needs and funding, and in developing approaches for cost-benefit analysis.� ��However, risk assessment is at a crossroads. Despite advances in the field, risk assessment faces a number of significant challenges including lengthy delays in making complex decisions; lack of data leading to significant uncertainty in risk assessments; and many chemicals in the marketplace that have not been evaluated and emerging agents requiring assessment. ��Science and Decisions makes practical scientific and technical recommendations to address these challenges. This book is a complement to the widely used 1983 National Academies book, Risk Assessment in he Federal Government (also known as the Red Book). The earlier book established a framework for the concepts and conduct of risk assessment that has been adopted by numerous expert committees, regulatory agencies, and public health institutions. The new book embeds these concepts within a broader framework for risk-based decision-making. Together, these are essential references for those working in the regulatory and public health fields. ��

Copyright © National Academy of Sciences. All rights reserved. Unless otherwise indicated, all materials in this PDF file are copyrighted by the National Academy of Sciences. Distribution or copying is strictly prohibited without permission of the National Academies Press http://www.nap.edu/permissions/ Permission is granted for this material to be posted on a secure password-protected Web site. The content may not be posted on a public Web site. �

http://www.nap.edu/catalog/12209.html

http://www.nap.edu/permissions/

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Science and Decisions: Advancing Risk Assessment

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Prepublication Copy 3

Summary

Virtually every aspect of life involves risk. How we deal with risk depends largely on how well

we understand it. The process of risk assessment has been used to help us understand and address a wide variety of hazards and has been instrumental to the U.S. Environmental Protection Agency (EPA), other federal and state agencies, industry, the academic community, and others in evaluating public-health and environmental concerns. From protecting air and water to ensuring the safety of food, drugs, and consumer products such as toys, risk assessment is an important public-policy tool for informing regulatory and technologic decisions, setting priorities among research needs, and developing approaches for considering the costs and benefits of regulatory policies.

Risk assessment, however, is at a crossroads, and its credibility is being challenged (Silbergeld 1993; Montague 2004; Michaels 2008).1 Because it provides a primary scientific rationale for informing regulations that will have national and global impact, risk assessment is subject to considerable scientific, political, and public scrutiny. The science of risk assessment is increasingly complex; improved analytic techniques have produced more data that lead to questions about how to address issues of, for example, multiple chemical exposures, multiple risks, and susceptibility in populations. In addition, risk assessment is now being extended to address broader environmental questions, such as life-cycle analysis and issues of costs, benefits, and risk-risk tradeoffs.

The regulatory risk assessment process is bogged down; major risk assessments for some chemicals take more than 10 years. In the case of trichloroethylene, which has been linked to cancer, the assessment has been under development since the 1980s, has undergone multiple independent reviews, and is not expected to be final until 2010. Assessments of formaldehyde and dioxin have had similar timelines. EPA is struggling to keep up with demands for hazard and dose-response information but is challenged by a lack of resources, including funding and trained staff.

Decision-making based on risk assessment is also bogged down. Uncertainty, an inherent property of scientific data, continues to lead to multiple interpretations and contribute to decision-making gridlock. Stakeholders—including community groups, environmental organizations, industry, and consumers—are often disengaged from the risk-assessment process at a time when risk assessment is increasingly intertwined with societal concerns. Disconnects between the available scientific data and the information needs of decision-makers hinder the use of risk assessment as a decision-making tool.

1Silbergeld, E.K. 1993. Risk assessment: The perspective and experience of U.S. environmentalists. Environ.

Health Perspect. 101(2):100-104; Montague, P. 2004. Reducing the harms associated with risk assessment. Environ. Impact Assess. Rev. 24:733-748; Michaels, D. 2008. Doubt Is Their Product: How Industry’s Assault on Science Threatens Your Health. New York: Oxford University Press.

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Science and Decisions: Advancing Risk Assessment

4 Prepublication Copy

the 21st Century: A Vision and a Strategy (2007). However, the realization of the promise is at least a decade away.

To address current challenges, EPA asked the National Research Council to perform an independent study on improving risk-analysis approaches, one of a number of studies by the National Research Council that have examined risk assessment in EPA. Specifically, the committee selected by the National Research Council was charged to identify practical improvements that EPA could make in the near term (2-5 years) and in the longer term (10-20 years). The committee focused primarily on human health risk assessment but also considered the implications of its conclusions and recommendations for ecologic risk assessment. The committee conducted its data gathering for this study between fall 2006 and winter 2008, so materials published after this were not considered in the committee’s evaluation.

COMMITTEE’S EVALUATION

The committee focused on two broad elements in its evaluation: (1) improving the technical

analysis that supports risk assessment (addressed in Chapters 4-7) and (2) improving the utility of risk assessment (addressed in Chapters 3 and 8). Improving technical analysis entails the development and use of scientific knowledge and information to promote more accurate characterizations of risk. Improving utility entails making risk assessment more relevant to and useful for risk-management decisions.

Regarding improvement in technical analysis, the committee considered such issues as how to improve uncertainty and variability analysis and dose-response assessment to ensure the best use of scientific data, and it concluded that technical improvements are necessary. The committee concluded that EPA’s overall concept of risk assessment, which is generally based on the National Research Council’s Risk Assessment in the Federal Government: Managing the Process (NRC 1983,2 also known as the Red Book), should be retained. The four steps of risk assessment (hazard identification, dose-response assessment, exposure assessment, and risk characterization) have been adopted by numerous expert committees, regulatory agencies, public-health institutions, and others.

With respect to improving utility, the committee considered such issues as how risk-related problems are identified and formulated before the development of risk assessments and how a broad set of options might be considered to ensure that risk assessments are most relevant to the problems.

CONCLUSIONS AND RECOMMENDATIONS

A number of improvements are needed to streamline EPA’s risk-assessment process to ensure

that risk assessments make better use of appropriate available science and are more relevant to decision- making. Implementing improvements will require building on EPA’s current practices and developing a long-term strategy that includes greater coordination and communication within the agency, training and building a workforce with the requisite expertise, and a commitment by EPA, the executive branch, and Congress to implement the framework for risk-based decision-making recommended in this report and to fund the needed improvements.

The committee recommends an important extension of the Red Book model to meet today’s challenges better—that risk assessment should be viewed as a method for evaluating the relative merits of various options for managing risk rather than as an end in itself. Risk assessment should continue to capture and accurately describe what various research findings do and do not tell us about threats to human health and to the environment, but only after the risk-management questions that risk assessment should address have been clearly posed, through careful evaluation of the options available to manage the environmental problems at hand, similar to what is done in ecologic risk assessment. That alteration in the current approach to risk assessment has the potential to increase its influence on decisions because it

2NRC (National Research Council). 1983. Risk Assessment in the Federal Government: Managing the Process. Washington DC: National Academy Press.

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Summary

Prepublication Copy 5

requires greater up-front planning to ensure that it is relevant to the specific problems being addressed and that it will cast light on a wider range of decision options than has traditionally been the case.

A second recommended shift in thinking is seen in the technical recommendations in this report that call for improvements in uncertainty and variability analysis and for a unified approach to dose- response assessment that will result in risk estimates for both cancer and noncancer end points. Just as a risk assessment itself should be more closely tied to the questions to be answered, so should the technical analyses supporting it. For example, descriptions of the uncertainty and variability inherent in all risk assessments may be complex or relatively simple; the level of detail in the descriptions should align with what is needed to inform risk-management decisions. Similarly, the results of a dose-response assessment should be relevant to the problem being addressed, whether it is informing risk-risk tradeoffs or a cost- benefit analysis. Ensuring that the technical analyses supporting a risk assessment are both supported by the science and relevant to the problem being addressed will go a long way to improving the value, timeliness, and credibility of the assessment.

The committee’s most important conclusions and recommendations are summarized below. The committee believes that implementation of its recommendations will do much to enhance the credibility and usefulness of risk assessment.

Design of Risk Assessment

The process of planning risk assessment and ensuring that its level and complexity are consistent

with the needs to inform decision-making can be thought of as the “design” of risk assessment. The committee encourages EPA to focus greater attention on design in the formative stages of risk assessment, specifically on planning and scoping and problem formulation, as articulated in EPA guidance for ecologic and cumulative risk assessment (EPA 1998, 2003).3 Good design involves bringing risk managers, risk assessors, and various stakeholders together early in the process to determine the major factors to be considered, the decision-making context, and the timeline and depth needed to ensure that the right questions are being asked in the context of the assessment.

Increased emphasis on planning and scoping and on problem formulation has been shown to lead to risk assessments that are more useful and better accepted by decision-makers (EPA 2002, 2003, 20044); however, incorporation of these stages in risk assessment has been inconsistent, as noted by their absence from various EPA guidance documents (EPA 2005a,b5). An important element of planning and scoping is definition of a clear set of options for consideration in decision-making where appropriate. This should be reinforced by the up-front involvement of decision-makers, stakeholders, and risk assessors, who together can evaluate whether the design of the assessment will address the identified problems.

3EPA (U.S. Environmental Protection Agency). 1998. Guidelines for Ecological Risk Assessment. EPA/630/R- 95/002F. Risk Assessment Forum, U.S. Environmental Protection Agency, Washington, DC; EPA (U.S. Environmental Protection Agency). 2003. Framework for Cumulative Risk Assessment. EPA/600/P-02/001F. National Center for Environmental Assessment, Risk Assessment Forum, U.S. Environmental Protection Agency, Washington, DC.

4EPA (U.S. Environmental Protection Agency). 2002. A Review of the Reference Dose and Reference Concentration Processes. EPA/630/P-02/002F. Risk Assessment Forum, U.S. Environmental Protection Agency, Washington, DC; EPA (U.S. Environmental Protection Agency). 2003. Framework for Cumulative Risk Assessment. EPA/600/P-02/001F. National Center for Environmental Assessment, Risk Assessment Forum, U.S. Environmental Protection Agency, Washington, DC; EPA (U.S. Environmental Protection Agency). 2004. Risk Assessment Principles and Practices. Staff Paper. EPA/100/B-04/001. Office of the Science Advisor, U.S. Environmental Protection Agency, Washington, DC.

5EPA (U.S. Environmental Protection Agency). 2005a. Guidelines for Carcinogen Risk Assessment. EPA/630/P- 03/001F. Risk Assessment Forum, U.S. Environmental Protection Agency, Washington, DC; EPA (U.S. Environmental Protection Agency). 2005b. Supplemental Guidance for Assessing Susceptibility for Early-Life Exposures to Carcinogens. EPA/630/R-03/003F. Risk Assessment Forum, U.S. Environmental Protection Agency, Washington, DC.

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Science and Decisions: Advancing Risk Assessment

6 Prepublication Copy

Recommendation: Increased attention to the design of risk assessment in its formative stages is needed. The committee recommends that planning and scoping and problem formulation, as articulated in EPA guidance documents (EPA 1998, 2003), should be formalized and implemented in EPA risk assessments.

Uncertainty and Variability

Addressing uncertainty and variability is critical for the risk-assessment process. Uncertainty

stems from lack of knowledge, so it can be characterized and managed but not eliminated. Uncertainty can be reduced by the use of more or better data. Variability is an inherent characteristic of a population, inasmuch as people vary substantially in their exposures and their susceptibility to potentially harmful effects of the exposures. Variability cannot be reduced, but it can be better characterized with improved information.

There have been substantial differences among EPA’s approaches to and guidance for addressing uncertainty in exposure and dose-response assessment. EPA does not have a consistent approach to determine the level of sophistication or the extent of uncertainty analysis needed to address a particular problem. The level of detail for characterizing uncertainty is appropriate only to the extent that it is needed to inform specific risk-management decisions appropriately. It is important to address the required extent and nature of uncertainty analysis in the planning and scoping phases of a risk assessment. Inconsistency in the treatment of uncertainty among components of a risk assessment can make the communication of overall uncertainty difficult and sometimes misleading.

Variability in human susceptibility has not received sufficient or consistent attention in many EPA health risk assessments although there are encouraging exceptions, such as those for lead, ozone, and sulfur oxides. For example, although EPA’s 2005 Guidelines for Carcinogen Risk Assessment acknowledges that susceptibility can depend on one’s stage in life, greater attention to susceptibility in practice is needed, particularly for specific population groups that may have greater susceptibility because of their age, ethnicity, or socioeconomic status. The committee encourages EPA to move toward the long- term goal of quantifying population variability more explicitly in exposure assessment and dose-response relationships. An example of progress that moves towards this goal is EPA’s draft risk assessment of trichloroethylene (EPA 2001; NRC 2006), which considers how differences in metabolism, disease, and other factors contribute to human variability in response to exposures. Recommendation: EPA should encourage risk assessments to characterize and communicate uncertainty and variability in all key computational steps of risk assessment—for example, exposure assessment and dose-response assessment. Uncertainty and variability analysis should be planned and managed to reflect the needs for comparative evaluation of the risk management options. In the short term, EPA should adopt a “tiered” approach for selecting the level of detail to be used in the uncertainty and variability assessments, and this should be made explicit in the planning stage. To facilitate the characterization and interpretation of uncertainty and variability in risk assessments, EPA should develop guidance to determine the appropriate level of detail needed in uncertainty and variability analyses to support decision-making and should provide clear definitions and methods for identifying and addressing different sources of uncertainty and variability.

Selection and Use of Defaults

Uncertainty is inherent in all stages of risk assessment, and EPA typically relies on assumptions

when chemical-specific data are not available. The 1983 Red Book recommended the development of guidelines to justify and select from among the available inference options, the assumptions—now called

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Summary

Prepublication Copy 7

defaults—to be used in agency risk assessments to ensure consistency and avoid manipulations in the risk-assessment process. The committee acknowledges EPA’s efforts to examine scientific data related to defaults (EPA 1992, 2004, 2005a),6 but recognizes that changes are needed to improve the agency’s use of them. Much of the scientific controversy and delay in completion of some risk assessments has stemmed from the long debates regarding the adequacy of the data to support a default or an alternative approach. The committee concludes that established defaults need to be maintained for the steps in risk assessment that require inferences and that clear criteria should be available for judging whether, in specific cases, data are adequate for direct use or to support an inference in place of a default. EPA, for the most part, has not yet published clear, general guidance on what level of evidence is needed to justify use of agent-specific data and not resort to a default. There are also a number of defaults (missing or implicit defaults) that are engrained in EPA risk-assessment practice but are absent from its risk- assessment guidelines. For example, chemicals that have not been examined sufficiently in epidemiologic or toxicologic studies are often insufficiently considered in or are even excluded from risk assessments; because no description of their risks is included in the risk characterization, they carry no weight in decision-making. That occurs in Superfund-site and other risk assessments, in which a relatively short list of chemicals on which there are epidemiologic and toxicologic data tends to drive the exposure and risk assessments. Recommendation: EPA should continue and expand use of the best, most current science to support and revise default assumptions. EPA should work toward the development of explicitly stated defaults to take the place of implicit defaults. EPA should develop clear, general standards for the level of evidence needed to justify the use of alternative assumptions in place of defaults. In addition, EPA should describe specific criteria that need to be addressed for the use of alternatives to each particular default assumption. When EPA elects to depart from a default assumption, it should quantify the implications of using an alternative assumption, including how use of the default and the selected alternative influences the risk estimate for risk management options under consideration. EPA needs to more clearly elucidate a policy on defaults and provide guidance on its implementation and on evaluation of its impact on risk decisions and on efforts to protect the environment and public health.

A Unified Approach to Dose-Response Assessment

A challenge to risk assessment is to evaluate risks in ways that are consistent among chemicals,

that account adequately for variability and uncertainty, and that provide information that is timely, efficient, and maximally useful for risk characterization and risk management. Historically, dose-response assessments at EPA have been conducted differently for cancer and noncancer effects, and the methods have been criticized for not providing the most useful results. Consequently, noncancer effects have been underemphasized, especially in benefit-cost analyses. A consistent approach to risk assessment for cancer and noncancer effects is scientifically feasible and needs to be implemented.

For cancer, it has generally been assumed that there is no dose threshold of effect, and dose- response assessments have focused on quantifying risk at low doses and estimating a population risk for a given magnitude of exposure. For noncancer effects, a dose threshold (low-dose nonlinearity) has been assumed, below which effects are not expected to occur or are extremely unlikely in an exposed

6EPA (U.S. Environmental Protection Agency). 1992. Guidelines for Exposure Assessment. EPA/600/Z-92/001. Risk Assessment Forum, Office of Research and Development, U.S. Environmental Protection Agency, Washington, DC.; EPA (U.S. Environmental Protection Agency). 2004. Risk Assessment Principles and Practices. Staff Paper. EPA/100/B-04/001. Office of the Science Advisor, U.S. Environmental Protection Agency, Washington, DC; EPA (U.S. Environmental Protection Agency). 2005a. Guidelines for Carcinogen Risk Assessment. EPA/630/P-03/001F. Risk Assessment Forum, U.S. Environmental Protection Agency, Washington, DC.

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Science and Decisions: Advancing Risk Assessment

8 Prepublication Copy

population; that dose is a reference dose (RfD) or a reference concentration (RfC)—it is thought “likely to be without an appreciable risk of deleterious effects” (EPA 2002).7

EPA’s treatment of noncancer and low-dose nonlinear cancer end points is a major step by the agency in an overall strategy to harmonize cancer and noncancer approaches to dose-response assessment; however, the committee finds scientific and operational limitations in the current approaches. Noncancer effects do not necessarily have a threshold, or low-dose nonlinearity, and the mode of action of carcinogens varies. Background exposures and underlying disease processes contribute to population background risk and can lead to linearity at the population doses of concern. Because the RfD and RfC do not quantify risk for different magnitudes of exposure but rather provide a bright line between possible harm and safety, their use in risk-risk and risk-benefit comparisons and in risk-management decision- making is limited. Cancer risk assessments usually do not account for differences among humans in cancer susceptibility other than possible differences in early-life susceptibility.

Scientific and risk-management considerations both support unification of cancer and noncancer dose-response assessment approaches. The committee therefore recommends a consistent, unified approach for dose-response modeling that includes formal, systematic assessment of background disease processes and exposures, possible vulnerable populations, and modes of action that may affect a chemical’s dose-response relationship in humans. That approach redefines the RfD or RfC as a risk- specific dose that provides information on the percentage of the population that can be expected to be above or below a defined acceptable risk with a specific degree of confidence. The risk-specific dose will allow risk managers to weigh alternative risk options with respect to that percentage of the population. It will also permit a quantitative estimate of benefits for different risk-management options. For example, a risk manager could consider various population risks associated with exposures resulting from different control strategies for a pollution source and the benefits associated with each strategy. The committee acknowledges the widespread applications and public-health utility of the RfD; the redefined RfD can still be used as the RfD has been to aid risk-management decisions.

Characteristics of the committee’s recommended unified dose-response approach include use of a spectrum of data from human, animal, mechanistic, and other relevant studies; a probabilistic characterization of risk; explicit consideration of human heterogeneity (including age, sex, and health status) for both cancer and noncancer end points; characterization (through distributions to the extent possible) of the most important uncertainties for cancer and noncancer end points; evaluation of background exposure and susceptibility; use of probabilistic distributions instead of uncertainty factors when possible; and characterization of sensitive populations.

The new unified approach will require implementation and development as new chemicals are assessed or old chemicals are reassessed, including the development of test cases to demonstrate proof of concept. Recommendation: The committee recommends that EPA implement a phased-in approach to consider chemicals under a unified dose-response assessment framework that includes a systematic evaluation of background exposures and disease processes, possible vulnerable populations, and modes of action that may affect human dose-response relationships. The RfD and RfC should be redefined to take into account the probability of harm. In developing test cases, the committee recommends a flexible approach in which different conceptual models can be applied in the unified approach.

7EPA (U.S. Environmental Protection Agency). 2002. A Review of the Reference Dose and Reference

Concentration Processes. EPA/630/P-02/002F. Risk Assessment Forum, U.S. Environmental Protection Agency, Washington, DC. December 2002.

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Summary

Prepublication Copy 9

Cumulative Risk Assessment

EPA is increasingly asked to address broader public-health and environmental-health questions involving multiple exposures, complex mixtures, and vulnerability of exposed populations—issues that stakeholder groups (such as communities affected by environmental exposures) often consider to be inadequately captured by current risk assessments. There is a need for cumulative risk assessments as defined by EPA (EPA 2003)8—assessments that include combined risks posed by aggregate exposure to multiple agents or stressors; aggregate exposure includes all routes, pathways, and sources of exposure to a given agent or stressor. Chemical, biologic, radiologic, physical, and psychologic stressors are considered in this definition (Callahan and Sexton 2007).9

The committee applauds the agency’s move toward the broader definition in making risk assessment more informative and relevant to decisions and stakeholders. However, in practice, EPA risk assessments often fall short of what is possible and is supported by agency guidelines in this regard. Although cumulative risk assessment has been used in various contexts, there has been little consideration of nonchemical stressors, vulnerability, and background risk factors. Because of the complexity of considering so many factors simultaneously, there is a need for simplified risk-assessment tools (such as databases, software packages, and other modeling resources) that would allow screening-level risk assessments and could allow communities and stakeholders to conduct assessments and thus increase stakeholder participation. Cumulative human health risk assessment should draw greater insights from ecologic risk assessment and social epidemiology, which have had to grapple with similar issues. (Cumulative risk assessment will be addressed in a forthcoming National Research Council report on phthalates.) Recommendation: EPA should draw on other approaches, including those from ecologic risk assessment and social epidemiology, to incorporate interactions between chemical and non- chemical stressors in assessments; increase the role of biomonitoring, epidemiologic, and surveillance data in cumulative risk assessments; and develop guidelines and methods for simpler analytical tools to support cumulative risk assessment and to provide for greater involvement of stakeholders. In the short-term, EPA should develop databases and default approaches to allow for incorporation of key non-chemical stressors in cumulative risk assessments in the absence of population-specific data, considering exposure patterns, contributions to relevant background processes, and interactions with chemical stressors. In the long-term, EPA should invest in research programs related to interactions between chemical and non-chemical stressors, including epidemiologic investigations and physiologically-based pharmacokinetic modeling.

Improving the Utility of Risk Assessment

Given the complexities of the current problems and potential decisions faced by EPA, the

committee grappled with designing a more coherent, consistent, and transparent process that would provide risk assessments that are relevant to the problems and decisions at hand and that would be sufficiently comprehensive to ensure that the best available options for managing risks were considered. To that end, the committee proposes a framework for risk-based decision-making (see Figure S-1). The framework consists of three phases: I, enhanced problem formulation and scoping, in which the available

8EPA (U.S. Environmental Protection Agency). 2003. Framework for Cumulative Risk Assessment. EPA/600/P-

02/001F. National Center for Environmental Assessment, Risk Assessment Forum, U.S. Environmental Protection Agency, Washington, DC.

9Callahan, M.A., and K. Sexton. 2007. If ‘cumulative risk assessment’ is the answer, what is the question? Environ. Health Perspect. 115(5):799-806.

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Science and Decisions: Advancing Risk Assessment

10 Prepublication Copy

• What are the relative health or environmental benefits of the proposed options?

• How are other decision- making factors (technologies, costs) affected by the proposed options?

• What is the decision, and its justification, in light of benefits, costs, and uncertainties in each?

• How should the decision be communicated?

• Is it necessary to evaluate the effectiveness of the decision?

• If so, how should this be done?

Stage 1: Planning

• For the given decision-context, what are the attributes of assessments necessary to characterize risks of existing conditions and the effects on risk of proposed options? What level of uncertainty and variability analysis is appropriate?

Stage 3: Confirmation of Utility

• Does the assessment have the attributes called for in planning?

• Does the assessment provide sufficient information to discriminate among risk management options?

• Has the assessment been satisfactorily peer reviewed?

FORMAL PROVISIONS FOR INTERNAL AND EXTERNAL STAKEHOLDER INVOLVEMENT AT ALL STAGES

• The involvement of decision-makers, technical specialists, and other stakeholders in all phases of the processes leading to decisions should in no way compromise the technical assessment of risk, which is carried out under its own standards and guidelines.

• What problem(s) are associated with existing environmental conditions?

• If existing conditions appear to pose a threat to human or environmental health, what options exist for altering those conditions?

• Under the given decision context, what risk and other technical assessments are necessary to evaluate the possible risk management options?

• Hazard Identification

What adverse health or environmental effects are associated with the agents of concern?

• Dose-Response Assessment

For each determining adverse effect, what is the relationship between dose and the probability of the occurrence of the adverse effects in the range of doses identified in the exposure assessment?

• Risk Characterization

What is the nature and magnitude of risk associated with existing conditions?

What risk decreases (benefits) are associated with each of the options?

Are any risks increased? What are the significant uncertainties?

• Exposure Assessment

What exposures/doses are incurred by each population of interest under existing conditions?

How does each option affect existing conditions and resulting exposures/doses?

Stage 2: Risk Assessment

NO YES

PHASE I: PROBLEM FORMULATION

AND SCOPING

PHASE II: PLANNING AND CONDUCT

OF RISK ASSESSMENT

PHASE III: RISK MANAGEMENT

FIGURE S-1 A framework for risk-based decision-making that maximizes the utility of risk assessment. risk-management options are identified; II, planning and assessment, in which risk-assessment tools are used to determine risks under existing conditions and under potential risk-management options; and III, risk management, in which risk and nonrisk information is integrated to inform choices among options.

The framework has at its core the risk-assessment paradigm (stage 2 of phase II) established in the Red Book (NRC 1983).10 However, the framework differs from the Red Book paradigm, primarily in its initial and final steps. The framework begins with a “signal” of potential harm (for example, a positive bioassay or epidemiologic study, a suspicious disease cluster, or findings of industrial contamination). Under the traditional paradigm, the question has been, What are the probability and consequence of an adverse health (or ecologic) effect posed by the signal?” In contrast, the recommended framework asks, implicitly, What options are there to reduce the hazards or exposures that have been identified, and how can risk assessment be used to evaluate the merits of the various options? The latter question focuses on the risk-management options (or interventions) designed to provide adequate public-health and environmental protection and to ensure well-supported decision-making. Under this framework, the questions posed arise from early and careful planning of the types of assessments (including risks, costs, and technical feasibility) and the required level of scientific depth that are needed to evaluate the relative

10NRC (National Research Council). 1983. Risk Assessment in the Federal Government: Managing the Process.

Washington DC: National Academy Press.

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Summary

Prepublication Copy 11

merits of the options being considered.11 Risk management involves choosing among the options after the appropriate assessments have been undertaken and evaluated.

The framework begins with enhanced problem formulation and scoping (phase I), in which risk- management options and the types of technical analyses, including risk assessments, needed to evaluate and discriminate among the options are identified. Phase II consists of three stages: planning, risk assessment, and confirmation of utility. Planning (stage 1) is done to ensure that the level and complexity of risk assessment (including uncertainty and variability analysis) are consistent with the goals of decision-making. After risk assessment (stage 2), stage 3 evaluates whether the assessment was appropriate and whether it allows discrimination among the risk-management options. If the assessment is determined not to be adequate, the framework calls for a return to planning (phase II, stage 1). Otherwise, phase III (risk management) is undertaken: the relative health or environmental benefits of the proposed risk-management options are evaluated for the purpose of reaching a decision.

The framework systematically identifies problems and options that risk assessors should evaluate at the earliest stages of decision-making. It expands the array of impacts assessed beyond individual effects (for example, cancer, respiratory problems, and individual species) to include broader questions of health status and ecosystem protection. It provides a formal process for stakeholder involvement throughout all stages but has time constraints to ensure that decisions are made. It increases understanding of the strengths and limitations of risk assessment by decision-makers at all levels, for example, by making uncertainties and choices more transparent.

The committee is mindful of concerns about political interference in the process, and the framework maintains the conceptual distinction between risk assessment and risk management articulated in the Red Book. It is imperative that risk assessments used to evaluate risk-management options not be inappropriately influenced by the preferences of risk managers.

With a focus on early and careful planning and problem formulation and on the options for managing the problem, implementation of the framework can improve the utility of risk assessment for decision-making. Although some aspects of the framework are achievable in the short term, its full implementation will require a substantial transition period. EPA should phase in the framework with a series of demonstration projects that apply it and that determine the degree to which it meets the needs of the agency risk managers, how risk-management conclusions differ as a result of its application, and the effectiveness of measures to ensure that risk managers and policy-makers do not inappropriately influence the scientific conduct of risk assessments. Recommendations: To make risk assessments most useful for risk management decisions, the committee recommends that EPA adopt a framework for risk-based decision-making (see Figure S-1) that embeds the Red Book risk assessment paradigm into a process with initial problem formulation and scoping, upfront identification of risk-management options, and use of risk assessment to discriminate among these options.

Stakeholder Involvement

Many stakeholders believe that the current process for developing and applying risk assessments

lacks credibility and transparency. That may be partly because of failure to involve stakeholders adequately as active participants at appropriate points in the risk-assessment and decision-making process rather than as passive recipients of the results. Previous National Research Council and other risk-

11The committee notes that not all decisions require or are amenable to risk assessment and that in most cases one

of the options explicitly considered is “no intervention.”

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Science and Decisions: Advancing Risk Assessment

12 Prepublication Copy

assessment reports (for example, NRC 1996; PCCRARM 1997)12 and comments received by the committee (Callahan 2007; Kyle 2007)13 echo such concerns.

The committee agrees that greater stakeholder involvement is necessary to ensure that the process is transparent and that risk-based decision-making proceeds effectively, efficiently, and credibly. Stakeholder involvement needs to be an integral part of the risk-based decision-making framework, beginning with problem formulation and scoping.

Although EPA has numerous programs and guidance documents related to stakeholder involvement, it is important that it adhere to its own guidance, particularly in the context of cumulative risk assessment, in which communities often have not been adequately involved. Recommendation: EPA should establish a formal process for stakeholder involvement in the framework for risk-based decision-making with time limits to ensure that decision-making schedules are met and with incentives to allow for balanced participation of stakeholders, including impacted communities and less advantaged stakeholders.

Capacity-Building

Improving risk-assessment practice and implementing the framework for risk-based decision- making will require a long-term plan and commitment to build the requisite capacity of information, skills, training, and other resources necessary to improve public-health and environmental decision- making. The committee’s recommendations call for considerable modification of EPA risk-assessment efforts (for example, implementation of the risk-based decision-making framework, emphasis on problem formulation and scoping as a discrete stage in risk assessment, and greater stakeholder participation) and of technical aspects of risk assessment (for example, unification of cancer and noncancer dose-response assessments, attention to quantitative uncertainty analysis, and development of methods for cumulative risk assessment). The recommendations are tantamount to “change-the-culture” transformations in risk assessment and decision-making in the agency.

EPA’s current institutional structure and resources may pose a challenge to implementation of the recommendations, and moving forward with them will require a commitment to leadership, cross- program coordination and communication, and training to ensure the requisite expertise. That will be possible only if leaders are determined to reverse the downward trend in budgeting, staffing, and training and to making high-quality, risk-based decision-making an agencywide goal. Recommendation: EPA should initiate a senior-level strategic re-examination of its risk-related structures and processes to ensure that it has the institutional capacity to implement the committee’s recommendations for improving the conduct and utility of risk assessment for meeting the 21st century environmental challenges. EPA should develop a capacity building plan that includes budget estimates required for implementing the committee’s recommendations, including transitioning to and effectively implementing the framework for risk-based decision-making.

12NRC (National Research Council). 1996. Understanding Risk: Informing Decisions in a Democratic Society.

Washington DC: National Academy Press; PCCRARM (Presidential/Congressional Commission on Risk Assessment and Risk Management). 1997. Framework for Environmental Health Risk Management - Final Report, Vol. 1.

13Callahan, M.A. 2007. Improving Risk Assessment: A Regional Perspective. Presentation at the Third Meeting of Improving Risk Analysis Approaches Used by EPA, February 26, 2007, Washington, DC; Kyle, A. 2007. Community Needs for Assessment of Environmental Problems. Presentation at the Fourth Meeting of Improving Risk Analysis Approaches Used by EPA, April 17, 2007, Washington, DC.

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Summary

Prepublication Copy 13

CONCLUDING REMARKS

Global impacts are combining with the high financial and political stakes of risk management to place unprecedented pressure on risk assessors in EPA. But risk assessment remains essential to the agency’s mission to ensure protection of public health and the environment. Much work is needed to improve the scientific status, utility, and public credibility of risk assessment. The committee’s recommendations focus on designing risk assessments to ensure that they make the best possible use of available science, are technically accurate, and address the appropriate risk-management options effectively to inform risk-based decision-making. The committee hopes that the recommendations and the proposed framework for risk-based decision-making will provide a template for the future of risk assessment in EPA and strengthen the scientific basis, credibility, and effectiveness of future risk- management decisions.

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

PREPUBLICATION COPY

Science and Decisions: Advancing Risk Assessment

Committee on Improving Risk Analysis Approaches Used by the U.S. EPA

Board on Environmental Studies and Toxicology

Division on Earth and Life Studies

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Prepublication Copy

THE NATIONAL ACADEMIES PRESS 500 Fifth Street, NW Washington, DC 20001 NOTICE: The project that is the subject of this report was approved by the Governing Board of the National Research Council, whose members are drawn from the councils of the National Academy of Sciences, the National Academy of Engineering, and the Institute of Medicine. The members of the committee responsible for the report were chosen for their special competences and with regard for appropriate balance. This project was supported by Contract EP-C-06-056 between the National Academy of Sciences and the U.S. Environmental Protection Agency. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the authors and do not necessarily reflect the view of the organizations or agencies that provided support for this project. Additional copies of this report are available from The National Academies Press 500 Fifth Street, NW Box 285 Washington, DC 20055 800-624-6242 202-334-3313 (in the Washington metropolitan area) http://www.nap.edu Copyright 2008 by the National Academy of Sciences. All rights reserved. Printed in the United States of America

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Prepublication Copy

The National Academy of Sciences is a private, nonprofit, self-perpetuating society of distinguished scholars engaged in scientific and engineering research, dedicated to the furtherance of science and technology and to their use for the general welfare. Upon the authority of the charter granted to it by the Congress in 1863, the Academy has a mandate that requires it to advise the federal government on scientific and technical matters. Dr. Ralph J. Cicerone is president of the National Academy of Sciences.

The National Academy of Engineering was established in 1964, under the charter of the National Academy of Sciences, as a parallel organization of outstanding engineers. It is autonomous in its administration and in the selection of its members, sharing with the National Academy of Sciences the responsibility for advising the federal government. The National Academy of Engineering also sponsors engineering programs aimed at meeting national needs, encourages education and research, and recognizes the superior achievements of engineers. Dr. Charles M. Vest is president of the National Academy of Engineering.

The Institute of Medicine was established in 1970 by the National Academy of Sciences to secure the services of eminent members of appropriate professions in the examination of policy matters pertaining to the health of the public. The Institute acts under the responsibility given to the National Academy of Sciences by its congressional charter to be an adviser to the federal government and, upon its own initiative, to identify issues of medical care, research, and education. Dr. Harvey V. Fineberg is president of the Institute of Medicine.

The National Research Council was organized by the National Academy of Sciences in 1916 to associate the broad community of science and technology with the Academy’s purposes of furthering knowledge and advising the federal government. Functioning in accordance with general policies determined by the Academy, the Council has become the principal operating agency of both the National Academy of Sciences and the National Academy of Engineering in providing services to the government, the public, and the scientific and engineering communities. The Council is administered jointly by both Academies and the Institute of Medicine. Dr. Ralph J. Cicerone and Dr. Charles M. Vest are chair and vice chair, respectively, of the National Research Council.

www.national-academies.org

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Prepublication Copy v

COMMITTEE ON IMPROVING RISK ANALYSIS APPROACHES USED BY THE U.S. ENVIRONMENTAL PROTECTION AGENCY

Members THOMAS A. BURKE (Chair), Johns Hopkins Bloomberg School of Public Health, Baltimore, MD A. JOHN BAILER, Miami University, Oxford, OH JOHN M. BALBUS, Environmental Defense, Washington, DC JOSHUA T. COHEN, Tufts Medical Center, Boston, MA ADAM M. FINKEL, University of Medicine and Dentistry of New Jersey, Piscataway, NJ GARY GINSBERG, Connecticut Department of Public Health, Hartford, CT BRUCE K. HOPE, Oregon Department of Environmental Quality, Portland, OR JONATHAN I. LEVY, Harvard School of Public Health, Boston, MA THOMAS E. MCKONE, University of California, Berkeley, CA GREGORY M. PAOLI, Risk Sciences International, Ottawa, ON, Canada CHARLES POOLE, University of North Carolina School of Public Health, Chapel Hill, NC JOSEPH V. RODRICKS, ENVIRON International Corporation, Arlington, VA BAILUS WALKER JR., Howard University Medical Center, Washington, DC TERRY F. YOSIE, World Environment Center, Washington, DC LAUREN ZEISE, California Environmental Protection Agency, Oakland, CA Staff EILEEN N. ABT, Senior Project Director JENNIFER SAUNDERS, Associate Program Officer (through December 2007) NORMAN GROSSBLATT, Senior Editor RUTH CROSSGROVE, Senior Editor MIRSADA KARALIC-LONCAREVIC, Manager, Technical Information Center RADIAH A. ROSE, Editorial Projects Manager MORGAN R. MOTTO, Senior Program Assistant (through February 2008) PANOLA GOLSON, Senior Program Assistant Sponsor U.S. ENVIRONMENTAL PROTECTION AGENCY

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

vi Prepublication Copy

BOARD ON ENVIRONMENTAL STUDIES AND TOXICOLOGY1 Members JONATHAN M. SAMET (Chair), University of Southern California, Los Angeles RAMO N ALVAREZ, Environmental Defense Fund, Austin, TX JOHN M. BALBUS, Environmental Defense Fund, Washington, DC DALLAS BURTRAW, Resources for the Future, Washington, DC JAMES S. BUS, Dow Chemical Company, Midland, MI RUTH DEFRIES, Columbia University, New York, NY COSTEL D. DENSON, University of Delaware, Newark E. DONALD ELLIOTT, Willkie, Farr & Gallagher LLP, Washington, DC MARY R. ENGLISH, University of Tennessee, Knoxville J. PAUL GILMAN, Covanta Energy Corporation, Fairfield, NJ JUDITH A. GRAHAM (Retired), Pittsboro, NC WILLIAM M. LEWIS, JR., University of Colorado, Boulder JUDITH L. MEYER, University of Georgia, Athens DENNIS D. MURPHY, University of Nevada, Reno DANNY D. REIBLE, University of Texas, Austin JOSEPH V. RODRICKS, ENVIRON International Corporation, Arlington, VA ARMISTEAD G. RUSSELL, Georgia Institute of Technology, Atlanta ROBERT F. SAWYER, University of California, Berkeley KIMBERLY M. THOMPSON, Harvard School of Public Health, Boston, MA MARK J. UTELL, University of Rochester Medical Center, Rochester, NY Senior Staff JAMES J. REISA, Director DAVID J. POLICANSKY, Scholar RAYMOND A. WASSEL, Senior Program Officer for Environmental Studies EILEEN N. ABT, Senior Program Officer for Risk Analysis SUSAN N.J. MARTEL, Senior Program Officer for Toxicology KULBIR BAKSHI, Senior Program Officer ELLEN K. MANTUS, Senior Program Officer RUTH E. CROSSGROVE, Senior Editor

1This study was planned, overseen, and supported by the Board on Environmental Studies and Toxicology.

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Prepublication Copy vii

OTHER REPORTS OF THE BOARD ON ENVIRONMENTAL STUDIES AND TOXICOLOGY

Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution

(2008) Respiratory Diseases Research at NIOSH (2008) Evaluating Research Efficiency in the U.S. Environmental Protection Agency (2008) Hydrology, Ecology, and Fishes of the Klamath River Basin (2008) Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment (2007) Models in Environmental Regulatory Decision Making (2007) Toxicity Testing in the Twenty-first Century: A Vision and a Strategy (2007) Sediment Dredging at Superfund Megasites: Assessing the Effectiveness (2007) Environmental Impacts of Wind-Energy Projects (2007) Scientific Review of the Proposed Risk Assessment Bulletin from the Office of Management and

Budget (2007) Assessing the Human Health Risks of Trichloroethylene: Key Scientific Issues (2006) New Source Review for Stationary Sources of Air Pollution (2006) Human Biomonitoring for Environmental Chemicals (2006) Health Risks from Dioxin and Related Compounds: Evaluation of the EPA Reassessment (2006) Fluoride in Drinking Water: A Scientific Review of EPA’s Standards (2006) State and Federal Standards for Mobile-Source Emissions (2006) Superfund and Mining Megasites—Lessons from the Coeur d’Alene River Basin (2005) Health Implications of Perchlorate Ingestion (2005) Air Quality Management in the United States (2004) Endangered and Threatened Species of the Platte River (2004) Atlantic Salmon in Maine (2004) Endangered and Threatened Fishes in the Klamath River Basin (2004) Cumulative Environmental Effects of Alaska North Slope Oil and Gas Development (2003) Estimating the Public Health Benefits of Proposed Air Pollution Regulations (2002) Biosolids Applied to Land: Advancing Standards and Practices (2002) The Airliner Cabin Environment and Health of Passengers and Crew (2002) Arsenic in Drinking Water: 2001 Update (2001) Evaluating Vehicle Emissions Inspection and Maintenance Programs (2001) Compensating for Wetland Losses Under the Clean Water Act (2001) A Risk-Management Strategy for PCB-Contaminated Sediments (2001) Acute Exposure Guideline Levels for Selected Airborne Chemicals (six volumes, 2000-2008) Toxicological Effects of Methylmercury (2000) Strengthening Science at the U.S. Environmental Protection Agency (2000) Scientific Frontiers in Developmental Toxicology and Risk Assessment (2000) Ecological Indicators for the Nation (2000) Waste Incineration and Public Health (2000) Hormonally Active Agents in the Environment (1999) Research Priorities for Airborne Particulate Matter (four volumes, 1998-2004) The National Research Council’s Committee on Toxicology: The First 50 Years (1997) Carcinogens and Anticarcinogens in the Human Diet (1996) Upstream: Salmon and Society in the Pacific Northwest (1996) Science and the Endangered Species Act (1995) Wetlands: Characteristics and Boundaries (1995) Biologic Markers (five volumes, 1989-1995) Science and Judgment in Risk Assessment (1994)

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

viii Prepublication Copy

Pesticides in the Diets of Infants and Children (1993) Dolphins and the Tuna Industry (1992) Science and the National Parks (1992) Human Exposure Assessment for Airborne Pollutants (1991) Rethinking the Ozone Problem in Urban and Regional Air Pollution (1991) Decline of the Sea Turtles (1990)

Copies of these reports may be ordered from the National Academies Press (800) 624-6242 or (202) 334-3313

www.nap.edu

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Prepublication Copy ix

Preface

Risk assessment has become a dominant public-policy tool for informing risk managers and the

public about the different policy options for protecting public health and the environment. Risk assessment has been instrumental in fulfilling the missions of the U.S. Environmental Protection Agency (EPA) and other federal and state agencies in evaluating public-health concerns, informing regulatory and technologic decisions, setting priorities for research and funding, and developing approaches for cost- benefit analyses.

However, risk assessment is at a crossroads. Despite advances in the field, it faces a number of substantial challenges, including long delays in completing complex risk assessments, some of which take decades to complete; lack of data, which leads to important uncertainty in risk assessments; and the need for risk assessment of many unevaluated chemicals in the marketplace and emerging agents. To address those challenges, EPA asked the National Academies to develop recommendations for improving the agency’s risk-analysis approaches.

In this report, the Committee on Improving Risk Analysis Approaches Used by the U.S. EPA conducts a scientific and technical review of EPA's current risk-analysis concepts and practices and offers recommendations for practical improvements that EPA could make in the near term (2-5 y) and in the longer term (10-20 y). The committee focused on human health risk assessment but considered the implications of its conclusions and recommendations for ecologic risk assessment.

This report has been reviewed in draft form by persons chosen for their diverse perspectives and technical expertise in accordance with procedures approved by the National Research Council’s Report Review Committee. The purpose of this independent review is to provide candid and critical comments that will assist the institution in making its published report as sound as possible and to ensure that the report meets institutional standards of objectivity, evidence, and responsiveness to the study charge. The review comments and draft manuscript remain confidential to protect the integrity of the deliberative process. We wish to thank the following for their review of this report: Lawrence W. Barnthouse; LWB Environmental Services, Inc.; Roger G. Bea, University of California, Berkeley; Allison C. Cullen, University of Washington; William H. Farland, Colorado State University; J. Paul Gilman, Convanta Energy Corporation; Bernard D. Goldstein, University of Pittsburgh; Lynn R. Goldman, Johns Hopkins University; Dale B. Hattis, Clark University; Carol J. Henry, American Chemistry Council (retired); Daniel Krewski, University of Ottawa; Amy D. Kyle, University of California, Berkeley; Ronald L. Melnick, National Institute of Environmental Health Sciences; Gilbert S. Omenn, University of Michigan Medical School; Louis Ryan, Harvard School of Public Health; and Detlof von Winterfeldt, University of Southern California.

Although the reviewers listed above have provided many constructive comments and suggestions, they were not asked to endorse the conclusions or recommendations, nor did they see the final draft of the report before its release. The review of the report was overseen by the review coordinator William Glaze, Georgetown, TX and the review monitor, John Ahearne, Sigma Xi. Appointed by the National Research Council, they were responsible for making certain that an independent examination of the report was

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Preface

x Prepublication Copy

carried out in accordance with institutional procedures and that all review comments were carefully considered. Responsibility for the final content of the report rests entirely with the committee and the institution.

The committee gratefully acknowledges the following for making presentations to the committee: Nicholas Ashford, Massachusetts Institute of Technology; Robert Brenner, Michael Callahan, George Gray, Jim Jones, Tina Levine, Robert Kavlock, Al McGartland, Peter Preuss, Michael Shapiro, Glenn Suter, and Harold Zenick, EPA; Douglas Crawford-Brown, University of North Carolina; Kenny Crump, ENVIRON International Corporation; Robert Donkers, Delegation of the European Commission to the United States; William Farland, Colorado State University; James A. Fava, Five Winds International; Penny Fenner-Crisp, International Life Sciences Institute Research Foundation; Dale Hattis, Clark University; Amy D. Kyle, University of California, Berkeley; Rebecca Parkin, George Washington University; Chris Portier, National Institute of Environmental Health Sciences; Lorenz Rhomberg, Gradient Corporation; Jennifer Sass, Natural Resources Defense Council; Jay Silkworth, General Electric Company; and Thomas Sinks, Centers for Disease Control and Prevention.

The committee is thankful for the useful input of Roger Cooke, Resources for the Future and Dorothy Patton, Environmental Protection Agency (retired) in the early deliberations of this study. The committee is also grateful for the assistance of the National Research Council staff in preparing this report. Staff members who contributed to this effort are Eileen Abt, project director; James Reisa, director of the Board on Environmental Studies and Toxicology; Jennifer Saunders, associate program officer; Norman Grossblatt, senior editor; Ruth Crossgrove, senior editor; Mirsada Karalic-Loncarevic, research associate; Radiah Rose, editorial projects manager; Morgan Motto, senior program assistant; and Panola Golson, senior program assistant.

I would especially like to thank the committee members for their efforts throughout the development of this report.

Thomas Burke, Chair Committee on Improving Risk Analysis Approaches Used by the U.S. EPA

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Prepublication Copy xi

Abbreviations

ARARs Applicable or Relevant and Appropriate Requirements ATSDR Agency for Toxic Substances and Disease Registry BMD benchmark dose CARE Community Action for a Renewed Environment CASAC Clean Air Scientific Advisory Committee CBPR community-based participatory research CERCLA Comprehensive Environmental Response Compensation and Liability Act CTE central tendency exposure DBP dibutyl phthalate DBPs disinfection byproducts EPA Environmental Protection Agency EPHT Environmental Public Health Tracking Program FIFRA Federal Insecticide, Fungicide and Rodenticide Act FQPA Food Quality Protection Act GAO Government Accountability Office GIS geographic information systems HAPs hazardous air pollutants HI hazard index IARC International Agency for Research on Cancer IPCS International Program on Chemical Safety IRIS Integrated Risk Information System LNT linear, no-threshold MACT maximum achievable control technology MCL maximum contaminant level MCLG maximum contaminant level goal MeCl2 methylene chloride MEI maximally exposed individual MOA mode of action MOE margin of exposure MTD maximum tolerated dose NAAQS National Ambient Air Quality Standards NCEA National Center for Environmental Assessment NEJAC National Environmental Justice Advisory Council NER National Exposure Registry NHANES National Health and Nutrition Examination Survey NOAEL no-observed-adverse-effect-level NPL National Priorities List

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Abbreviations

xii Prepublication Copy

NRC National Research Council NTP National Toxicology Program OAR Office of Air and Radiation OP organophosphate OPPTS Office of Prevention, Pesticides and Toxic Substances OSWER Office of Solid Waste and Emergency Response OW Office of Water PBPK physiologically based pharmacokinetic PD pharmacodynamic PDF probability density function PK pharmacokinetic POD point of departure PPDG Pesticide Program Dialogue Group RAGS Risk Assessment Guidance for Superfund Red Book Risk Assessment in the Federal Government: Managing the Process RfC reference concentration RfD reference dose RI/FS remedial investigation and feasibility study RME reasonable maximum exposure ROD record of decision RR relative risk RRM relative risk model SDWA Safe Drinking Water Act SEP socioeconomic position TCA 1,1,1-trichloroethane TCE trichloroethylene TSCA Toxic Substances Control Act UF uncertainty factor VOI value-of-information WHO World Health Organization WOE weight-of-evidence

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Prepublication Copy xiii

Contents

SUMMARY.................................................................................................................................................................3 1 INTRODUCTION ..........................................................................................................................................14 Background, 14 Challenges, 15 Traditional and Emerging Views of the Roles of Risk Assessment, 16 Technical Impediments to Risk Assessment, 18 Improving Risk Analysis, 20 The National Research Council Committee, 21 Organization of the Report, 21 References, 23 2 EVOLUTION AND USE OF RISK ASSESSMENT IN THE ENVIRONMENTAL

PROTECTION AGENCY: CURRENT PRACTICE AND FUTURE PROSPECTS.........................24 Overview, 24 Statutory Plan and Regulatory Structure, 24 The Pivotal Role of the Red Book, 27 Current Concepts and Practices, 33 Institutional Arrangements for Managing the Process, 43 Extramural Influences and Participants, 48 Conclusions and Recommendations, 52

References, 54 3 THE DESIGN OF RISK ASSESSMENTS.................................................................................................60 Risk Assessment as a Design Challenge, 60 Design Considerations: Objectives, Constraints, and Tradeoffs, 62 Environmental Protection Agency’s Current Guidance Related to Risk-Assessment Design, 67 Incorporating Value-of-Information Principles in Formative and Iterative Design, 74 Conclusions, 82 Recommendations, 82 References, 83 4 UNCERTAINTY AND VARIABILITY: THE RECURRING AND

RECALCITRANT ELEMENTS OF RISK ASSESSMENT...................................................................86 Introduction to the Issues and Terminology, 86

Uncertainty in Risk Assessment, 91 Variability and Vulnerability in Risk Assessment, 99 Uncertainty and Variability in Specific Components of Risk Assessment, 104 Principles for Addressing Uncertainty and Variability, 110 Recommendations, 112 References, 112

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Contents

xiv Prepublication Copy

5 TOWARD A UNIFIED APPROACH TO DOSE-RESPONSE ASSESSMENT ...............................118 The Need for an Improved Dose-Response Framework, 118 A Unified Framework and Approach for Dose-Response Assessment, 125 Case Studies and Possible Modeling Approaches, 139 Implementation, 160 Conclusions and Recommendations, 163 References, 167

6 SELECTION AND USE OF DEFAULTS................................................................................................174 Current Environmental Protection Agency Policy on Defaults, 175 The Environmental Protection Agency’s System of Defaults, 178 Complications Introduced by Use of Defaults, 180 Enhancements of Environmental Protection Agency’s Default Approach, 183 Performing Multiple Risk Characterizations for Alternative Models, 189 Conclusions and Recommendations, 191

References, 192 7 IMPLEMENTING CUMULATIVE RISK ASSESSMENT..................................................................197

Introduction and Definitions, 197 History of Cumulative Risk Assessment, 199 Approaches to Cumulative Risk Assessment, 202 Key Concerns and Proposed Modifications, 206 Recommendations, 217

References, 218 8 IMPROVING THE UTILITY OF RISK ASSESSMENT .....................................................................221 Beyond the Red Book, 222 A Decision-Making Framework that Maximizes the Utility of Risk Assessment, 222 The Framework: An Overview, 225 Additional Improvements Offered by the Framework, 231 Potential Concerns Raised by the Framework, 233 Conclusions and Recommendations, 235

References, 235 9 TOWARDS IMPROVED RISK-BASED DECISION-MAKING ........................................................237 Transition to the Framework for Risk-Based Decision-Making, 238 Institutional Processes, 238 Leadership and Management, 240

Conclusions and Recommendations, 241 References, 248

APPENDIX A: BIOGRAPHIC INFORMATION ON THE COMMITTEE ON IMPROVING

RISK ANALYSIS APPROACHES USED BY THE U.S. ENVIRONMENTAL PROTECTION AGENCY .......................................................................................................250

APPENDIX B: STATEMENT OF TASK OF THE COMMITTEE ON IMPROVING

RISK ANALYSIS APPROACHES USED BY THE U.S. ENVIRONMENTAL PROTECTION AGENCY .......................................................................................................255

APPENDIX C: TIMELINE OF SELECTED ENVIRONMENTAL PROTECTION AGENCY

RISK-ASSESSMENT ACTIVITIES .....................................................................................256

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Contents

Prepublication Copy xv

APPENDIX D: ENVIRONMENTAL PROTECTION AGENCY RESPONSE TO RECOMMENDATIONS FROM SELECTED NRC REPORTS: POLICY, ACTIVITY, AND PRACTICE ...............................................................................................272

APPENDIX E: EPA PROGRAM AND REGION RESPONSES TO QUESTIONS FROM

THE COMMITTEE .................................................................................................................330 APPENDIX F: CASE STUDIES OF THE FRAMEWORK FOR RISK-BASED

DECISION-MAKING ..............................................................................................................360

BOXES, FIGURES, AND TABLES BOXES

2-1 Agencywide Risk-Assessment Guidelines, 30 2-2 Science Policy and Defaults, 33 2-3 Agency Guidance on Risk Characterization: Attention to Uncertainty, 36 2-4 Commentary on Risk Characterization for the Dioxin Reassessment, 37 2-5 Guideline Implementation and Risk-Assessment Impacts, 38 2-6 Choices and a Reference Dose Value for Perchlorate, 40 2-7 Impact of New Studies, 43 2-8 Arsenic in Drinking Water: Uncertainties and Standard-Setting, 47 2-9 Risk Assessment Planning: Multiple Participants, 51 2-10 After Peer Review, 51 3-1 Selected Elements of Scope Considered During Planning and Scoping, 68 3-2 Selected Methodological Considerations in Problem Formulation, 69 3-3 Planning and Scoping: An Exampled Summary Statement, 71 3-4 Major Elements of an Analysis Plan, 73 4-1 Terminology Relating to Uncertainty and Variability, 88 4-2 Why is It Important to Quantify Uncertainty and Variability, 90 4-3 Cognitive Tendencies that Affect Expert Judgment, 95 4-4 Levels of Uncertainty Analysis, 96 4-5 Examples of Uncertainties for Comparisons of Discrete and Continuous Possibilities, 98 4-6 Expressing and Distinguishing Model and Parameter Uncertainty, 100 4-7 Recommended Principles for Uncertainty and Variability Analysis, 111 5-1 A Risk-Specific Reference Dose, 130 5-2 Conceptual Model 1: Default Linear Low-Dose Extrapolation for Phosgene, 147 5-3 Calculating a Risk-Specific Dose and Confidence Bound in Conceptual Model 2, 151 6-1 Boron: Use of Data Derived Uncertainty Factors, 185 8-1 Key Definitions Used in the Framework for Risk-Based Decision-Making, 227 8-2 Phase I of the Framework for Risk-Based Decision-Making (Problem Formulation

and Planning), 227 8-3 Phase II of the Framework for Risk-Based Decision-Making (Planning and Conduct of

Risk Assessment), 228 8-4 Other Technical Analysis Necessary for the Framework for Risk-Based Decision-Making, 229 8-5 Elements of Phase III of the Framework for Risk-Based Decision-Making (Risk Management), 229 FIGURES S-1 A framework for risk-based decision-making that maximizes the utility of risk assessment, 10 2-1 The National Research Council risk-assessment-risk-management paradigm, 29

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Contents

xvi Prepublication Copy

2-2 The World Health Organization’s framework for integrated health and ecological risk assessment, 32

2-3 Timeline of major documentary milestones, 34 3-1 Schematic representation of the formative stages of risk-assessment design, 68 3-2 Illustration of the scope of a risk assessment, indicating both pathways considered (bold lines) and

pathways not considered, 72 3-3 Schematic of the application of value-of-information analysis to assess the impacts of additional

studies in a specific decision context. Information opportunities that address uncertainties in the baseline model are considered with respect to the changes they would have on the decision-maker’s preferred decision option and the associated change in net benefits, 76

3-4 Schematic of an analysis of the value of various methodologic opportunities (or “value of methods” analysis) to enhance the risk-assessment process and products, 80

4-1 Illustration of key components evaluated in human health risk assessment, tracking pollutants from environmental release to health effects, 87

4-2 Factors contributing to variability in risk in the population, 102 5-1 Current approach to noncancer and cancer dose-response assessment, 120 5-2 Value of physiologic parameter for three hypothetical populations, illustrating that population

responses depend on milieu of endogenous and exogenous exposures and on vulnerability of population due to health status and other biologic factors, 121

5-3a New conceptual framework for dose-response assessment, 126 5-3b Risk estimation and description under the new conceptual framework for dose-response

assessment, 127 5-4 Linear low-dose response in the population dose-response relationship resulting from background

xenobiotic and endogenous exposures and variable susceptibility in the population, 131 5-5 Nonlinear or threshold low-dose response relationships for individuals and populations, 131 5-6 Linear low-dose response models for individuals and population, 132 5-7 Dose-response relationships involving a continuous effect variable, 132 5-8 New unified process for selecting approach and methods for dose-response assessment for

cancer and noncancer end points involves evaluation of background exposure and population vulnerability to ascertain potential for linearity in dose-response relationship at low doses and to ascertain vulnerable populations for possible assessment, 133

5-9 Population vulnerability distribution, 136 5-10 Examples of conceptual models to describe individual and population dose-response relationships, 137 5-11 Widely differing sensitivity can create a bimodal distribution of risk, 140 5-12 Three example conceptual models lead to different descriptions of dose-response relationship at

individual or population levels, 140 5-13 Baseline airway reactivity as vulnerability factor for allergen-induced respiratory effects expressed

as relative risk, 143 5-14 Effect of asthma-related gene polymorphisms on human vulnerability to asthma, 144 5-15 Dose-response relationship for liver spongiosis in 1,4-dioxane-exposed rats, 145 5-16 Steps in derivation of risk estimates for low-dose nonlinear end points, 148 5-17 Steps to derive population and individual risk estimates, with uncertainty in estimates from animal

data, 154 5-18 Left, AUC for proximate carcinogen in bladder in units of nanograms-minutes simulated for 500

people. Right, simulated fraction bound in bladder, presumed to indicate differences in susceptibility due to PK and physiologic parameters, 157

8-1 A framework for risk-based decision-making that maximizes the utility of risk assessment, 223 9-1 A framework for risk-based decision-making that maximizes the utility of risk assessment, 246 E-1 Community involvement activities at NPL sites, 339 E-2 The framework for ecological risk assessment (Modified from EPA 1992), 350 E-3 Conditions for Regulation Under SDWA 1996, 350

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

Contents

Prepublication Copy xvii

TABLES 3-1 Transition in EPA Human Health Risk-Assessment Characteristics According to EPA, 63 4-1 Examples of Factors Affecting Susceptibility to Effects of Environmental Toxicants, 101 5-1 Potential Approaches to Establish Defaults to Implement the Unified Framework for

Dose-Response Assessment, 161 6-1 Examples of Explicit EPA Default Carcinogen Risk-Assessment Assumptions, 179 6-2 Examples of Explicit EPA Default Noncarcinogen Risk-Assessment Assumptions, 180 6-3 Examples of “Missing” Defaults in EPA “Default” Dose-Response Assessments, 181 7-1 Modified Version of Stressor-Based Cumulative-Risk-Assessment Approach from Menzie et al.

(2007) Oriented Around Discriminating among Risk-Management Options, 204 C-1 Timeline of Selected EPA Risk-Assessment Activities, 257 D-1 Environmental Protection Agency Response to National Research Council Recommendations

of 1983-2006: Policy, Activity, and Practice, 274 E-1 Nine Evaluation Criteria for Superfund Remedial Alternatives, 340 E-2 Examples of Default Exposure Values with Percentiles, 342

Science and Decisions: Advancing Risk Assessment http://books.nap.edu/catalog/12209.html

__MACOSX/._advancing+risk+assessment+full+report.pdf

choosing+pleasures+risk.doc

`Choosing Our Pleasures And Our Poisons: Risk Assessment For The 1980’s’

by William Lowrance

At the root of many contemporary concerns about technology is the question of risk. Most of the controversies over nuclear power, pesticides, and asbestos (to cite but a few examples) focus on the possibility of harm to humans and the environment, the issue of who might be harmed and how much, the matter of responsibility, and the relative merits of various steps to alleviate the risks. As such controversies have become part of the political scene, techniques for risk assessment have emerged and gained in sophistication. In "Choosing Our Pleasures and Our Poisons: Risk Assessment for the 1980s, " William W. Lowrance reviews the state of the art of this important new field. Originally prepared in 1980 for the American Association for the Advancement of Science contribution to the federal government's Five Year Outlook for Science and Technology, Lou Tance's paper explains the concept of risk and shows how quantitative measures of risk are developed and employed. It stresses the need to be explicit in characterizing risks, suggests means of dealing with uncertainties, and places special emphasis on the need to treat risks comparatively so that minor ones do not displace more significant ones on the policy agenda. William W. Lowrance has been a senior fellow and director of the Life Sciences and Public Policy Program at the Rockefeller University in New York City since 1980. Prior to that he held policy, research, and teaching positions at Stanford University, Harvard, the U.S. State Department, and the National Academy of Sciences. A biochemist by training, he holds a Ph. D. from Rockefeller University and is the author of Modern Science and Human Values (Oxford University Press, 1985).

INTRODUCTION It takes only a few highly charged terms to evoke the risk -assessment milieu of the past decade: DDT, the pill, saccharin, "Tris," asbestos, nuclear waste, Three Mile Island, smoking, black lung, Clean Air Act, Delaney clause, recombinant DNA, 2,4,5-T, "Research Mining versus EPA," Teton Dam, DC-10....

This rash of accidents, disruptions and disputes has left the public and its leaders fearful that the world is awfully risky and that, although science can raise warnings, when crucial decisions have to be made, science backs away in uncertainty. Further, there is a feeling that as with medical catalepsy, in which the simultaneous firing of too many nerves draws the body into spasms, the body politic has been drawn into a kind of regulatory catalepsy by too many health scares, too many consumer warnings, too many environmental lawsuits, too many bans, too many reversals. A related complaint is that we are afflicted with excessive government intervention, often of a naive, or trifling or nay saying sort. Among professional analysts as well as members of the public, there is a conviction. that many risk-reduction efforts are disproportionate to the relative social burden of the hazards.

Public apprehensiveness has a number of causes. Is life becoming riskier? Not in any simple sense.... Many classical scourges have been conquered; infants get a healthier start in life; on average people live longer lives than ever before. The historical record of floods, hurricanes, typhoons, tornadoes, earthquakes and other geophysical disasters shows a relatively constant pattern of occurrence over the centuries. I... What we are menaced by now are enormous increases in the physical and temporal scale and complexity of sociotechnical hazards. Of these, the most threatening are risks having low probability and high consequence, such as genetic disaster, nuclear war and global climate change. Too, alarm arises, in an almost paradoxical sense, because science has become so much better at detecting traces of chemicals and rare viruses and at identifying birth defects, diseases and mental stress. Often we know enough to worry but not enough to be able to ameliorate the threat.' Warnings and accusations are amplified by the public media, often with unseemly haste. Worse, scientific hunches are announced as scientific fact, only to have to be withdrawn later. With all this, it would be surprising if the public's sensibilities were not battered....

THE EVOLUTION OF MORTAL AFFLICTIONS In his 1803 Essay on Population Thomas Malthus observed of Jenner's new vaccine- "I have not the slightest doubt that if the introduction of cowpox should extirpate the smallpox, we shall find ... increased mortality of some other disease." This general expectation holds true today if, in addition to disease, we include noninfectious threats. The communicable diseases of smallpox, diphtheria, typhus, cholera, tuberculosis and polio have been conquered. So have scurvy, pellagra and other nutritional deficiency diseases. Infant mortality has dropped dramatically. As the toll from these causes has lessened, mortality has shifted toward degenerative diseases-notably heart disease and cancer-which are attributable either to personal life style or to causative agents in the environment. While the causes of death have changed, the average age of onset of fatal illness has moved higher. Life span has lengthened. Put crudely, we die now of stroke and cancer in part because we live long enough to do so.

Thus at present in the United States the leading cause of death is heart disease, followed by cancer. The rest of mortality is accounted for by other diseases and by accidents, homicide and natural disasters (in that order).2 Within these gross statistics, however, there is great variability by age and socioeconomic status: Motor vehicles and other accidents kill the most children under 14; for black mates between the ages of 15 and 24, homicide is the largest threat; cirrhosis of the liver is the fourth leading cause of death for people between 25 and 64.

In a recent analysis of the prospects for saving lives in this country, James Vaupel developed the 'concept of "early deaths." (The definitional problem is fully treated in his report; for short, early death can be taken to refer to death before the age of 65.) Vaupel concluded:

The statistics indicate that the aggregate social losses due to death are largely attributable to early death and that the losses due to early death are immense, that the early dead suffer an egregious inequality in life-chances compared with those who die in old age, and that non-whites, the poor, and males suffer disproportionately from early death. Furthermore, statistics on the leading causes of death and statistics comparing non-whites and whites, males and females, current mortality with mortality earlier in this country, and the United States with Sweden and other countries suggest that early deaths could be significantly decreased.3

Extrapolation of life expectancy data has led to another provocative observation about survival. Some analysts now speculate that the human species is approaching a "natural" life span limit of about 85 years. These analyses have led James Fries to predict that "the number of very old persons will not increase, that the average period of diminished vigor will decrease, that chronic disease will occupy a smaller proportion of the typical life span, and that the need for medical care in later life will decrease."4

Surely, coming to terms with these trends will lead us as a society to strive less to fend off full-lifetime mortality and to attend more to illness, accidents and quality of life. Among occupational diseases demanding attention, for instance, are the pneumoconioses.- black lung disease, asbestosis and brown lung (textile dust) disease; among the most debilitating, lingering and painful conditions are arthritis, emphysema and allergies; among "life style" diseases, cirrhosis of the liver and the venereal diseases.

IMPROVEMENTS IN ASSESSMENT

Becoming More Comparative

As a society we find ourselves, relative to all previous human confrontation with mortal risk, in the enviable but emotionally unsettling situation of living longer and healthier lives than ever before; of not having to remain ignorant and vaguely apprehensive of hazards but of understanding many of their causes, likelihoods and effects; and of having now accumulated substantial experience in predicting, assessing, reducing, buffering and redressing harm. Blissfulness is prevented by our having too many options. If we still lived only on the margin of survival, we would not have the luxury of worrying about microwaves and hairdryers. If we lacked scientific understanding and the prospect of taking preventative action, we would be more fatalistic about Legionnaires ‘ disease and toxic shock syndrome. If we had not established the hurricane warning network and the national air traffic control system, we would not have to argue about their budgets.

Howard Raiffa made the central analytical point recently in congressional hearings:

We must not pay attention to those voices that say one life is just as precious as 100 lives, or that no amount of money is as important as saving one life. Numbers do count. Such rhetoric leads to emotional, irrational inefficiencies and when life is at stake we should be extremely careful test we fail to save lives that could have easily been saved with the same resources, or lest we force our disadvantaged poor to spend money that they can ill afford in order to gain a measure of safety that they don't want in comparison to their other more pressing needs.5

To proceed in dealing with risks without making comparisons, both of import of threats and of marginal risk reduction effectiveness (and cost-effectiveness) of public programs, makes little sense. Yet surprisingly little sophisticated comparative work has been done.

In studies meant to be illustrative, Bernard Cohen, Richard Wilson and others have assembled catalogues of common risks.6 Cohen and Lee have calculated effects from different hazards upon life expectancy (for people at specified ages). They found that cigarette smoking reduces U.S. male life expectancy by six years on average. Being 30 percent overweight reduces life expectancy by about four years. Motor vehicle accidents cut off 207 days. And assuming that all U.S. electricity came from nuclear power and that the unoptimistic risk estimates published by the Union of Concerned Scientists are correct, nuclear accidents would claim 2 days from the life of an average citizen.... Although these studies are flawed in numerous ways, their most valuable lesson has been to illustrate how difficult it is to reduce complex social phenomena, such as cigarette smoking and nuclear power generation, to single scalar risk rankings.

Stimulated in part by the early contributions of Chauncey Starr, …. . assessors have attempted to compare technological hazard to natural hazard.7 For example, the so-called Rasmussen Report attempted to compare nuclear reactor accident risks to those of meteorite impacts and other natural hazards in order to provide some intuitive groundings The difficulty is that reliable numbers are hard to compute, and because polls have shown that most people, including scientists, do not have a very accurate intuitive sense of the likelihood and magnitude of natural hazards, such grounding may not be very useful anyway.9

The next logical step has been to “to compare the relative impacts various risk-reduction measures make on longevity. Shan Pou Tsai and colleagues, for example, have examined the question of what gains in life expectancy would result if certain major causes of death were partially eliminated. They calculated that for a newborn child, reduction of cardiovascular disease by 30 percent nationally would add 1.98 years to life expectancy at birth; 30 Percent reduction of malignant cancers would add 0.71 years; and 30 percent reduction of motor vehicle accidents would add 0.21 years. If such 30 percent. causative reduction were to exert effect during the working years of 15 to 60, there would be gains of 1.43 years (cardiovascular), 26 years (cancer), and 0. 14 years (motor vehicle accidents). "Even with a scientific breakthrough in combating these causes of death, " the authors concluded, "it appears that future gains in life expectancies for the working ages will not be spectacular.

Obviously the outcome of comparisons is heavily dependent on the way the boundaries of comparison are set.... In calculating the risks of coal, do we count deaths from train wrecks, air pollution or release of radioactive radon from the burning fuel? In assessing nuclear power, do we include terrorist abuse or nuclear weapons proliferation? In appraising solar sources, do we include health effects on copper and glass workers? There is no avoiding such analyses. The problem is to learn how to perform them with technical sophistication and to take due account of all relevant social considerations. Overreaching is hard to avoid. The consolation of most such ambitious studies has been that the process of assessment has itself sharpened the social debate and clarified technical-analytic needs.

That the general public is sophisticated enough to understand and endorse the idea of comparative risk assessment has been demonstrated in such situations as Canvey Island in Britain. Within an area of 15 square miles on that island in the Thames near London are oil refineries, petroleum tanks, ammonia and hydrogen fluoride plants and a liquefied natural gas facility. When a few years ago controversy arose as to whether Canvey's 33,000 people were exposed to unusually high risks, a thorough government inquiry was conducted. Upon deliberation the residents passed a resolution that no further construction be accepted until the overall industrial accident risk on the island had been reduced to the average level for the United Kingdom. But they did not demand that their neighborhood be risk free. 11

That the same toleration for comparative approaches holds in the United States is evident in industrial areas, such as Ohio and New Jersey, where residents are demanding cleanup, but not closing, of industries. Similar moderation led the voters of Maine, an environmentally sensitive state that has had to deal with cold winters but also with proposals of supertanker ports, in their 1980 referendum to vote against measures that would have had the effect of being more restrictive of nuclear power.

If this country is to move toward more "rational" apportionment of risk-reduction and -management efforts, we must assure ourselves that there is reasonable parallel between the burden, in whatever terms, of particular risks and the avidity with which we defend against them, and that programs take into consideration age of onset of harm, degree of debilitation, longevity erosion and cost-effectiveness of ameliorative programs. Before any of this can be done, hazards have to be stated explicitly and goals of hazard reduction agreed upon.

Facing Hazards Explicitly

Comparative approaches are necessarily more quantitative, and they tend to force the revelation of specific consequences. As it dawns on social consciousness that even strict protection inevitably admits some residual harm, even if only by inducing exposure to the hazards of alternatives, little by little public officials have moved toward explicitness.

One of the most widely discussed test cases is that of DES (diethylstilbestrol, the growth hormone sometimes fed to beef cattle). The Food and Drug Administration (FDA) has formally proposed to allow beef producers to use this putatively carcinogenic but economically important agent, if they remove it from feed sufficiently in advance of slaughter that residual DES in marketed beef does not exceed a specified, extremely low concentration. In its proposal the FDA argued that "the acceptable risk level should (1) not significantly increase the human cancer risk and, (2) subject to that constraint, be as high as possible in order to permit the use of carcinogenic animal drugs and food additives as decreed by Congress.... A risk level of 1 in 1 million over a lifetime meets these criteria better than does any other that would differ significantly from it." The agency noted that further reduction "would not significantly increase human protection from cancer."12 This proposal and similar ones are predicated on a conviction that the underlying carcinogen assessments are worst possible case overestimates of human risk. The DES standard is still under discussion. In March 1980 FDA Commissioner Jere E. Goyan stated that he would favor amending the food additives laws so that the chemicals testing out under the level of one chance in a million would be permitted (the Delaney clause prohibits even minute traces of very weakly testing carcinogenic additives-a prohibition honored mostly in the breach, because of its absolutist nature).

One by one, as cases have developed-the 1979 Pinto lawsuit, the national review of earthwork dams, amendment of the Clear Air Act-there has been a tendency to require that an upper bound on the estimated actual hazard be stated.

Specifying Risk-Management Goals

Although industrial and legislative programs usually operate under guidelines mandating "reduction of harm" or "protection of consumers," the degree of reduction or protection is often not specified (except when absolute protection is called for, which, usually being impossible, simply amounts to defaulting). Goal ambiguities may remain even when program objectives are spelled out. Different goals may come into conflict: reducing use of asbestos insulation, in order to protect miners and insulation installers, may have the effect of increasing fire hazard in buildings; forbidding black airmen who are sickle-cell-trait carriers to serve as Air Force Pilots, to avoid the possibility of their becoming functionally impaired under emergency oxygen loss, conflicts with equal opportunity goals.

A recent RAND Corporation study for the Department of Energy (DOE), Issues and Problems in Inferring a Level of Acceptable Risk, lists types of risk-reduction goals that can be considered, such as minimization of maximum accident consequences, minimization of probability of most probable accident and so on. After describing ways in which goal choices can make a difference to programs, the report urges that "DOE and other agencies need to be self-aware in specifying risk-reduction goals, as well as in relating them to goals of other agencies and interested parties, and understanding their implications for the choice of energy alternatives."13

Skeptics may be tempted to dismiss the topic, saying that we in this country do not have a consensus on social goals. Rebuttal to that too- simple dismissal is evidenced, for example, in the way our medical X ray protection practices, which are the result of decades of reassessment and improvement by industry, medicine and government, pursue goals.- minimization of probability of damage (by decrease in frequency of use of diagnostic X rays, compensated for by more sensitive films), minimization of potentially irreversible damage to the human gene pool (special protection of gonads) and minimization of threat to infants in utero (again, special protection). The typically American goal of helping disadvantaged citizens underlies special health programs for minority groups. The goal of preserving maximum consumer choice can be seen as a goal of food quality programs.

Setting goals is not impossible, but setting realistically attainable goals is not easy. It is imperative that programs be tailored to goals more precise than "protection of all Americans against all harm."

Weighing Risks in Context with Benefits and Costs

All decisions, indirectly or directly, rely on judgments of the sort Benjamin Franklin referred to as "prudential algebra." Under the Toxic Substances Control Act, the Environmental Protection Agency (EPA) must protect the public against "unreasonable risk of injury"; under the stationary-sources provisions of the Clean Air Act, it must ensure "an ample margin of safety"; under the Safe Drinking Water Act, it must protect the public "to the extent feasible ... (taking costs into consideration)." "Unreasonable," "ample" and "feasible" are not defined in these laws. For the EPA the question is not whether analysis but what form of analysis, taking what considerations into account. For all such risk reduction regimes, the day has passed when benefits and costs could be ignored.

Every segment of industry and government-food, energy, transportation-has to ask:

•Are there ways to take benefits and costs into consideration along with risk? Do existing policy and managerial rules allow consideration of all such factors? Should they?

•Which methodological approaches (cost-benefit analysis, decision theory, cost-effectiveness analysis, etc.) are appropriate?

•How should secondary, indirect and intangible effects be taken into consideration?

•Are formal, explicit, published analyses required to form the basis of decision, or should they be used as informational background only?

•What are the procedural rules by which definitions, analytic boundaries and conceptual assumptions are established?

•Should those reviewing a technological option be required to review the attributes of alternatives also?

After a decade of concentrating on the negative side of the ledger, society is now trying to learn how to measure benefits. The National Academy of Sciences (NAS) 1977 study of ionizing radiation ("BEIR 11") struggled with the issue of how to appraise the benefits of such applications as medical X rayS.14 Its 1979 food safety policy report analyzed the benefits of saccharin and of food-safety policies regarding mercury, nitrites and aflatoxin (in peanut butter)15 and its 1980 report, `Regulating Pesticides’, described the methods available for estimating marginal gains in crop yield and benefit expected from a candidate pesticide. 16

Several methods, usually referred to in shorthand as "risk-benefit" or "cost-benefit analysis," are available for constructing a balance sheet of desirable and undesirable attributes. Analysis is thus a problem of handicapping what will happen (the odds of a destructive flood, the probable incidence of a disease) and comparing quantities that are rarely expressible in common-denominator terms (social cost of lives shortened, benefits of production, risks of genetic mutation).

With a few well-defined projects, for which goals and constraints are agreed upon by the major affected parties, for which health and environmental risks, costs and benefits are well known and understood (not only in magnitude but in social distribution, over both the near and long term), risk-benefit accounting has proven itself useful. Under such rare circumstances of certainty, commonsensical estimates as well as more formal analyses derived from operations research are applicable. The latter tend to be favored by specialists, technical or otherwise, who have been given a specific task to accomplish (the Army Corps of Engineers has pioneered in their use). The occasional "successful" application of such techniques-and, one suspects, also the all-embracing ring of their title-tempts legislators, administrators, managers and judges to call for their use.

The griefs of analysis could fill a large set of books. Most reviews conclude that such approaches are very use" for structuring discussion but are less useful, or even subject to misuse, when granted formal, legalistic weight. In their Primer f6r Policy Analysis, Edith Stokey and Richard Zeckhauser warned that:

Benefit-cost analysis is especially vulnerable to misapplication through carelessness, naiveté, or outright deception. The techniques are potentially dangerous to the extent that they convey an aura of precision and objectivity. Logically they can be no more precise than the assumptions and valuations that they employ; frequently, through the compounding of errors, they may be less so. Deception is quite a different matter, involving submerged assumptions, unfairly chosen valuations, and purposeful misestimates. Bureaucratic agencies, for example, have powerful incentives to underestimate the costs of proposed projects. Any procedure for making policy choices, from divine guidance to computer algorithms, can be manipulated unfairly. 17

These and other critics respond to their own complaint by acknowledging that "prudential algebra" of one form or another must be resorted to, nevertheless.

All analytic approaches have difficulty with scientific uncertainties, with fair and full description of societal problems, with predicting all possible consequences, with placing a "price" on human life and environmental goods, with taking into account intangibles and amenities in general and with assessing the social costs of opportunities precluded. 18...

Furthermore, formal analysis is still helpless to accommodate many major effects: the weapons-proliferation and terrorist risks of the spread of civilian nuclear power, the highly touted and ambipotent benefits and risks of recombinant DNA development, the opportunity costs from undue conservativeness in regulation of contraceptive and pharmaceutical development.

Defining "Negligible" and "Intolerable" and Setting Priorities

A disturbing feature of the 1960s and 1970s was that as each sector of manufacturing, or municipal governance, or research or purchasing found itself having to confront risk problems, each had to develop its own approach and work through hearings, scientific studies, economic reviews, lawsuits and insurance disputes. The social teaming process was, unavoidably, painful. So were the disruption and unpredictability caused by the lack of defensible priorities. Industries and agencies found themselves so distracted by disputes over sensational cases that they could hardly pursue their main tasks, even if their charter was to reduce major risks: Neither "major" nor "minor" had been defined. Expressed in a metaphor of the time, smoldering barn fires had to be neglected while brushfires were fought.

Chastening has been accomplished. Now the challenge is to develop ways of keeping priorities clear: to avoid frittering away worry-capital on very small hazards, to prohibit unbearably large hazards and to concentrate decision-making attention on problems that affect large numbers of people in important ways. This admonition may appear an obvious one, but our failure to protect appropriate priorities is just what has set us up for the regulatory "overload" and disproportionateness we now labor under.

This concern was expressed in the 1980 NAS report, Regulating Pesticides:

A serious flaw in the current procedure is that those compounds that receive the most publicity or pressure-group attention may not necessarily be those that present the greatest public health or environmental hazards. The current procedure does not provide for a broad comparison of the hazards posed by the large number of registered pesticides. At the same time, outside pressures to regulate a specific compound rarely arise from careful evaluation of comparative risks of alternative pesticides. To the extent that external pressures are influential in determining the order in which the [Office of Pesticides Programs] evaluates compounds, the consequence may well be that considerable resources are devoted to regulation of minor, low-risk compounds while important high-risk ones remain unreviewed for periods longer than would otherwise be the case . 19

Naturally, regulatory agencies do try to apply their most vigorous attention to the most important issues, but their problem is to set protectable priorities (ones that are buffered from sporadic undermining) so that all parties involved know the analytic and legal agenda and can allocate resources accordingly. OSHA has tried to do this with occupational carcinogens, as has EPA with chemicals regulated under the Toxic Substances Control Act. The new National Toxicology Program is taking over some of the priority-setting tasks and will try to rationalize them across agency lines. The Consumer Product Safety Commission bases its priorities in part on a "frequency- severity index" derived from a computerized sampling system of hospital emergency-room admissions.

"Intolerable" and "unacceptable" are being invested with real-world connotations, as are "negligible" and "insignificant." These boundary- setting adjectives gain meaning in two ways- as experts, insurers and others rank hazards in hierarchies by severity, incidence and overall social exposure (hazards at the top and bottom of lists thus becoming obvious candidates for prohibition or acceptance); and as public opinion, lawsuits and so on indicate endorsement of the ranking. This helps administrators and managers allocate attention to the difficult cases in the middle....

In a striking case recently, the FDA approved the hair-dye chemical lead acetate. While acknowledging that in high doses the material is carcinogenic to rodents, the agency concluded that human exposure is so small, especially relative to overall lead intake, as not to warrant prohibition. 20

Risk ceilings also can be established. In this country and many others polychlorinated biphenyls (PCBS) have been banned from commerce because their carcinogenic potency is judged to be absolutely intolerable. From time to time, high-technology projects have been vetoed because their risks were unthinkably high- some macro- engineering modifications of the environment and certain potentially disastrous recombinant DNA experiments are landmark examples. The issue may not only be whether the hazards are actuarially high, but whether the threat would have an intolerably disruptive effect, physically or psychologically, on the fabric of society....

Seeking Accommodation Between Technical and Lay Perceptions

It is evident that "the public" often views risks differently from the way technical analysts do. Of course, consensus is also rare, even within relatively closed circles of experts.)

From what do these differences of opinion stem? First, science itself is, in effect, simply a matter of "voting"; the scientifically "true" is no more than what scientists endorse to be true. Empirical knowledge is developed systematically within the scientific community, subject to criteria of repeatability, controlled observation, statistical significance, openness and the other guides of western science. By itself, procedure guarantees nothing, though. Good science is science that "works"- science that can predict with consistency and generality and accuracy what will happen in the physical and social world. The weighing of facts remains subjective; perfect objectivity is a myth.

And second, judgments of hazards involve consideration not only of "size" of risks-likelihood and magnitude-but also of social value.21 This, of course, leaves much room for disagreement.

Researchers have speculated that people's opinions about risks depend on many biasing factors, such as voluntariness of exposure, frequency of occurrence, amenability to personal control, reversibility, immediacy, bizarreness, catastrophic nature and so on.22

Social scientists such as Paul Slovic, Baruch Fischhoff and Sarah Lichtenstein have used polling techniques to survey risk perceptions and risk-taking proclivities. What they find, to neither their surprise nor ours, is that people have different perceptual biases. This research has concluded that human beings' brains, whether expert or lay, get overloaded with risk information and have trouble comparing risks; that the media accentuate social reverberations in risk disputes; and that, in essence, people believe what they want to believe. Person- in-the-street interviews of technical people show them to be not much better than nontechnical people at guessing, for example, how many fatalities are incurred annually from tornadoes, contraceptives or lawnmowers. 23

Many of these polling studies are open to criticism. They suffer from the usual shortcomings of questionnaire design and the generic weaknesses of polling. Often they ask about only a single hazard at a time, which, by failing to foster or force comparison and by allowing people to express self-contradictory views, provides little guidance for policymaking. They force people artificially to break down their views into components. And these studies are vulnerable to being assumed (not necessarily by their authors) to imply findings about "the public,” when in fact most of them have dealt with only small population samples. …

In the risk-assessment domain, as in others, we are being forced to realize that "the public" is a very elusive construct. No one per- son or group of people fully represents, or is representative of, all of our citizenry; and the "organized public" remains small and keeps changing in composition and opinion. For this reason, and others, the notion of "public participation" lacks conceptual shape. To oppose closed bureaucratic proceedings is usually legitimate, but it is a lot harder to devise proceedings that are not only open to the affected policy but that encourage extensive "public" participation without just opening channels for special-interest lobbying. A recent Organization for Economic Cooperation and Development study of public participation, entitled Technology on Trial, concluded: "the general thrust of participatory demand would appear to be for a greater degree of public accountability; freer public access to technical information; more timely consultation on policy options; a more holistic approach to the

assessment of impacts: all of which amounts, of course, to more direct public participation in the exercise of decision-making Power.” 24

In recent years both governmental and nongovernmental bodies have been taking steps to seek accommodation between lay perceptions and technical-analytic ones. 25 Regulatory agencies have opened up their proceedings and have solicited public input. Professional organizations have explored perceptual issues. In 1979 the National Council on Radiation Protection and Measurement held a symposium resulting in a volume entitled `Perceptions of Risk’. 26

If an attitudinal bias emerges, it can be incorporated into standards. In recognition of the public's extraordinary concern about catastrophic potential (as opposed to diffuse chronic risks) of nuclear reactors, for example, industry and its regulators have incorporated "risk aversiveness," or disproportionate conservatism, into reactor safeguarcis.27...

It is worth surmising that what is under perceptual dispute in many cases is not only the hazard itself but the social "management" of it. Nowhere has this been more bluntly evidenced than in the overall conclusion of the President's Commission on the Accident at Three Mile Island: "To prevent nuclear accidents as serious as Three Mile Island, fundamental changes will be necessary in the organization, procedures, and practices and-above all-in the attitudes of the Nuclear Regulatory Commission and, to the extent that the institutions we investigated are typical, of the nuclear industry."28 Too, one suspects that risk opinions often may in effect be proxies for more deeply seated opinions about corporate bigness, or bureaucratic inaction or erosion of personal control.

INSTITUTIONAL ATTENTION

Congressional Actions

As though swatting at swarms of hazards on all sides, during the 1970s the Congress passed, inter alia, the Consumer Product Safety Act, the Fire Prevention and Control Act, the Occupational Safety and Health Act, the Federal Water Pollution Control Act, the Toxic Substances Control Act, the Mine Safety and Health Act, the (aircraft) Noise Control Act, the Federal Environmental Pesticide Act, the National

Earthquake Hazards Reduction Act, the Medical Devices Amendment to the Food, Drug, and Cosmetic Act, the Safe Drinking Water Act, the Resource Conservation and Recovery Act and various Clean Air Act amendments. To ensure independence of control, Congress split off the Nuclear Regulatory Commission from the old Atomic Energy Commission. And it established the Environmental Protection Agency, the Occupational Safety and Health Administration, the Consumer Product Safety Commission, the National Fire Prevention and Control Administration and the Federal Emergency Management Agency to administer all the new laws.

The effect of this legislative crusade has been to bring tens of thousands of hazards into regulatory frameworks of many kinds, based on science, medicine, engineering, law and economics, that were- and still are-inadequate bases for decision.

The Congress has chosen a variety of roles for itself in risk assessment. It has established the regulatory agencies and overseen their work. With some issues, such as automobile emissions, it has insisted on reviewing the scientific and economic evidence in detail and on itself setting primary standards. With others, such as the arcane questions of recombinant DNA research, it has held hearings to establish a record but has refrained from instituting strong control. Occasionally, in response to constituent pressure or political opportunity, it has intervened precipitously in regulatory action, as it has repeatedly done with saccharin, directing the FDA to stay an action or requesting the NAS to conduct another study. In emergencies it has held high-level inquiries, as it did during the Three Mile Island accident.

Recently the Office of Technology Assessment, the General Accounting Office, and the Congressional Research Service have all gotten more involved in preparing risk-related reports for the Congress. Congressman Don Ritter and others have proposed man- dating that cost-benefit analysis be used as the basis for regulatory action.... Prompted by such flaps as that over the questionable studies of health risks at Love Canal, legislators are considering establishing guidelines for scientific peer review of assessments used in regulation. Congressional concern over risk issues remains high, but it tends to focus on individual hazards rather than on a comparative high-risk-reduction agenda, and it tends to favor regulation as its best instrument.

Administration Actions

Various Executive Branch sagas in risk decision making have been described elsewhere and will not be reviewed here. We should, however, notice several trends that go beyond the straightforward execution of regulatory mandates.

There is some movement toward interagency coordination of regulatory actions. The complexity of the administrative task is illustrated by the fact that the Interagency Review Group on Nuclear Waste Management had to be constituted from 14 major entities of government (the Departments of Commerce, Energy, Interior, State and Transportation; National Aeronautics and Space Administration, Arms Control and Disarmament Agency, Environmental Protection Agency, Office of Management and Budget, Council on Environmental Quality, Office of Science and Technology Policy, Office of Domes- tic Affairs and Policy, National Security Council and Nuclear Regulatory Commission).29 The Interagency Regulatory Liaison Group (Consumer Product Safety Commission, Environmental Protection Agency, Food and Drug Administration, Occupational Safety and Health Administration and Department of Agriculture) has developed coordinated guidelines on carcinogenicity assessment.30 A National Toxicology Program has been established to serve the needs of a number of agencies.

Fundamental research in support of regulatory work may be improving: The National Institutes of Health have become more involved in such matters as development of reliable and practical screening tests for carcinogens; the National Science Foundation now sponsors risk-related policy studies; the National Bureau of Standards conducts fire research for the benefit of many agencies. How to marshall such support effectively is still a challenge: The basic research agencies don't have specific mission mandates, and the regulatory agencies lack strong fundamental research capabilities....

Court Actions

Thousands of tort cases are heard every year. For the present review, what is important are the ongoing debates over the role of the courts and the landmark decisions handed down by the high courts. One respected view of the role of the judiciary is that championed by judge David Bazelon- "Courts cannot second-guess the decisions made by those who, by virtue of their expertise or their political accountability, have been entrusted with ultimate decisions. But courts can and have played a critical role in fostering the kind of dialogue and reflection that can improve the quality of those decisions. "31 Others disagree, believing that courts should be free to review the substantive evidence and logic of assessments and decisions. The extent of judicial intrusion into agency decision making will remain an issue.

Recent years have seen the courts interpreting legislative mandates (as to whether, for instance, regulation under the Clean Air Act must consider costs, or whether the FDA properly interpreted its mandate in banning laetrile) and refereeing territorial disputes between agencies. A crucial issue that continues to work its way up to the Supreme Court relates to the imperative for cost-benefit analysis in regulatory decisions. The recent case of Industrial Union Department, AFL-CIO versus American Petroleum Institute sidestepped the issue of whether OSHA must, under its statutes, base' its decision-in this case, over whether to tighten occupational exposure limits for benzene from 10 parts per million to 1 part per million-on formal, explicit, published cost-benefit analyses, the issue that many observers hoped the court would address.32 The justices have, however, agreed to hear an analogous case, on cotton dust. The legislative background from which the Supreme Court has to work does not provide much guidance.

Nongovernmental Actions

Several recent developments exemplify the increasingly collective initiatives being taken by nongovernmental bodies. An impressive contribution has been made by the Food Safety Council, a non-profit coalition of industrial, consumerist and other members, which has developed and published a thorough review of the technical problems associated with food risk assessment and made proposals that are now under consideration by regulatory and other bodies.33 The American Industrial Health Council, a coalition of 140 companies and 80 trade associations, has developed concerted positions on regulatory issues and is now proposing structural and procedural reforms.34 In the aftermath of the Three Mile Island accident, the country's electric utilities and nuclear industry pooled their interests and established a Nuclear Safety Analysis Center, associated with the Electric Power Research Institute, to serve as an industry-wide reactor performance clearinghouse. Some 35 major chemical firms have recently established the Chemical Industry Institute of Toxicology, a research center charged with performing state-of-the-art toxicological research and assessment of large-volume commodity chemicals (not proprietary products) for the benefit of the industry as a whole. The major U.S. automobile and truck companies have joined the EPA in establishing a Health Effects Institute to study the effects of motor vehicle pollution. 35

It is not yet -possible to evaluate the promise of these new institutions. They deserve watching because they typify efforts to develop techniques, procedures, databases and focal centers for risk assessment outside of government. The question will be whether the work they produce is of high technical quality, whether they develop reputations of integrity and whether government and the courts can effectively accommodate the work of these hybrid institutions as alternatives to direct regulation and government sponsored assessment.

SCIENTIFIC INTEGRITY AND AUTHORITY

Serious criticism is currently being leveled at the manner and quality with which scientific analysis is brought to bear on public hazards. Not to be interpreted as disaffection with science per se, this dismay reflects confidence that science can indeed help assess these problems, if it is property applied.

Proposals are gathering for establishment of central authority structures to which technical disputes can be appealed. For example, the New York governors' panel (chaired by Lewis Thomas) formed to review the Love Canal fiasco found that “only further questions and debates on scientific credibility have been the result" of the "inadequate research designs" and "inadequate intergovernmental coordination and cooperation in the design and implementation of health effects studies" at the dump; as a remedy it recommended establishment of a Scientific Advisory Panel responsible to the governor.36 Editorials have appeared in Science and elsewhere calling for reincarnation of the President's Science Advisory Committee to referee such disputes.... In somewhat the same vein, the American Industrial Health Council has urged Congress to establish a Science Panel:

AIHC advocates that in the development of carcinogen and other federal chronic health control policies scientific determinations should be made separate from regulatory considerations and that such determinations, assessing the most probable human risk should be made by the best scientists available following a review of all relevant data. These determinations should be made by a Panel of eminent scientists located centrally somewhere within government or elsewhere as appropriate but separate from the regulatory agencies whose actions would be affected by the determinations. 37

Two questions must be asked of such proposals: whether "scientific and technical determinations" can legitimately be separated from "political and social determinations" and whether centralization of authority assures higher quality science.

To the first the answer is probably, yes, to a considerable extent, as long as it is understood that the very process of defining the problem is subjective and that scientific assessments usually have to be conducted iteratively. For example, to view the problem of liquefied natural gas facilities as one of time-averaged risk is different from worrying about the potentially massive social disruption one large accident could cause. Complex issues, such as energy policy, have to go many rounds of assessment, criticism, redefinition and reassessment.

To the second question, the answer is that communal scientific assessments do tend to gain critical analytic strength and social legitimacy over assessments made by individuals alone, but that plural- ism and variety within the scientific community should be encouraged: recruiting more skilled policy-analytic scientists and engineers in industry, government and other organizations; appointing able advisory panels to many different administrative, legislative and managerial bodies; upgrading assessment work in academies, professional societies and trade organizations; and so on. Pluralism remains an essential safeguard against narrowness. Centralization and consistency are not always good in themselves. Besides, high-level bodies will always be limited to handling only a few contentious issues at a time. What they can do is raise warning flags about hazardous situations, draw attention to suspect scientific studies and help set the national agenda of assessment.

One of the more encouraging developments of the last few years has been a willingness of technical people, acting as professional communities, to review major assessments. When the original "Rasmussen Report" on reactor safety was issued, for example, it was subjected to detailed critique by a panel of the American Physical Society, by an ad hoc review group (the "Lewis Panel") chartered by the Nuclear Regulatory Commission, by the Union of Concerned Scientists and by others. Currently the Society of Toxicology is reviewing the controversial "ED-01" effective-carcinogen-dose experiment performed by the National Center for Toxicological Research....

RECOMMENDATIONS

1. The overall urging of this essay is that bodies responsible for appraising public risk ask of their assessment efforts:

*Are risks, benefits and costs characterized as explicitly as possible?

*Are uncertainties and intangibles acknowledged and, where possible,

estimated?

*Are programs oriented to agreed-upon societal goals?

*Do procedures guarantee that high-quality technical evidence is made

available and used as the basis for decision?

*Are risks examined in a property comparative context along with benefits

and costs?

*Are precautions taken to prevent minor hazards from displacing larger

ones on the protection agenda?

*Are the formality and legal bindingness of the analytic base appropriate?

2. Excerpts of well regarded risk-assessment studies should be collected and published with commentary. (The NAS food safety study published several examples, and the NAS current review of some of its past projects-the "Kates study"-will provide more.) Critique should be made not only of analytic methodology but also of how boundaries of assessment were set, how assessors were chosen, how conflicts-of- interest and biases were dealt with, how findings were expressed and how the study groups maintained their relationships with patrons and clients.

The causal connection between environment and health deserves continued investigation. As part of this, baseline surveys like the "LaLonde Report" (Health of Canadians) or the 1980 California Health Plan should be developed for the United States; this would be an extension of the 1979 Report of the U.S. Surgeon General on Health Promotion and Disease Prevention.38 Then those determinants of health that are amenable to environmental influence should be evaluated.

4. The Office of Management and Budget, the Congressional Bud, get Office or others might direct or commission comparative evaluations of the marginal longevity gains and other benefits from the key regulatory programs.

5. Evaluation should be made of such longstanding risk- management regimes as food inspection programs, fire-prevention pro- visions of building codes, flood plains insurance, black lung insurance and the like, asking whether they accomplish their risk-spreading or risk-reduction goals.

6. As the nation contemplates deregulation, sectoral net-assessment of regulatory policies should be conducted and reviewed. Alternatives to regulation should be examined, especially hybrid nongovernmental- governmental approaches.39 In this regard the experiences of other countries, such as Sweden's in food safety, should be reviewed.

7. High-level scientific leadership needs continual renewal. One function of an upgraded White House scientific advisory body should be to identify major risk issues needing attention (such as, for example, the underattended issues cited at the end of this paper). This body, or other groups, should consider setting up a watchdog commission like the United Kingdom's Advisory Committee on Major Hazards to lead in the anticipation and assessment of important, long-term hazards.

8. There are many specific research needs, ranging from toxicology to policy analysis. Broad topics deserving attention include:

• Evaluation of the overall predictive usefulness of the toxicological gauntlet through which chemical products now are required to be run.40

• Improvement of epidemiology as an analytic complement to toxicological testing and continued development of the necessary databases.

• Refinement and comparison of such analytic techniques as cost- benefit analysis, decision theory cost-effectiveness analysis.

• Evaluation of the validity of fault-tree and event tree analysis as applied to nuclear reactors and other engineered structures.41

• Investigation of ways in which human error (maintenance error, operation error, emergency-response error) can be taken into account in probabilistic assessment of technological systems.

NOTES 1. Ian Burton, Robert W. Kates, and Gilbert F. White, The Environment as Hazard (New York- Oxford University Press, 1978). 2. National Safety Council, Accident Facts (National Safety Council, 425 North Michigan Avenue, Chicago, Ill. 60611, 1980). 3. James W. Vaupel, "The Prospects for Saving Lives- A Policy Analysis," printed as pp. 44-199 of the U.S. House of Representatives, Subcommittee on Science, Research and Techno- logy (of the Committee on Science and Technology), Hearings on Comparative Risk Assessment, Ninety-Sixth Congress, Second Session (14-15 May 1980). 4. James F. Fries, "Aging, Natural Death, and the Compression of Morbidity," New England journal of Medicine, vol. 303 (1980), pp. 130-35. 5. U.S. House of Representatives, Committee on Science and Technology, Subcommittee on Science, Research and Techno- logy, Hearings on Comparative Risk Assessment, Ninety-Sixth Congress, Second Session, (14-15 May 1980). 6. Bernard L. Cohen and I-Sing Lee, "A Catalog of Risks," Health Physics, vot. 36 (1979), pp. 707-22; Richard Wilson, "Analyzing the Daily Risks of Life," Technology Review (February 1979), pp. 41-46. 7. Chauncey Staff, Richard Rudman and Chris Whipple, "Philosophical Basis for Risk Analysis," Annual Review of Energy, vol. 1 (1976), pp. 629-62. 8. N. Rasmussen, et al.., Reactor Safety Study: An Assessment of Accident Risks in U.S. Commercial Nuclear Power Plants (Washington, D.C.- Nuclear Regulatory Commission, 1975), document number WASH-1400 (NUREG-75/014). 9. Paul Slovic, Baruch Fischhoff and Sarah Lichtenstein, "Rating the Risks," Environment, vol. 21 (1979), pp. 14ff. 10. Shan Pou Tsai, Eun Sul Lee and Robert J. Hardy, "The Effect of Reduction in Leading Causes of Death- Potential Gains in Life Expectancy," American journal of Public Health, vol. 68 (1978), pp. 966-71. See also Nathan Keyfitz, "What Difference Would It Make If Cancer Were Eliminated? An Examination of the Taeuber Paradox," Demography, vol. 14 (1977), pp. 411-18. 1 1. U. K. Health and Safety Executive, Convey: An Investigation of

Potential Hazards from Operations in the Canvey Island/Thurrock Area (London- Her Majesty's Stationery Office, 1978). 12. U.S. Food and Drug Administration, "Chemical Compounds in Food-Producing Animals: Criteria and Procedures for Evaluating Assays for Carcinogenic Residues," Federal Register, Vol. 44 (1979), pp. 17070-114. 13. Steven L. Salem, Kenneth A. Solomon and Michael S. Yes'leyl Issues and Problems in Inferring a Level of Acceptable Risk, Report R-2561-DOE (Santa Monica, Calif. RAND Corporation, 1980). 14. National Academy of Sciences National Research Council, Advisory Committee on the Biological Effects of Ionizing Radiation, Considerations of Health Benefit-Cost Analysis for Activities Involving Ionizing Radiation Exposure and Alternatives (Washing- ton, D.C.: National Academy of Sciences, 1977). 15. National Academy of Sciences National Research Council, Committee for a Study on Saccharin and Food Safety Policy, Food Safety Policy: Scientific and Societal Considerations (Washington, D.C.-. National Academy of Sciences, 1979). 16. National Academy of Sciences National Research Council, Committee on Prototype Explicit Analyses for Pesticides, Regulating Pesticides (Washington, D.C.: National Academy of Sciences, 1980). 17. Edith Stokey and Richard Zeckhauser, A Primer for Policy Analysis (New York: W. W. Norton, 1978). 18. David Okrent and Chris Whipple, "An Approach to Societal Risk Acceptance Criteria and Risk Management," #UCLA- ENG-7746 (Los Angeles: UCLA School of Engineering and Applied Science, June 1977); Baruch Fischhoff, "Cost Benefit Analysis and the Art of Motorcycle Maintenance," Policy Sciences, Vol. 8 (1977), pp. 177-202; Dan Litai, "A Risk Comparison Methodology for the Assessment of Acceptable Risk" (Ph.D. dissertation, Massachusetts Institute of Technology, 1980); David Okrent and Chris Whipple An on Social Risk," Science, Vol. 208 (25 April 1980), pp. 372-75; Chauncey Starr and Chris Whipple, "Risks of Risk Decisions," Science, Vol. 208 (6 June 1980), pp. 1114-19. 19. National Academy of Sciences/National Research Council, Committee on Prototype Explicit Analyses for Pesticides, Regulating Pesticides (Washington, D.C.- National Academy of Sciences, 1980). 20. U.S. Food and Drug Administration, "Lead Acetate: Listing as a Color Additive in Cosmetics that Color the Hair on the Scalp," Federal Register, vol. 45 (1980), pp. 72112-18. The decision was lauded in a New York Times editorial, 9 November 1980. 21. William W. Lowrance, Of Acceptable Risk: Science and the Determination of Safety (Los Altos, Calif.: William Kaufmann, Inc., 1978); and William W. Lowrance, "The Nature of Risk," in Richard C. Schwing and Walter A. Albers, eds., Societal Risk Assessment: How Safe Is Safe Enough? (New York: Plenum Press, 1980), pp. 5-14. 22. Paul Slovic, Baruch Fischhoff and Sarah Lichtenstein, "Facts and Fears: Understanding Perceived Risk," in Schwing and Albers, Societal Risk Assessment, pp. 181-216; Charles Viek and, Pieteran Stallen, "Rational and Personal Aspects of Risk," Acta Psychologica, vol. 45 (1980), pp. 273-300. 23. Baruch Fischhoff, Paul Slovic, Sarah Lichtenstein, Stephen Read and Barbara Combs, "How Safe Is Safe Enough? A Psychometric Study of Attitudes Toward Technological Risks and Benefits," Policy Sciences, vol. 9 (1978), pp. 127-52; "Labile Values: A Challenge for Risk Assessment," Society, Technology and Risk Assessment, Jobs; Conrad, ed. (New York: Academic Press, 1980), pp. 57-66. 24. Organization for Economic Cooperation and Development, Technology on Trial: Public Participation in Decision-Making Related to Science and Technology (Paris- OECD, 1979). 25. Nancy E. Abrams and Joel Primack, "Helping the Public Decide: The Case of Radioactive Waste Management," Environment, vol. 22 (April 1980), pp. 14ff. 26. Perceptions of Risk (Washington, D.C.: National Council on Radiation Protection and Measurement, 15 March 1980). 27. J. M. Griesmeyer, M. Simpson; and D. Okrent, The Use of Risk Aversion in Risk Acceptance Criteria, #UCLA-ENG-7970 (Los Angeles.- UCLA School of Engineering and Applied Science, October 1979). 28. President's Commission on the Accident at Three Mile Island, The Need for Change: The Legacy of TMI (Washington, D.C: U.S. Government Printing Office, 1979).

29. U.S. Interagency Review Group on Nuclear Waste Management, Report to the President (Washington, D.C.: U.S. Department of Energy, 1979). 30. U.S. Interagency Regulatory Liaison Group, "Scientific Bases for Identification of Potential Carcinogens and Estimation of Their Risks," Journal of the National Cancer Institute, Vol. 63 (1979), pp. 241-68. 31. David L. Bazelon, "Risk and Responsibility," Science, vol. 205 (1979), pp. .277-80. 32. Industrial Union Department, AFL-CIO versus American Petroleum Institute (U.S. Supreme Court, decided 2 July 1980). Described in R. Jeffrey Smith, "A Light Rein Falls on OSHA," Science, vol. 209 (1980), pp. 567-68. 33. Proposed System for Food Safety Assessment: Final Report of the Scientific Committee of the Food Safety Council (Washington, D.C.: Food Safety Council, 1980). 34. AIHC Recommended Alternatives to OSHA's Generic Carcinogen Proposal (Scarsdale, N.Y.: American Industrial Health Council, 24 February 1978, OSHA Docket No. H-090). 35. Philip Shabecoff, "Health Institute to Study Motor Vehicle Emissions," New York Times, 12 December 1980. 36. Governor of New York, Panel to Review Scientific Studies and the Development of Public Policy on Problems Resulting from Hazardous Waste, Report (8 October 1980). 37. "AIHC Proposal for a Science Panel" (Scarsdale, N.Y.: American Industrial Health Council, 26 March 1980). 38. Marc LaLonde, A New Perspective on the Health of Canadians (Ottawa: Health and Welfare Canada Working Document, 1975); Office of Statewide Health Planning and Development, California State Health Plan 1980-85 (Sacramento, 1980); and U.S. Surgeon General, Healthy People: The Surgeon General's Report on Health Promotion and Disease Prevention, 1979, U.S. Department of Health, Education and Welfare Publication No. 79-55071 (Washington, D.C., 1979). 39. Michael S. Baram, Alternatives to Regulation for Managing Risks to Health, Safety, and Environment (a report to the Ford Foundation from the Program on Government Regulation, Franklin Pierce Law Center, White Street, Concord, N.H., I September 1980). 40. U.S. Congress, Office of Technology Assessment, Assessment

of Technologies for Determining Cancer Risk from the Environment (Washington, D.C., 1981). 41. Issac Levi, "A Brief Sermon on Assessing Accident Risks in U.S. Commercial Nuclear Power Plants," The Enterprise of Knowledge: An Essay on Knowledge, Credal Possibility, and Chance (Cambridge, M.I.T. Press, 1980). Also, enlightening analyses were presented in staff reports to the President's Com- mission on Three Mile Island.

__MACOSX/._choosing+pleasures+risk.doc

env+risk+yucca+mt.pdf

ENVIRONMENTAL RISK AND THE IRON TRIANGLE: THE CASE OF YUCCA MOUNTAIN

Kristin S. Shrader-Frechette

Abstract: Despite significant scientific uncertainties and strong public op- position, there appears to be an "iron triangle" of industry, government, and consultants/contractors promoting the siting of the world's first per- manent geological repository for high-level nuclear waste and spent fuel, proposed for Yucca Mountain, Nevada. Arguing that representatives of this iron triangle have ignored important epistemological and ethical difficulties with the proposed facility, I conclude that the business cli- mate surrounding this triangle appears to leave little room for considera- tion of ethical issues related to public safety, environmental welfare, and citizen consent to risk. If my analysis of the Yucca Mountain case is correct and typical, then some of the most pressing questions of busi- ness ethics may concern how to break the iron triangle or, at least, how to expand it into a quadrilateral that includes the public.

1. Introduction

IN late 1991 someone leaked a confidential letter written by Allen Keesler, President of Florida Power and Chair of the utility industry's American

Committee on Radwaste Disposal. Keesler's letter to other US utility execu- tives revealed that nuclear utilities in the US were about to begin a $9 million "advertising blitz in Nevada designed to overcome its resistance to serving as the dumping ground for other states' nuclear wastes." Recognizing that the profits of nuclear utilities are tied to the existence of radwaste repositories, Keesler was eager to promote the proposed Nevada repository. He also re- vealed, in his letter to the other nuclear-utility executives, that the federal waste-disposal program being run by the US Department of Energy (DOE) is progressing only "because of the active support, guidance, and involvement of our industry" in re-educating the people of Nevada.'

According to Keesler's plan, each utility owning a nuclear unit in the US would be assessed $50,000 per year, per unit, for the cost of Nevada advertising designed to "convert" the Nevada citizens to favoring the proposed Yucca Mountain high-level nuclear waste repository. For the 112 commercial nuclear reactors in the US, this assessment comes to $5.6 million annually. Keesler asked the executives to keep his letter "confidential" because "all costs for the utility campaign" are to be charged to utility "customers, not stockholders."^

Keesler's actions raise a host of ethical questions.^ Central among them is whether a particular industry ought to attempt to coerce both citizens of Nevada

754 BUSINESS ETHICS QUARTERLY

and the US DOE to accept a risk (the repository) whose central benefits accrue to that industry. Another question is whether such one-sided "educational" ef- forts (directed by a regulated monopoly) ought to be funded by the ratepayers, without their knowledge, when a subset of these ratepayers are those likely to be put at risk because of the repository. A broader question is whether the behavior of the nuclear-utility representatives points to the failure of profit- based market allocation."^ An equally broad question is whether Keesler's plan takes adequate account of public welfare and public consent to industrial risks. Or, is the public effectively shut out of the "iron triangle" of industry, govem- ment, and contractors/subcontractors—an iron triangle of cooperation, influ- ence, persuasion, and money that is "beyond the control of existing laws"?^

In this essay I argue that there appears to be an "iron triangle" promoting the siting of the world's first permanent geological repository for high-level nuclear waste and spent fuel, proposed for Yucca Mountain, Nevada. Government con- tractors, scientists and consultants, US DOE officials, and nuclear industry representatives are all eager to build Yucca Mountain. Noting that 80 percent of Nevadans are opposed to the proposed facility, I argue (1) that scientists cannot guarantee Yucca Mountain safety; (2) that uncertainty regarding Yucca Moun- tain is so great that it is not quantifiable; (3) that no other country in the world is moving to permanent geological disposal of radioactive waste as quickly as the US; and (4) that ethics requires, in such a situation of uncertainty, that industry, government, and scientists attempt to limit false negatives (type-II risks), false assurances that Yucca Mountain will cause no serious harm. More- over, I conclude that, because the "iron triangle" of industry, govemment, and consultants/contractors is heavily promoting Yucca Mountain, despite significant scientific uncertainties, the business climate surrounding this triangle appears to leave little room for consideration of ethical issues related to public safety, environ- mental welfare, and citizen consent to risk. If these speculations about the "iron triangle" are correct, then some of the most pressing questions of business ethics concem the acceptability of the industry-govemment-contractor triad.

2. Historical Background

For nearly four decades, virtually all scientists and public policymakers have agreed that permanent geological disposal is the preferred method of dealing with high-level radioactive waste during the 10,000 years that it remains a serious threat to health and safety. Because Yucca Mountain, Nevada has been proposed as the location of the first permanent geological repository for high- level radioactive waste anywhere in the world, the US is spending billions of dollars to study and engineer the site. Indeed, during the last five years, the formalities of site study and selection have cost more than $2.5 billion,^ and the US government is nowhere close to final approval of a single site. Because of the scientific and financial preeminence of Yucca Mountain, it provides a para- digm case of the ethical, policy, and scientific questions associated with perma- nent disposal. Although the US Department of Energy (DOE) studies of the

ENVIRONMENTAL RISK AND THE IRON TRIANGLE 755

Nevada location are state-of-the-art quantitative risk assessments, this essay argues that the optimistic assessment conclusions about site suitability both conflict with fundamental scientific uncertainties about Yucca Mountain and raise questions about how far the "iron triangle" (of industry, govemment and contractors) controls repository siting.

As early as 1955, researchers representing the US National Academy of Sci- ences (NAS) recommended permanent isolation of high-level radioactive wastes in mined geological repositories, a position the NAS spokespersons hold today.^ This basic approach to disposal of high-level radioactive wastes is still being pursued in virtually every nation in the world. As Don U. Deere, Chair of the US Nuclear Waste Technical Review Board of the NAS, expressed this position in 1990: "There is currently a world-wide scientific consensus that a deep geologic repository is the best option for disposal of high-level waste. The Board believes that there are no insurmountable technical reasons why an acceptable deep geologic repository cannot be developed."^

The most fundamental reason that virtually all govemments and nuclear-risk experts have pursued a policy of developing repositories for permanent geologi- cal disposal of high-level radioactive wastes is that they wish to maximize waste isolation. Other arguments in favor of permanent geological disposal are that it minimizes both costs and hazards, especially transport risks to and from a storage facility. Still other reasons for permanent disposal are that we, members of the present generation, should solve the high-level radioactive waste problem, not merely store the waste and thus leave the burden to members of future generations.^ The underlying assumption of this rationale for disposal is that only a permanent geological repository addresses important ethical obligations to future persons. The technical disadvantages of permanent geological disposal of high-level radioactive wastes are the lack of experience with long-term iso- lation and the difficulty of knowing geological features and processes at the great depths and over the long time periods required. Some persons also oppose permanent geological disposal because they claim that it is impossible to assure isolation of the wastes underground. Other arguments against permanent dis- posal focus on technical uncertainties, on political difficulties associated with siting the facilities, on ethical problems related to imposing such a risk on members of future generations, and on the importance of the retrievability of the waste, so as to leave open the options for future storage or disposal.'"

In 1982, Congress passed the Nuclear Waste Policy (NWPA), perhaps the single most important piece of legislation affecting high-level radioactive-waste disposal. The act mandated permanent disposal of radwaste, a policy that had for years been the conventional wisdom. Containing timetables for the Department of Energy (DOE) to accomplish permanent, underground disposal of high-level waste, the NWPA govems commercially generated materials but allows for disposal of defense wastes, given Presidential approval. The NWPA also requires an Office of Civilian Radioactive Waste Management, with its director reporting to the Sec- retary of Energy. Perhaps most importantly, the act provides guidelines for site selection of possible high-level radioactive-waste repositories."

756 BUSINESS ETHICS QUARTERLY

Under the guidelines of the 1982 NWPA, the DOE selected a number of sites as potentially acceptable for the first permanent high-level radwaste repository in the US. They were in Washington, Utah, Texas, Mississippi, Louisiana, Ne- vada, the Great Lakes area, and the Appalachian range. In 1987, the choice of sites was narrowed to Hanford (Washington), Yucca Mountain (Nevada), and Deaf Smith (Texas). After much political compromise, the US Congress passed the Nuclear Waste Policy Amendments Act of 1987; one of its main provisions was to mandate study of only one site. Yucca Mountain, Nevada. Other special features in the act are the requirements to create a Nuclear Waste Review Board in the National Academy of Sciences; to ship spent fuel in NRC-approved packages, with state and local authorities notified of shipments; and to provide an analysis, between the years 2007 and 2010, of the need for a second reposi- tory.'^ Only if the Nevada site is found unacceptable will other possible loca- tions be considered. Currently scientists and engineers are studying the hydrogeology, seismicity, volcanism, and climate of the Nevada location. How- ever, on January 5, 1990, the Nevada Attorney General filed a court petition seeking a "notice of disapproval" of the Yucca Mountain site under the NWPA. The petition failed, and Nevada has appealed it to the US Supreme Court.'^ The US Supreme Court, however, denied further review. It said that discussion of constitutional issues (related to Nevada's support of an absolute right to veto the selection of the Yucca Mountain site) was premature. In other words, Nevada's alleged right to veto the site can be discussed only after the site is formally selected for a repository, after all licensing and permitting procedures are com- pleted.'"* Hence, the DOE plans for Yucca Mountain remain in question.'^ Some persons have even argued that the DOE may have to abandon its current plans and consider other options, such as sub-seabed disposal or above-ground stor- age.'^ Evaluation of the Yucca Mountain site continues, however, despite the opposition of 80 percent of Nevadans to the repository.'^ Site studies will cost several billion more before site evaluation is complete.'^

Part of the controversy driving the opposition of Nevadans to the proposed Yucca Mountain facility is not only the possibility of repository failure and radioactive contamination but also the questionable way in which the nuclear industry, the DOE, and its contractors—the iron triangle—are performing the Yucca Mountain environmental risk assessments. At the heart of this controversy is disagreement over the assessment methods and the data that are being used.

3. Assessors Cannot Guarantee Yucca Mountain Safety

The authors of a recent US Geological Survey (USGS) study of the proposed Yucca Mountain site warned that site "data are not sufficient to predict accu- rately rates of [ground] water movement and travel times."'^ One question raised by the USGS warning is whether the Yucca Mountain predictions, al- though inaccurate, are accurate enough for us to build the repository. Are the questionable inferences in the repository risk estimates and evaluations signifi- cant? Or, are the quantitative risk assessments (QRAs) nevertheless accurate enough to justify permanent geological disposal of high-level radwaste?

ENVIRONMENTAL RISK AND THE IRON TRIANGLE 757

If no scientific result is ever certain or completely objective, and if no policy is ever perfectly just, a reasonable person ought fault neither science nor policy merely for uncertainty, subjectivity, or incomplete justice. The real issue is the significance of the apparent problems in the Yucca Mountain assessments. How objective is objective enough? How certain is certain enough? How just is just enough? Do the available data and site characteristics lead one to believe that QRAs of Yucca Mountain can be done with sensitivity and precision adequate to insure credible regulation and long-term safety?

Many risk assessors believe that the data and the site are adequate to insure safety. They say that Yucca Mountain would comply with the regulations.̂ ** This judgment, however, is quite controversial given all the ways in which incom- plete data, inadequate theory, uncertainty, and site heterogeneity threaten accu- rate knowledge of Yucca Mountain. Even US DOE assessors use language that suggests their largely qualitative and imprecise knowledge of the site is a prob- lem. Note, for example, the US DOE's use of the terms 'estimate,' 'likely,' and 'significant' in the following claim:

estimates of groundwater travel time along any path of likely and significant radionuclide travel from the disturbed zone to the accessible environment are more than 1,000 years. Therefore, the evidence does not support a finding that the site is disqualified.̂ *

Presumably, if DOE officials were more certain about Yucca Mountain safety, they would speak of "calculations" or definite "probabilities" of certain ground- water travel times and not of "estimates." Likewise, if their data were more accurate, presumably they would speak of threats posed by "any path of radionu- clide travel," rather than of threats "along any path of likely and significant radionuclide travel." As the DOE's own works illustrate, its claims of safety are laden with methodological judgments about "likely" travel, for example, and with language that avoids assigning any probabilities to regulatory compliance. The DOE officially admits, for example:

The characteristics of the Yucca Mountain site and the processes operating there permit, and probably ensure, compliance with the limits on radionuclide release to the accessible environment.

When one is considering a potentially catastrophic threat to health and safety, however, one requires a very high probability that the site in question will comply with regulations. One of the main reasons why the methodological judgment—that site knowledge is adequate for regulation and for safety—is questionable is that the various DOE probabilities allegedly associated with site characteristics are already very close to the limits of regulatory acceptability. We shall argue that, given a variety of questionable inferences, assumptions, and value judgments made by assessors,^^ actual site characteristics might not com- ply with regulations. Changes of only one order of magnitude in some of the parameters dealing with fracture flow, infiltration, precipitation, or volcanic and seismic activity could initiate disastrous changes—such as flooding or unaccept- ably rapid groundwater transport—in the Yucca Mountain repository. As Amory

758 BUSINESS ETHICS QUARTERLY

Lovins warned, an error factor of two at each stage of a twenty-step methodol- ogy permits a possible millionfold mistake. '̂* For example, increasing the al- leged percolation rate by only one order of magnitude could initiate fracture flow and speed groundwater-travel time.^^ Such sensitive numbers, together with the two to six orders of uncertainty of characterizing many risk assess- ments, show that the margin for error at Yucca Mountain may be too slim to insure adequate government regulation and safety. Even the US National Acad- emy of Science (NAS) noted that the DOE assumes, incorrectly, "that the prop- erties and future behavior of a geological repository can be determined and specified with a very high degree of certainty. In reality," said the NAS, "the inherent variability of the geological environment will necessitate frequent changes in the specifications."^^ But if geological variability necessitates changes in repository specifications, then there is question whether a facility like Yucca Mountain can meet the pre-determined US safety regulations.

Porous flow alone onsite would mean leachate could reach the water table at Yucca Mountain in 10,000 to 20,000 years.^' Fracture fiow, however, could enhance transport of water and radioactive leachate, above the flux at Yucca Mountain, by as much as 5 orders of magnitude.^^ Assessors have confirmed that "fractures do exist of sufficient width to allow significant water flow in the unsaturated region."^^ Moreover, with a large fracture-fiow rate,^'C, ^^^, and 237 Np could get through to the water table in less than 10,000 years.3° Hence, understanding fracture fiow is a crucial determinant of site safety. Yet, knowl- edge of fractured zones, particularly for unsaturated regions, is very limited. Likewise, the seismicity at Yucca Mountain, prior to 1960, is virtually unknown even though seismic failure is possible.3' One wonders how a possibly seismic, fractured site, even in an arid climate like Yucca Mountain, could be acceptable if volcanism, intruding water, and seismic activity were not highly improbable during the life of the repository.^^ At Yucca Mountain, these conditions do not appear to be highly improbable.

A person who makes the value judgment that site knowledge is sufficient for regulation and for safety is in the questionable position of knowing that signifi- cant problems could occur with fracture flow, seismicity, and volcanism, yet not being able to predict any of them accurately—because of numerous difficulties with modelling, sampling, extrapolation, and so on. Even the Nuclear Regula- tory Commission (NRC) officials recognized some of these problems when they complained that the Yucca Mountain risk assessments fail to recognize ade- quately the uncertainty in the data. Likewise, the US NAS warned that "uncer- tainty is treated inappropriately" in the Yucca Mountain assessments.^^ Indeed, the NRC said that the environmental assessments of the DOE for its proposed radwaste facilities are, in general, "overly optimistic."^'^ Such optimism often appears almost gratuitous, because it is not based on precise, quantitative pre- dictions. For example, an official DOE document claims that the site can protect the safety of ail future generations from radiological hazards:

ENVIRONMENTAL RISK AND THE IRON TRIANGLE 759

The quality of the environment during this and future generations can be adequately protected. Estimates of radiation releases during normal operation and worst-case accident scenarios provide confidence that the public and the environment can be adequately protected from the potential hazards of radio- active-waste disposal.

Equally gratuitous is the DOE claim that no future groundwater conditions will disrupt the site:

Currently available engineering measures are considered more than adequate to guarantee that no disruption of constniction and operation will occur be- cause of groundwater conditions at Yucca Mountain.''

Such assurances are highly questionable, given DOE assessors' admissions of uncertainties about basic hydrological and geological conditions at the site. For example, at Yucca Mountain, "in most cases, hydraulic data are insufficient for performing geostatistical analyses,"^' and "traditional fiow path chemical evalu- ation does not directly apply to tuffaceous volcanic environments."'^ Likewise, there is "no known mechanical model that describes nonuniform corrosion well enough to use in performance assessment" of the waste canisters.^' In areas of hydrology, geology, canister security, climate, volcanism, and seismicity, no techniques exist, at the present time, that are adequate for removing the uncer- tainties at Yucca Mountain or even for quantifying them.'*'' Basic questions conceming the reliability of the studies remain unanswered.^^ Indeed, how could significant uncertainties be removed if one required precise predictive power and regulatory guarantees regarding the site for 10,000 years?

The long time period of storage is one reason that Yucca-Mountain reviewers have claimed that "compliance with US [radiation-dose] limits cannot be shown objectively by PRA [probabilistic risk assessment] methods."^^ One reason for this problem is that the precise, probabilistic standards of the Environmental Protection Agency (EPA) for the management of spent fuel and high-level and transuranic radioactive wastes cannot be confirmed with current data. The stand- ards set limits for releases when events have more than a 1 in 10 chance of occurring over the 10,000 years.''^ Such precise probabilistic standards cannot be guaranteed for so long a time, however. As one DOE reviewer put it: "no assurance can be given that all significant factors have been examined here.'"*'* Other reviewers maintain that it is doubtful whether we can model or predict long-term behavior at all, given the heterogeneities and uncertainties at the site.'*^ Still other evaluators, including those from the utility industry and the NAS, have proclaimed that the limits of environmental science have been ex- ceeded by the goals set by the nation's radioactive waste program.^* Perhaps the most significant analysis of how scientific uncertainties undercut assurances of repository safety is that of the DOE team of 14 peer reviewers who in 1992 analyzed the DOE's Early Site Suitability Evaluation for Yucca Mountain. The "consensus position" of the 14 DOE-selected peer reviewers is telling:

It is the opinion of the panel that many aspects of site suitability are not well suited for quantitative risk assessment. In particular are predictions involv- ing future geological activity, future value of mineral deposits and mineral

760 BUSINESS ETHICS QUARTERLY

occurrence models. Any projections of the rates of tectonic activity and vol- canism, as well as natural resource occurrence and value, will be fraught with substantial uncertainties that cannot be quantified using standard statistical methods.

If uncertainties at any proposed site are so severe that they cannot be quantified, then it is arguable that they force those who currently favor a permanent reposi- tory—some members of the iron triangle—into either begging the question or appealing to ignorance in defending site suitability. Indeed, anyone who main- tains that there is, at present, a compelling scientific basis for permanent geo- logical disposal is unavoidably forced to use incomplete and short-term data (on seismicity, volcanism, hydrogeology, and so on) as a basis for extraordinarily precise, long-term predictions—tens of thousands of years—about site suitabil- ity. We are able to make general predictions about the future, of course, and geologists do so all the time. Precise predictions, however, are a problem. Be- cause of the imprecision of our hydrogeological and climate models, we are at present unable to predict the geological and hydrological situation at Yucca Mountain with any degree of reliability and precision, 10,000 years into the future. As a result, we cannot quantify the claim that we shall be able to meet current US repository standards for safety 10,000 years from now. We cannot be reasonably assured that a permanent repository might not cause catastrophe hundreds or thousands of years into the future. Indeed, to claim the ability to predict very precise geological events, 10,000 years into the future, when one's precise, site-specific evidential base for doing so covers only tens of years, has little scientific justification. Although we can reconstruct geological histories spanning millions of years, geology is primarily an explanatory and not a pre- dictive science, as we argued earlier. Hence, it seems prima facie evident that one ought not base arguments for the safety of a permanent repository on an uncertain judgment about our ability to make precise geological predictions.

Another reason that it is difficult to know the distant future in great detail is that we humans and our institutions are not precisely predictable. Anyone who argues for permanent geological disposal must discount the effects (on reposi- tory safety) of human error and the social amplification of risk that might occur in thousands of years. Discounting these effects is problematic, as the Chair of the US NAS overview committee (for the WIPP project for storage of weapons- related radwaste in New Mexico) noted before Congress:

current feeling is that the WIPP site could probably meet EPA standards with the exception of the so-called "human-intrusion" scenario. This is the idea that some- time in the future somebody comes and drills directly into a repository. . .'*

As the NAS committee warned, dismissing the effects of human activities such as terrorism, sabotage, or ignorance, tens of thousands of years into the future, is highly problematic. Indeed, given the prevalence of fiaws in humans and their institutions, it might be more reasonable to assume that terrorism or ignorance would be a major problem for a facility storing radiotoxic materials. Moreover, whether about climate and hydrogeology, or about human errors and institutions.

ENVIRONMENTAL RISK AND THE IRON TRIANGLE 761

precise predictions about the long-term future are highly questionable, at least at present, because our generalizations are built on such a limited empirical base.

If it is impossible to know the long-term future with great precision, then any claims to precision (as US radwaste regulations require) about the long-term future must rely in part on ignorance. Yet, from ignorance about a particular claim, it is logically invalid to conclude that the claim is either true or false. From our ignorance about future, long-term, repository safety, it is logically invalid to conclude that a repository would be either safe or unsafe. Like many scientific claims, conclusions about the safety of repositories—tens of thou- sands of years into the future—are uncertain. Based on data from the present or even from several decades, there can be no empirically compelling argument for the safety of such repositories in the distant future. The best our experiments can do is to confirm that, if permanent repositories meet certain safety standards in the future, then our current experiments are likely to exhibit these same features. Be- cause affirming the consequent does not invariably lead to valid conclusions, how- ever, the reverse is not true. We cannot infer that because of the success of current, short-term experiments, therefore repositories will avoid catastrophic releases of radionuclides and will meet safety standards thousands of years from now.

Because of all the uncertainties in the Yucca Mountain data and methods, assessors typically are not able to determine the degree of accuracy in their models.'*' They are able, for example, merely to say that there is a "high level of probability" that groundwater travel time to the water table will exceed 10,000 years.̂ *^ In other words, the degree of uncertainty regarding groundwater travel time is very great. Likewise, the margin of safety necessary to prevent signifi- cant problems, such as fracture flow, is quite slim. Yet, despite this narrow "window," some persons appear to believe that Yucca Mountain will be predict- ably safe or in compliance with govemment regulations requiring a groundwater travel time greater than 1,000 years.^' There is also only a "narrow window," or slim margin, of safety because groundwater travel time is extremely sensitive to fracture flow, and fracture flow is extremely sensitive to percolation rate. If either flow or percolation increase by even a small amount, then the travel time of leachate from the waste will increase significantly.^^ In the world of ground- water flow, where risk assessments "are highly uncertain,"^^ a factor of 10 as a window of safety is quite small. Indeed, in some of the simulated cases, water travel time from the repository to the water table is less than 1,000 years.5^* Hence, the methodological judgment that current and near-future knowledge about Yucca Mountain can guarantee safety and compliance with govemment regulations—for example, requiring groundwater travel time of more than 1,000 years—may be questionable.

The judgment about travel time is not only factually questionable but also inconsistent. One well known group of assessors, for example, found that, ac- cording to their models, some calculated groundwater travel times are less than 10,000 years. They also admitted that hydraulic data were insufficient, and that there has not been enough time to estimate cumulative radioactive releases.^^

762 BUSINESS ETHICS QUARTERLY

Nevertheless, they concluded that the "evidence indicates that the Yucca Moun- tain repository site would be in compliance with regulatory requirements,"^^ and that "no radioactivity from the repository will migrate even to the water table immediately beneath the repository for about 30,000 years."^'' How do some migration values of less than 10,000 years translate to a migration time of "about" 30,000 years? How can the same DOE assessors claim that the reposi- tory will be in compliance with govemment regulations^* when they also assert that low flux "will probably limit fiow velocities to the extent that no leachate will reach the water table for tens to hundreds of thousands of years"?^^ Such poorly grounded "probable" knowledge of something that may occur within tens to hundreds of thousands of years (a wide range) is hardly consistent with precise claims about safety and regulatory compliance! Likewise, how can the same DOE assessors conclude, with confidence, that no radioactivity will mi- grate to the water table for at least 30,000 years,^° and yet claim: "Because data and understanding about water flow and contaminant transport in deep unsatu- rated fractured environments are just beginning to emerge, complete dismissal of the rapid-release scenarios is not possible at this time"?^' How is the 30,000- year claim consistent with the assertion about not dismissing the rapid-release scenarios?

Assessors investigating the uncertainties in the Yucca Mountain hydro- geological data also have admitted that, for the unsaturated zone, uncertainties in groundwater velocities may be as much as 100 percent above or below the mean value.̂ -̂ They likewise claim that a change in percolation of a factor of only 10 is sufficient to initiate fracture fiow, that groundwater travel time is extremely sensitive to fracture flow,̂ ^ and that heat from the waste could cause fractures. '̂* Given such admissions, how can the same DOE assessors consis- tently claim that fracture fiow is not a credible process,^^ and that groundwater flow will be "well within the limits set by the NRC"?^^ Similar inconsistencies appear, when the same assessors, after acknowledging (1) that they have incom- plete data,^^ (2) that they have had no time to estimate cumulative radioactive releases,^^ and (3) that they may "have underestimated the cumulative releases of all nuclides during 100,000 years, by an amount that is unknown,"^^ never- theless draw a contradictory conclusion. They conclude that only one ten-mil- lionth of allowable releases of radionuclides will reach the water table.™

Likewise, Yucca-Mountain assessors admit that solubility limits and retarda- tion factors are site- and (radioactive) species-dependent.''' They also claim that they may have underestimated radioactive releases.^^ If the same DOE assessors do not know the degree to which they may have underestimated radioactive releases,^' how do they know so precisely that only one ten-millionth of allow- able releases will be released? Similar inconsistencies and unsupported extrapo- lations occur throughout the Yucca Mountain analyses, with DOE assessors confidently affirming that there will be "less than one health effect every 1,400 years." '̂̂ A more precise and consistent appraisal, given the problems with the data and models at Yucca Mountain, might be that of the assessors who con-

ENVIRONMENTAL RISK AND THE IRON TRIANGLE 763

eluded: "Even though we have tried to use the best data and models available at this time, we make no claims that these results have any value in the perform- ance assessment of the Yucca Mountain repository site."'*

Instead of using such precise language, however, the DOE's final 1992 Early Site Suitability Evaluation (ESSE) for Yucca Mountain continues to formulate site risks in terms of words such as "likely" and "unlikely," rather than by using numerical probabilities.̂ *^ Similarly, when DOE reviewer M. T. Einaudi com- plained that the ESSE had vaguely defined the "foreseeable future" as "the next few years to 10 years, and occasionally as long as 30 years,"^' the DOE ESSE team responded by removing from the document all language mentioning the number of years. Next the team noted:

The evaluation and definition of the terms, such as "reasonable projections" and "likely future activities" will receive considerable attention in the future and is likely to utilize the review of a panel of experts.

This response, however, does not solve the problem with vague language, both because the DOE team uses the language to argue for site suitability, and pre- sumably such usage must have implications. Indeed, if the language did not have certain implications regarding future time periods, then it would not be part of an effective argument for site suitability. Hence, if the terms are used effectively, they must have some precise, implicit meaning. If they do not have a precise, implicit meaning, then it is arguable that they are not effective in supporting the site-suitability conclusions and ought not be used. Indeed, by using indefinable terms to defend conclusions about site suitability, the ESSE renders its conclu- sions nonfalsifiable and therefore ineffective, because vague claims cannot be falsified. And if the ESSE site- suitability claims are not falsifiable, then this suggests that they are a priori rather than empirical and scientific.

Another reviewer (of the 1992 ESSE for Yucca Mountain), J. I. Drever, also complained about the failure of the ESSE to provide rigorous definitions of words such as "likely" and "significant."^^ Again, the final ESSE document did not alleviate the difficulty. Instead the ESSE Core Team responded to Drever's criticism:

The terms 'likely' and 'significant' should be defined in the context of the overall postclosure performance objectives. Because the evaluations of sys- tem performance cannot be definitive at this time, the ESSE Core Team be-

80 lieved it inappropriate to define those terms precisely for this evaluation.

This response by the DOE team, however, creates more questions than it an- swers. For one thing, to say that terms like "likely" should be defined in terms of overall postclosure performance is not coherent, because the term "likely," for example, is rarely if ever used in the context of "total system performance." Rather, it is used in radically different, but specific contexts, such as probability of human interference at the site, or the probability of a route of radionuclide transport.*' Hence, terms like "likely" not only do not refer to "overall perform- ance," as the DOE team claimed, but, second, they are not univocal. They clearly mean different things in different ESSE contexts. Third, although the ESSE team says that such terms cannot be defined precisely because the system evaluations

764 BUSINESS ETHICS QUARTERLY

are incomplete, this response is puzzling because the ESSE team obviously has already used the terms to mean something. Fourth, if the system-performance evaluations are not definitive enough to allow the ESSE team to define the very terms that it uses, then it is unclear why the system-performance evaluations are definitive enough to support a lower-level suitability finding, rather than an unsuitability finding, for Yucca Mountain. Fifth, contrary to the response of the DOE ESSE Core Team, the terms used by the team clearly presuppose some precise meanings, because words like "likely" are often used in precise regula- tory contexts, such as "not likely to exceed a small fraction of [radiation dose] limits."*^ If such terms were not used somewhat precisely, then it would be impossible for the claims in which they are imbedded not to be false. Likewise, the ESSE Core Team claims, for example, that "although confidence is substan- tial, it is not yet sufficient to support the higher-level suitability finding for this qualifying condition."^^ Such a claim appears to presuppose some precise level or cut-off of confidence or likelihood. It appears to presuppose that lower-level findings are justified below this level, and that higher level findings are justified above it. For all these reasons, there appears to be a mismatch between the science and the regulations discussed in DOE assessments such as the ESSE. Because of this mismatch, it is questionable whether the science discussed in repository assessments is adequate to the regulatory task.

Previous experiences at the Maxey Flats low-level radwaste facility show that similar problems with value judgments about hydrogeological accuracy—and the ability of QRA to meet regulatory guidelines—may have occurred there. Envi- ronmental Protection Agency (EPA) assessors believed that the knowledge of the Maxey Flats site was adequate to insure containment, credible regulation, and safety, largely because "the general soil characteristics" at the facility have been "very impermeable." '̂* Yet, such general assurances failed to address the problem of leachate migration with sufficient precision and accuracy. Other US EPA geolo- gists noted that precise determination of hydraulic conductivity is impossible at a site, like Maxey Flats, with fractures.̂ ^ US Geological Survey (USGS) scientists claimed that the Maxey Flats hydrogeology, because of the fractures, was "too complex for accurate quantitative description."^^ Given the complexity and uncer- tainty associated with much information about Yucca Mountain, there is reason to believe that optimistic judgments, about the accuracy of site studies, may err just as they did at Maxey Flats. Because inaccurate knowledge of the Yucca Mountain facility prevents scientists from being able to predict precisely migration rates of the waste thousands of years into the f̂ uture, it also prevents them from guarantee- ing that the proposed repository will comply with very specific, US radiation- dose limits. Because compliance with government regulations is unknown, and because the consequences of repository failure could be catastrophic, it is argu- able that tbe Yucca Mountain facility ought not be built, at least not until there is significantly more knowledge about the future risks likely to be associated with the installation. The fact that nuclear industry, DOE, and contractor repre- sentatives support siting the facility suggests that this "iron triangle" may be taking inadequate account of scientific concerns about the site.

ENVIRONMENTAL RISK AND THE IRON TRIANGLE 765

4, Nonquantifiable Uncertainty at Yucca Mountain Argues Against Disposal

US NAS panelists said that perhaps the US should delay any permanent radwaste facility until we have more knowledge about long-term repository behavior. Likewise, a major US government commission, studying policy for dealing with high-level radioactive waste, concluded recently that Congress should reconsider the subject of interim [rather than permanent] high-level rad- waste storage by the year 2000 so as to "take into account uncertainties that exist today and which might be resolved or clarified within 10 years." Indeed, said the commission, "despite the considerable time and money already expended to site a repository, none has been sited yet, and the date by which a permanent repository will be available is uncertain...the most notable uncertainty" is the "date of opening a permanent repository" in the US.*''

At least part of the reason for the commission's worries, it appears, are the scientific uncertainties associated with the proposed facility at Yucca Mountain, some of which have been outlined in the preceding section. Moreover, to the degree that this nonquantifiable uncertainty precludes assurance that precise radiation-control standards can be met during the thousands of years of opera- tion of the proposed Nevada repository, to that extent it is arguable that we cannot yet guarantee the safety of permanent waste disposal. And if we cannot guarantee the long-term safety of proposed repositories, like Yucca Mountain, then the "dig now, pay later" approach of repository supporters is highly ques- tionable. Part of the rationale for delay or avoidance of a permanent US reposi- tory is a basic legal premise: res inter alios acta alteri nocere non debet: no one ought to suffer from what others have done.** Unless we can guarantee that many others in the future will not suffer unreasonably from what we have done in building a permanent repository, then our scientific uncertainty may be suffi- cient to argue against building the Yucca Mountain permanent repository.

Why does our uncertainty about whether Yucca Mountain will lead to catas- trophe in the future argue against the facility? Brian Berry has provided one of the simplest rationales for the claim that the possibility of causing future catas- trophe is a decisive reason for not acting in the present. He argues that, (1) in the case of an individual making a possibly lethal choice that affects only himself we should regard anyone who chooses the potentially fatal action—who claims that uncertainty makes it premature to decide against the action—as crazy. Likewise, says Barry, (2) when we change the case to one that involves millions of people and extends over many centuries, the same reasoning applies with increased force. Barry's rationale for (1) is that no rational person gambles with his own life except to gain a comparable benefit, to save it. Rock climbers, sky divers, and other risk enthusiasts, however, might claim that they are skilled and well trained and hence not gambling with their lives since the probability of death for such a skilled person is low. Risk enthusiasts probably would also argue that they gain great benefits from their activities. Both Barry and these enthusiasts would likely agree, however, that as the benefits decreased, and as

766 BUSINESS ETHICS QUARTERLY

the probability of death increased, the risky actions become more foolish. Hence, (I) is reasonable. Barry's rationale for (2) is that, because the numbers of persons potentially at risk of death are larger, the impetus for choosing against the risk is likewise even greater. Despite reasoning such as Barry's, official US DOE documents have argued for permanent repositories on exactly the grounds that Barry says are most questionable. He claims that anyone in this position —who argues that uncertainty makes it premature to decide against a potentially catastrophe action—is "crazy." Yet, the US DOE repeatedly has argued for such a claim, for example:

A final conclusion on the qualifying condition for climatic changes cannot be made based on available data. However, the evidence does not support a finding that the reference repository location is not likely to meet the qualify- ing condition.

In other words, DOE officials have used uncertainty about climatic changes as an argument for the thesis that the repository ought not be disqualified. Such an argument, an appeal to ignorance, is problematic on both logical grounds and for the ethical reasons outlined by Barry. Moreover, in cases of future catastrophic risk, Barry's reasons (1) and (2) likewise are compelling, because a repository catastrophe presumably could wipe out an entire culture, not just many persons, and destroying a culture may be worse than merely killing many people. Also, in the case of our threatening future generations, the repository risk is imposed without the consent of the possible victims, and it is not confined to the benefi- ciaries—a point that we shall not take time to discuss here. For all these reasons, scientific uncertainty raises numerous questions regarding siting permanent rad- waste facilities like Yucca Mountain.^°

5. Uncertainty and Permanent Disposal: Other Countries

Despite the uncertainties associated with Yucca Mountain, the US could have a permanent geological facility for storage of high-level radioactive waste there as early as 2010.^' No other country is moving so quickly to permanent reposi- tories. Officials in other nations have openly admitted that they are proceeding more slowly with high-level radioactive waste disposal, precisely because of the scientific uncertainties involved. As the Board on Radioactive Waste Manage- ment of the National Research Council of the US National Acadethy of Sciences (NAS) put it:

The US program is unique among those of all nations in its rigid schedule, in its insistence on defining in advance the technical requirements for every part of the multibarrier system, and in its major emphasis on the geological com- ponent of the barrier as detailed in 10 CFR 60. Because one is predicting the fate of the HLW into the distant future, the undertaking is necessarily full of uncertainties.... It may even tum out to be appropriate to delay permanent closure of a waste repository until adequate assurances concerning its long- term behavior can be obtained through continued in-situ geological studies.... There are scientific reasons to think that a satisfactory HLW repository can be built and licensed. But for the reasons described earlier, the current US pro- gram seems unlikely to achieve that desirable ^̂

ENVIRONMENTAL RISK AND THE IRON TRIANGLE 767

What can we learn about the likelihood of success in permanent geological disposal, on the basis of activities in the various countries considering the repository option?

In eight of the nations with the most radioactive waste, uncertainties have forced the countries to postpone permanent geological disposal. In Canada, for example, although nuclear reactors supply about 13 percent of the country's electricity, there has been no decision about spent reactor fuel, although Canada will have approximately 34,000 MTU by the end of the century. Given no decision about permanent disposal, the earliest Canadians could have such a repository is 2010, even assuming that it wanted one.'^

Similarly, the French plan to use interim storage for a minimum of 20 years before moving to permanent disposal. Nuclear reactors currently supply more than 70 percent of French electricity. The earliest a permanent facility could be ready in France is 2010. The French rationale for delaying decisions about permanent storage is that cooling the waste would reduce the thermal impact on the host rock where it might be stored. In the Yucca Mountain studies, many problems have arisen because of the ability of the high-temperature wastes to induce thermal fractures in the host rock, thereby increasing the probability of fracture flow of the leachate. Because of such diH'iculties, "the French believe that the period [of interim storage] could be extended as long as needed because of the safety of monitored storage."'"*

Nuclear reactors supply approximately 40 percent of electricity in Germany. Like France, Germany is building interim storage facilities for high-level radio- active wastes, although the Germans hope to use deep geological disposal at the Gorleben salt dome. Even if the German plans are not delayed, the earliest a permanent repository could be ready is 2008. The Gorleben facility was licensed in 1983, but litigation conceming safety and scientific uncertainty has, so far, prevented its use as repository for spent fuel.̂ ^ In Japan, approximately 32 percent of the nation's electricity is supplied by nuclear reactors. Despite this fact, the Japanese appear to be quite concemed about a premature leap to an inadequately tested technology for permanent waste disposal. They plan to store their vitrified waste for 30 to 50 years before considering deep geological em- placement. In fact, the Japanese do not plan even to try to develop regulations for siting a permanent repository until after the year 2000. Hence, if approved, the earliest date at which a Japanese repository could operate is 2030.'^

Spain is following a strategy similar to that of its European neighbors. With 36 percent of its electricity supplied by nuclear reactors, the Spaniards plan to store spent fuel at the reactors for 10 years, and then to use interim storage for another 40 years. Sometime around the tum of the century, they plan to consider possible candidate sites for permanent geological disposal. Their explicit strat- egy is to gain more experience dealing with the wastes before deciding what to do with them.^'

In Sweden, approximately 50 percent of electricity is supplied by nuclear reactors. Because of scientific uncertainties and because they want to achieve a

768 BUSINESS ETHICS QUARTERLY

tenfold reduction in radiation and heat output from the waste, the Swedes are storing their spent fuel for 30 to 40 years in centralized, interim storage facili- ties. They do not expect to have a permanent facility available until some time after 2020.^* Like the Swedes, the Swiss plan to store their spent fuel in interim facilities for 40 years. Approximately 38 percent of electricity in Switzerland is supplied by nuclear reactors. The earliest a permanent repository could be avail- able in Switzerland is sometime after 2025. Like the Swedes, the Swiss have laws and regulations that make it impossible to site a new commercial nuclear plant unless operators can demonstrate safe disposal of spent fuel. As a result, no new plants have been sited in either country.^^

The United Kingdom (UK), with 17 percent of its electricity coming from nuclear reactors, has one of the longest periods of interim storage of spent fuel, 50 years. Using interim storage at Sellafield has been necessary, in part, because of opposition in the UK to permanent disposal and because of scientific uncer- tainties associated with deep geological facilities. The earliest date by which the British could have a permanent repository ready is 2030, although the have not begun the siting process.'*^°

Although all eight countries just surveyed are some of the world's major users of nuclear electricity, and even though all of them plan to use permanent geo- logical disposal of spent fuel in the future, none of them expects to do so as quickly as the United States. Indeed, the preferred altemative is to reduce uncer- tainties about behavior of the waste. As the US review commission put it: "In general, deferred disposal is viewed as beneficial because it reduces the heat output of the wastes." As a result, centralized, monitored, interim storage facili- ties have been built or planned in all but one country, Canada, where plans are to use at-reactor interim storage."" If the experience of eight major nuclear countries is correct, then one powerful argument (for not pursuing permanent disposal at present and for postponing a decision about a geological repository) is that no nation, except the US, has plans for rapid permanent disposal of nuclear waste. If the plans of most countries refiect a scientific consensus about our inability, at present, to handle the uncertainties associated with permanent disposal of high-level nuclear waste, then these uncertainties may undercut arguments for permanent disposal anywhere at present.

6. Uncertainty and Permanent Disposal: An Objection

In response to these arguments about the scientific uncertainty associated with the safety of permanent geological disposal, a proponent of the repositories could argue that no science is ever certain, and that scientific certainty is not always required before one acts. In other words, one could argue that reasonable assurance of safety, not scientific certainty, is a precondition for ethically defen- sible behavior. On this view, one could argue that certainty is impossible, and therefore that one need merely follow the best available scientific opinion or the course of action leading to the best estimated results.

The heart of this objection to our analysis is correct. One does not need certainty before one acts, because certainty is unattainable. Our argument, how-

ENVIRONMENTAL RISK AND THE IRON TRIANGLE 769

ever, is not that permanent disposal requires certainty. Rather, the argument is that permanent disposal requires more certainty than we have now, and that at present, the uncertainties associated with permanent disposal are extreme. For now, we wish to raise the issue of what behavior is ethically defensible under conditions of uncertainty. Following Barry's insights already mentioned, our presupposition is that, in cases of extensive scientific and probabilistic uncer- tainty—like those concerning precise geological predictions 10,000 years from now or like those concerning events whose uncertainty cannot be quanti- fied—we ought to behave in an ethically conservative way. But what is ethically conservative behavior? On one view, ethically conservative behavior, in a situ- ation of uncertainty, is behavior that does not reject the null (no-effect) hypothe- sis. That is, if we are uncertain about a catastrophic event in the future, for example, ethical conservatives do not assume there will be no effect. In other words, we ought to minimize type-II statistical errors. Although we shall not take the time to provide the arguments in full here,'"^ there are a number of reasons for minimizing type-II error in situations of uncertainty, like those associated with permanent geological disposal of radioactive waste.

7. Uncertainty and Permanent Disposal: Type-II Error

In a situation of uncertainty, errors of type I occur when one rejects a null hypothesis that is true; errors of type II occur when one fails to reject a null hypothesis that is false. (One null hypothesis might be, for example, "the pro- posed Yucca Mountain repository will secure high-level radwastes so that only one ten-millionth of allowable releases of radionuclides will reach the water table over 100,000 years.")i«>3

Given a situation of uncertainty, which is the more serious error, type I or type II? An analogous issue arises in law. Is the more serious error to acquit a guilty person or to convict an innocent person? Ought one to run the risk of rejecting a true null hypothesis, of not using the Yucca Mountain technology that is really acceptable and safe? Or, ought one to run the risk of not rejecting a false null hypothesis, of employing the Yucca Mountain technology that is really unac- ceptable and unsafe? The basic problem is that to decrease type-I risk might hurt the public, especially members of future generations, and to decrease type-II risk might hurt both present persons and especially those dependent on the industries promoting the permanent repository.

In the area of pure science and statistics, most persons believe that in a situation of uncertainty one ought to minimize type-I risks, so as to limit false positives, assertions of effects where there are none. Pure scientists often attach a greater loss to accepting a falsehood than to failing to acknowledge a truth.'""^ Societal decisionmaking under uncertainty, as in cases involving siting perma- nent radwaste facilities, however, is arguably not analogous to decisionmaking in pure science. Societal decisionmaking involves rights, duties, and ethical consequences that affect the welfare of persons, whereas purely scientific deci- sionmaking involves largely epistemological consequences. For this reason, it

770 BUSINESS ETHICS QUARTERLY

is not clear that in societal cases under uncertainty, one ought to minimize type-I risks. Instead, there are a number of prima facie reasons for minimizing type-II errors. For one thing, it is arguably more important to protect the public from harm (from possible catastrophic radwaste releases) than to provide, in some positive sense, for welfare (building permanent repositories), because protecting from harm seems to be a necessary condition for enjoying other freedoms.'^^ Admittedly, it is difficult to draw the line between providing benefits and pro- tecting from harm, between positive and negative laws or duties. Nevertheless, just as there is a basic distinction between welfare rights and negative rights,•"^ so there is an analogous distinction between welfare policies (that provide some good) and protective policies that prohibit some infringement). Moral philoso- phers continue to honor related distinctions, such as that between letting die and killing someone. It therefore seems more important to protect citizens from public hazards, like a catastrophic leak at a permanent radwaste facility, than to attempt to enhance their welfare, over the short term, by implementing a tech- nology such as permanent geological disposal of radwaste."'^ A second reason for minimizing type-II errors under uncertainty is that the public typically needs more risk protection than do the industry or government proponents of the risky technology, like Yucca Mountain. The public usually has fewer financial re- sources and less information to deal with societal hazards that affect it, and laypersons are often faced with bureaucratic denials of public danger. Certainly members of future generations are likely to have less information to deal with a permanent repository since, by definition (US regulations), it will not be moni- tored. Hence, their needs for protection seem larger, and the importance of minimizing type-II errors appears greater."*^

Third, it is more important to minimize type-II error, especially in cases of great uncertainty, because laypersons ought to be accorded legal rights to pro- tection against technological decisions that could threaten their health and physical security. These legal rights arise out of the considerations that everyone has both due-process rights and rights to bodily security. In cases where those responsible or liable cannot redress the harm done to others by their faulty decisions—as they cannot in the case of repositories' harming future genera- tions—there are strong arguments for minimizing the public risk. Industrial and technological decisionmakers cannot adequately compensate or insure their po- tential victims from bad consequences in the case of permanent disposal, be- cause the risks involve death. Therefore, they are what Judith Jarvis Thomson calls "incompensable." Surely incompensable risks ought to be minimized for those who fail to give free, informed consent to them. Whenever risks are incompensable, (e.g., imposing a significant probability of death on another), failure to minimize the risks is typically morally unjustifiable without the free, informed consent of the victim. "'̂ A final reason for minimizing type-II error in cases of uncertainty is that failure to do so would result in using members of future generations as means to the ends of present persons. It would result in their bearing a significantly higher risk from radwaste than members of present generations, despite the fact that present persons have received most of the

ENVIRONMENTAL RISK AND THE IRON TRIANGLE ' 771

benefits associated with generating the waste. Such discrimination (in this case, against future persons), as Frankena has pointed out, is justified only if it would work to the advantage of everyone, including those discriminated against. Any other attempt to justify discrimination fails because it would amount to sanction- ing t he use of some humans as means to the ends of other humans. *'°

Because the imposition of Yucca Mountain risks discriminates against future persons, it would not work to the advantage of everyone. Because it would not, discrimination against members of future generations likely to be affected by Yucca Mountain appears not to be justified. And if it is not justified, then failure to minimize type-II errors—that cause such discrimination—is also not justi- fied. Hence, in situations of uncertainty, such as those concemed with perma- nent radwaste disposal, the ethically preferable course of action is to minimize type-II, rather than type-I, error. This course of action, in a situation of uncer- tainty, requires us to run the risk of rejecting the null hypothesis, to run the risk of not using permanent, high-level radwaste repositories, at least not until sig- nificant uncertainties are removed.

8. Conclusions

If the arguments of this essay are correct, then permanent geological disposal of radwaste is highly questionable on epistemological, ethical, and political grounds. The epistemological grounds are the tremendous uncertainties associ- ated with permanent disposal, uncertainties arising because of the 10,000-year time frame, the precision of safety predictions required by existing regulations, and our inability even to quantify these uncertainties. The ethical grounds are the necessity to behave in a morally conservative way and to minimize type II errors in situations of uncertainty. The political grounds are the fact that other countries are postponing decisions about permanent disposal of nuclear wastes. All of these grounds raise questions about the fact that members of the "iron triangle" appear to be promoting permanent disposal at Yucca Mountain.

These questions are especially troubling because it is impossible to justify building a permanent radwaste repository, at present, without employing at least two logically invalid inferences: the appeal to ignorance and affirming the con- sequent. Policy based on questionable logical and scientific inferences is highly problematic. Hence, all those who currently support using permanent radwaste repositories—especially representatives of nuclear utilities, the DOE, and DOE contractors—appear to err. Their behavior in the Yucca Mountain case suggests that, in such situations, the iron triangle needs to be either broken or expanded to include the scientific and ethical concerns of the public.

University of South Florida

Notes

'Olinger, D.: 1991, "Nuclear Industry Targets Nevada," St. Petersburg Times (Dec. 1), p. Dl. See also Keesler, A.: 'Testimony,' in C. Fairhurst: 1990, Board on Radioactive Waste

772 BUSINESS ETHICS QUARTERLY

Management, National Research Council, The Federal Program for the Disposal of Spent Nuclear Fuel and High-Level Radioactive Waste, Hearing before the Subcommittee on Nu- clear Regulation of the Committee on Environment and Public Works, U.S. Senate, 101 Congress (U.S. Govemment Printing Office, Washington, DC), pp. 01-02, and Schneider, K.: 1991, 'Nuclear Industry Plans Ads to Counter Critics,' New York Times (Nov. 13), p. A18.

^Olinger, op. cit., p. DL

^See Shrader-Frechette, K. S.: 1993, Burying Uncertainty: Risk and the Case Against Geological Disposal of Nuclear Waste (University of Califomia Press, Berkeley).

'*See Sen, A.: 1993, "Does Business Ethics Make Economic Sense?," Business Ethics Quarterly 3 (Jan.), no. 1, pp. 45-54.

^SeeMarquiss, K.: 1991, "Defense Contracts: Operation III Wind," in Case Studies in Business Ethics, T. Donaldson and A. R. Gini, eds. (Prentice Hall, Englewood Cliffs, NJ), p. 90.

^Rosen, M. E.: 1991, "Nevada v. Watkins: Who Gets the Shaft?," Virginia Environmental Law Journal 10, pp. 239-309.

'Fairhurst, C: 1990, "National Research Council and National Academy of Sciences, 'State- ment'," in US Congress, The Federal Program for the Disposal of Spent Nuclear Fuel and High-Level Radioactive Waste, Hearing Before the Subcommittee on Nuclear Regulation of the Committee on Environment and Public Works, US Senate, Wist Congress, Second Session, October 2, 1990 (US Government Printing Office, Washington, DC), p. 18.

^Deere, D.: 1990, "US Nuclear Waste Technical Review Board, 'Statement'," in US Congress, The Federal Program for the Disposal of Spent Nuclear Fuel and High-Level Radioactive Waste, Hearing Before the Subcommittee on Nuclear Regulation of Committee on Environment and Public Works, US Senate, 101st Congress, Second Session, October 2, 1990 (US Government Printing Office, Washington, DC), p. 18. The position in favor of permanent geological disposal is also confirmed by Blowers, A., D. Lowry, and B. Solomon: 1991, The International Politics of Nuclear Waste (St. Martin's Press, New York), p. 318, and by the US National Academy of Sciences, Commission on Geosciences, Environment, and Resources, National Research Council: 1990, Rethinking High-Level Radioactive Waste Dis- posal (National Academy Press, Washington, DC), pp. v, 6. See Waste Isolation Systems Panel, Board on Radioactive Waste Management: 1983, A Study of the Isolation System for Geologic Disposal of Radioactive Wastes (National Academy Press, Washington, DC).

^Blowers, Lowry, and Solomon, op. cit., pp. 318-19.

l^Murray, R. L.: 1983, Understanding Radioactive Waste (Batelle Press, Columbus), p. 127, p. 142; Blowers, Lowry, and Solomon, op. cit., p. 318.

"For discussion of the 1982 Nuclear Waste Policy Act, see US Congress: 1978, High-Level Nuclear Waste Issues, Hearings Before the Subcommittee on Nuclear Regulation of the Committee on Environment and Public Works, US Senate, 100th Congress, First Session, April 23. June 2. 3, 18. 1987 (US Government Printing Office, Washington, DC). See also US Congress: 1981, Radioactive Waste Legislation Hearings Before the Subcommittee on Energy and Environment of the Committee on Interior and Insular Affairs, 'House of Representatives, 97th Congress, First Session, June 23, 25; July 9, 1981 (US Government Printing Office, Washington, DC).

'^For discussion of the 1987 Act, see Raeber, J. D.: 1989, "Federal Nuclear Waste Policy as Defined by the Nuclear Waste Policy Amendments Act of 1987," Saint Louis University Law Journal 3A, no. 1 (Fall), pp. 111-31.

•^US DOE: 1991, Site Characterization Progress Report: Yucca Mountain, Nevada, DOE/ RW-0307P (US DOE, Washington, DC), p. xiv.

'''Swainston, H. W.: 1992, "Yucca Mountain: A Study of Conflicts in Federalism," Inter Alia 57, no. 1 (October), pp. 11-16. Sawyer, G. and the State of Nevada Commission on Nuclear

ENVIRONMENTAL RISK AND THE IRON TRIANGLE 773

Projects: 1992, Report of the Nevada Commission on Nuclear Projects (Nuclear Waste Project Office, Carson City), pp. 57-58. See also US Congress: 1991, Nuclear Waste Policy Amendment Act of 1991, Report of Mr. Johnston (US Govemment Printing Office, Washington, DC), pp. 1-5.

^'Bryan, R. H., Govemor of Nevada: 1987, "Statement," in US Congress, Nuclear Waste Program. Hearings Before the Committee on Energy and Natural Resources, US Senate, 100th Congress, April 29 and May 7, 1987, Part 3 (US Govemment Printing Office, Washington, DC), p. 41, provides the 80-percent figure. See also Yates, M.: 1990, "DOE Reassess Civilian Radioactive Waste Management Program," Public Utilities Fortnightly (February 15), pp. 36-38, esp. 36; Loux, R.: 1990, "Will the Nation's Nuclear Waste Policy Succeed at Yucca Mountain?," Public Utilities Fortnightly (November 22), p. 27, p. 52.

'^For discussion of sub-seabed disposal, see, for example, Kaplan, R. A.: 1991, "Into the Abyss: Intemational Regulation of Sub-seabed Nuclear Waste Disposal," University of Penn- sylvania Law Review, no. 3 (January), pp. 769-800; and US Congress: 1987, Civilian Radio- active Waste Disposal, Hearings Before the Committee on Energy and Natural Resources, 100th Congress, First Session, July 16, 17, 1987 {US Govemment Printing Office, Washington, DC), pp. 244ff., 309ff.

^^For data on the 80-percent opposition figure, see Bryan, op. cit., p. 41.

'^For the cost figures, see Johnston, J. B.: 1991, "Statement," in US Congress, Nuclear Waste Program (US Govemment Printing Office, Washington, DC), p. 147.

l%addeil, R. K., J. H. Robison, and R. K. Blankennagel: 1984, Hydrology of Yucca Mountain and Vicinity, Nevada -Califomia—Investigative Results through Mid-1983 (US Geological Survey, Water Resources Investigations Report 84-4267, Denver, CO).

^^Sinnock, S. et al.: 1986, Preliminary Estimates of Groundwater Travel Time and Radionu- clide Transport at the Yucca Mountain Repository Site, SAND 85-2701 (Sandia National Labs., Albuquerque, NM), p. i.

^'US DOE: 1986, Nuclear Waste Policy Act, Environmental Assessment, Yucca Mountain Site, Nevada Research and Development Area, Nevada, DOE/RW-0073, 3 vols. (US DOE, Washington, DC), vol. 2, pp. 06-165.

., vol. 2, pp. 06-167. examples of such difficulties, see Hamilton, L., D. Hill, M. D. Rowe, and E. Stern:

1986, Toward a Risk Assessment of the Spent Fuel and High-Level Nuclear Waste Disposal System, Contract DE-AC02-76CH00016 (US DOE, Washington, DC), pp. 09-12. See also Shrader-Frechette, K. S.: 1993, Burying Uncertainty: Risk and the Case Against Geological Disposal of Nuclear Waste (University of Califomia Press, Berkeley), chs. 4-7.

'̂̂ Lovins is quoted in Bates, A. K.: 1988, "The Karma of Kerma: Nuclear Wastes and Natural Rights," Environmental Law and Litigation 3 (Nov.), p. 19.

^'Peters, R., J. H. Gauthier, and A. L. Dudley: 1985, "Effect of Percolation Rate on Water-Travel Time in Deep, Partially Saturated Zones," in Symposium on Groundwater Flow and Transport Modeling for Performance Assessment of Deep Geologic Disposal of Radio- active Waste (Sandia National Labs., Albuquerque, NM), Item 227 in US DOE, DE88OO4834.

^^Board on Radioactive Waste Management, US NAS: 1990, Rethinking High-Level Radio- active Waste Disposal (National Academy Press, Washington, DC), p. v; see also p. 27.

^''Travis, B., S. W. Hodson, H. E. Nuttall, T. L. Cook, and R. S. Rundberg: 1984, Preliminary Estimates of Water Flow and Radionuclide Transport in Yucca Mountain (Los Alamos National Lab, Los Alamos, NM), pp. 03-04.

^*Dudley, A., R. Peters, J. Gauthier, M. Wilson, M. Tierney, and E. Klavetter: 1988, Total System Performance Assessment Code {TOSPAC): Volume 1, Physical and Mathematical Bases: Yucca Mountain Project, SAND85-0(X)2 UC-70 (Sandia National Labs.. Albuquerque. NM), Item 182 in US DOE, DE9OOO6793, p. 92.

774 BUSINESS ETHICS QUARTERLY

vis, Hodson, Nuttall, Cook, and Rundberg, op. cit., p. 16. ., p. 25; Sinnock et ai, op. cit., p. i.

DOE: 1985, Tectonic Stability and Expected Ground Motion at Yucca Mountain. Final Report. Revision 1. August 7-8, 1984-January 25-26, 1985 (Science Applications Intemational Corp., La Jolla, CA), Item 18 in US DOE, DE88004834; Emel, J., B. Cooke, R. Kasperson, H. Brown, R. Goble, J. Himmelberger, and S. Tuller: 1988, Risk Management and Organizational Systems for High-Level Radioactive Waste Disposal: Issues and Priorities, NWPO-SE-008-88 (Carson City, NV: State of Nevada, Agency for I^ojects/Nuclear Waste Project Office, Carson City, Nevada), September; Emel, J., R. Kasperson, R. Goble, and O. Rennet: 1988, Postclosure Risks at the Proposed Yucca Mountain Repository: A Review of Methodological and Technical Issues, NWPO-SE-011-88 (State of Nevada, Agency for Nuclear Projects/Nuclear Waste Pro- ject Office, Carson City, NV), June.

^^O'Brien, P.: 1977, Technical Support for High-Level Radioactive Waste Management, Task C Report: Assessment of Migration Pathways, EPA 520/4-79-997C (US EPA, Washington, DC), p. 68.

33NRC: 1987, "Staff Comments," in US Congress, Nuclear Waste Program, Hearings Before the Committee on Energy and Natural Resources, US Senate, 100th Congress, First Session on the Current Status of the Department of Energy's Civilian Nuclear Waste Activities, January 29, February 4 and 5, 1987, Part 1 (US Government Printing Office, Washington, DC), p. 204. For the NAS claim, see Board, NAS, op. cit., p. 4.

^•^Rusche, B.: "Statement," in US Congress, Nuclear Waste Program (US Government Printing Office, Washington, DC), p. 917.

DOE, NWPA-Yucca, op. cit., vol. 2, pp. 06-78.

id., vol. 2, pp. 06-334, pp. 06-335.

et al., op. cit., p. 58.

r, S., and R. Jacobson: 1987, Chemistry of Groundwater in Tuffaceous Rocks, Central Nevada, NWPO-TR-006-87 (State of Nevada, Agency for Projects/Nuclear Waste Project Office, Carson City, NV), January, p. 72.

^^Stephens, K., L. Boesch, B. Crane, R. Johnson, R. Moler, S. Smith, and L. Zaremba: 1986, Methodologies for Assessing Long-Term Performance of High-Level Radioactive Waste Pack- ages, NUREG/CR-4477 ATR-85(5810-01)IND (US NRC, Division of Waste Management, Office of Nuclear Material Safety and Safeguards, Washington, DC), January, p. xvi, p. 8-2.

^'^Maione, C. 1990: "Geologic and Hydrologic Issues Related to Siting a Repository for High-Level Nuclear Waste at Yucca Mountain, Nevada, USA," Journal of Environmental Management 30, p. 381; Brown, D. and J. Lemons: 1990, "Scientific Certainty and the Laws That Govem Location of a Potential High-Level Nuclear Waste Repository," Environmental Management 15, no. 3, p. 319.

'*'Lemons, J. and D. Brown: 1990, 'The Role of Science in the Decision to Site a High-Level Nuclear Waste Repository at Yucca Mountain, Nevada, USA," The Environmentalist 10, no. 1, p. 10.

'*2Emel, J., B. Cook, R. Kasperson, and O. Renn: 1990, Nuclear Waste Management: A Comparative Analysis of Six Countries, NWPO-SE-034-90 (State of Nevada, Agency for Projects/Nuclear Waste Project Office, Carson City, NV), November, p. 5.

^^Hunter, R., and C. Mann: 1989, Techniques for Determining Probabilities of Events and Processes Affecting the Performance of Geologic Repositories, NUREG/CR-3964 SAND86- 0196, vol. 1, June (US NRC, Division of High-Level Waste Management, Office of Nuclear Material Safety and Safeguards, Washington, DC), p. 1.

ENVIRONMENTAL RISK AND THE IRON TRIANGLE 775

., p. 2.

'*^Thompson Engineering Company: 1988, Review and Comment on the US Department of Energy Site Characterization Plan Conceptual Design Report. NWPO-TR-009-88 (State of Nevada, Agency for Projects/Nuclear Waste Project Office, Carson City, NV), item 329 in US DOE. DE90006793 (October), p. 13.

^Malone, C: 1989, "The Yucca Mountain Project," Environmental Science and Technology 23, no. 12, p. 1453. For the utility-industry claim, see Yates, M.: 1990, "Council Report Finds High-Level Nuclear Waste Repository Rules 'Unrealistic'," Public Utilities Fortnightly (Au- gust 16), pp. 40-41. For the NAS worries, see Board, NAS, op. cit.

'̂ ''Younker, J. L., S. L. Albrecht, W. J. Arabasz, J. H. Bell, F W. Cambray, S. W. Carothers, J. I. Drever, J. T. Einaudi, D. E. French, K. V. Hodges, R. H. Jones, D. K. Kreamer, W. G. Pariseau, T. A. Vogel, T. Webb, W. B. Andrews, G. A. Fasano, S. R, Mattson, R. C. Murray, L. B. Ballou, M. A. Revelli, A. R. Ducharme, L. E. Shephard, W. W. Dudley, D. T. Hoxie, R. J. Herbst, E. A. Patera, B. R. Judd, J. A. Docka, L. R. Rickertsen, J. M. Boak, and J. R. Stockey: 1992, Report of the Peer Review Panel on the Early Site Suitability Evaluation of the Potential Repository Site at Yucca Mountain, Nevada, SAIC-91/8001 (US DOE, Washington, DC), p. B-2: hereafter cited as Younker, Albrecht, et al.

^^Fairhurst, C: 1990, Board on Radioactive Waste Management, National Research Council, in The Federal Program for the Disposal of Spent Nuclear Fuel and High-Level Radioactive Waste, Hearing Before the Subcommittee on Nuclear Regulation of the Committee on Environ- ment and Public Works, US Senate, 101st Congress (US Govemment Printing Office, Wash- ington, DC), p. 35.

'̂ ^Thompson Engineering Company, op. cit., p. 5. et al, op. cit., p. 57.

p. i.

R.: 1986, The Effect of Percolation Rate on Water Travel in Deep, Partially Saturated Zones, SAND85-0854 (Sandia National Labs., Albuquerque, NM), p. i.

^^Reichard, E., C. Cranor, R. Raucher, and G. Zapponi: 1990, Groundwater Contamination Risk Assessment (Intemational Association of Hydroiogical Sciences, Oxfordshire, England), p. 101.

''*Bryan, R.: 1985, State of Nevada Comments on the US Department of Energy Draft Environmental Assessment for the Proposed High-Level Nuclear Waste Site at Yucca Moun- tain, 2 vols. (Nuclear Waste Project Office, Office of the Govemor, Carson City, NV), vol. 1, p. 1-42, p. 1-43; Peters, op. cit., p. 32; see Sawyer, G.: 1987, "Statement," in US Congress, Nuclear Waste Program, Hearings Before the Committees on Energy and Natural Resources, US Senate, 100th Congress, First Session on the Current Status of the Department of Energy's Civilian Nuclear Waste Activities, January 29, February 4 and 5, 1987, Part 1 (US Govern- ment Printing Office, Washington, DC), p. 709, p. 712.

et al., op. cit., p. 58, p.75. pp. i-ii.

, S. and T. Lin: 1984, Preliminary Bounds on the Expected Postclosure Perform- ance of the Yucca Mountain Repository Site, Southem Nevada, SAND84-1492 (Sandia National Labs., Albuquerque, NM), p. 41.

p. 37. p. 53. p. 41. p. 53.

776 BUSINESS ETHICS QUARTERLY

"^Jacobson, E.: 1985, Investigation of Sensitivity and Uncertainty in Some Hydrologic Models of Yucca Mountain and Vicinity, SAND84-7212 (Sandia National Labs., Albuquerque, NM), p. 90.

and Lin, op. cit., p. 29.

and Lin, op. cit., p. 24; Smith, C. B., D. J. Egan, Jr., W. A. Williams, J. M. Gnihlke, and C-Y. Hung, and B. L. Serini: 1982, Population Risks from Disposal of High-Level Radio- active Wastes in Geologic Repositories, EPA-520/3-80-006 (US EPA, Washington, DC), p. 91.

and Lin, op. cit., p. 16.

p. 37.

et al, op. cit., p. 58.

p 75.

^'^Ibid., p. 11.

'^^Ibid., p. 80.

^^Smith, Egan, Williams, Gruhlke, Hung, and Serini, op. cit., p. 49.

^hbid.,p. 183. '^Sinnock et al, op. cit., p. 77.

'"^Thompson, F. L., F. H. Dove, and K. M. Krupka: 1984, Preliminary Upper-Bound Conse- quence Analysis for a Waste Repository at Yucca Mountain, Nevada, SAND83-7475 (Sandia National Labs., Albuquerque, NM), pp. v-vi.

^^Dudley, Peters, Gauthier, Wilson, Tierney, and Klavetter, op. cit., p. 56. '^See, for example, Younker, J. L., W. B. Andrews, G. A. Fasano, C. C. Herrington, S. R.

Mattson, R. C. Murray, L. B. Ballou, M. A. Revelli, A. R. Ducharme, L. E. Shephard, W. W. Dudley, D. T. Hoxie, R. J. Herbst, E. A. Patera, B. R. Judd, J. A. Docka, and L. R. Rickertsen: 1992, Report of Early Site Suitability Evaluation of the Potential Repository Site at Yucca Mountain, Nevada, SAIC-91/8000 (US Department of Energy, Washington, DC), pp. 2-94, pp. 2-163; hereafter cited as: Younker, Andrews, et al.

'^Younker, Aibrecht, et al., op. cit., p. 25.

id., p. 2\A.

,d., p. 214.

^'Younker, Andrews, et al, op. cit., pp. 2-121, pp. 1-3.

82/{,W., pp. 01-09.

^^Ibid., pp. 2-m. 8'*US EPA: 1973, Report to Congress on Hazardous Waste Disposal (US Govemment

Printing Office, Washington, DC), June, p. 133. ^^Papadopulos, S., and I. Winograd: 1974, Storage of Low-Level Radioactive Wastes in the

Ground: Hydrogeologic and Hydrochemical Factors, EPA-520/3-74-009 (US EPA, Office of Radiation Programs, Washington, DC), p. 29, p. 33.

^^Zehner, H.: 1979, Preliminary Hydrogeologic Investigation of the Maxey Flats Radioactive Waste Burial Site, USGS 79-1329 (US Department of the Interior, USGS, Louisville, KY), pp. 48-52.

^'See Board, NAS, op. cit., p. 4. See also Radin, A., Chair: 1989, Monitored Retrievable Storage Review Commission, Nuclear Waste: Is There a Need for Federal Interim Storage? (US Government Printing Office, Washington, DC), p. 103, p. 10, p. xvii.

s, op. cit., p. 13.

ENVIRONMENTAL RISK AND THE IRON TRIANGLE 777

DOE: 1986, Nuclear Waste Policy Act, Environmental Assessment, Reference Reposi- tory Location, Hanford Site, Washington, 3 vols., DOE/RW-0070 (US DOE, Washington, DC), vol. 2, pp. 6-148.

^ a r r y , B.: 1991, Liberty and Justice (Clarendon Press, Oxford), pp. 271-73; See Shrader- Frechette, op. cit., esp. chs. 8-10.

'Wates, M.: 1990, "DOE Reassesses Civilian Radioactive Waste Management Program," PubUc Utilities Fortnightly (February 15), pp. 36-38.

'^Parker, F. L. etal.: 1990, Board on Radioactive Waste Management, US National Research Council, Rethinking High-Level Radioactive Waste Disposal (National Academy Press, Wash- ington, DC), p. 1, p. 4, p. 6.

^ ^ i n , op. cit., p. D3, p. D18.

., p. D5, see also p. D4, p. D18.

., p. D6, p. D7, p. D18.

id., p. D8, p. D9, p. D18.

jrf., p. DIO, p. D18.

id., p. Dl l , p. D12, p. D18. See Milnes, A.: 1985, Geology and Radwaste (Academic Press, New York), pp. 286ff. See also US NAS: 1980, A Review of the Swedish KBS-II Plan for Disposal of Spent Nuclear Fuel (US NAS, Washington, DC); and Nyquist, C. E.: 1987, "Nuclear Waste Disposal in Sweden," Public Utilities Fortnightly (May 14), pp. 34-35.

in, op. cit., pp. D13-D15, p. D18. id., pp. m5-D\%.

^^^Ibid., p. D\7.

'"^See, for example, Shrader-Frechette, K.: 1991, Risk and Rationality (University of Califomia Press, Berkeley, CA), chap. 9.

^̂ •̂ See Sinnock et al., op. cit., p. 80.

^*^Shrader-Frechette, Risk and Rationality, op. cit., pp. 132-34.

^*^^Shue, H.: 1981, "Exporting Hazards," in P. Brown and H. Shue, eds.. Boundaries: National Autonomy and Its Limits (Rowman and Littlefield, Totowa, NJ), pp. 107-45; Lichten- berg, J.: 1981, "National Boundaries and Moral Boundaries," in P. Brown and H. Shue, eds.. Boundaries: National Autonomy and Its Limits (Rowman and Littlefield, Totowa, NJ), pp. 79-100.

'°^See, for example, Becker, L.: 1984, "Rights," in L. Becker and K. Kipnis, eds.. Property (Prentice-Hall, Englewood Cliffs, NJ), p. 76. For a discussion of the flaws in this view of rights, see Baier, A.: 1986, 'Poisoning the Wells,' in R. MacLean, ed.. Values at Risk (Rowman and AUenheld, Totowa, NJ), pp. 49-74.

^^^Shrader-Frechette, Risk and Rationality, op. cit., pp. 136-37.

id., pp. 137-38.

id., pp. 138-39. ' '"For a discussion of this argument, see Frankena, W. K.: 1962, "Concept of Social Justice,"

in R. Brandt, ed.. Social Justice (Prentice-Hall, Englewood Cliffs, NJ), p. 15; Shrader-Frechette, Risk and Rationality, op. cit., chap. 8.

__MACOSX/._env+risk+yucca+mt.pdf

fecundity+as+basis+for+risk+assessment+in+mulloscs.pdf

Fecundity as a Basis for Risk Assessment of Nonindigenous Freshwater Molluscs REUBEN P. KELLER,∗‡ JOHN M. DRAKE,∗§ AND DAVID M. LODGE†∗∗ ∗Department of Biological Sciences, University of Notre Dame, Notre Dame, IN 46556, U.S.A. †National Center for Ecological Analysis and Synthesis, 735 State Street Suite 300, Santa Barbara, CA 93101, U.S.A.

Abstract: The most efficient way to reduce future damages from nonindigenous species is to prevent the introduction of harmful species. Although ecologists have long sought to predict the identity of such species, recent methodological advances promise success where previous attempts failed. We applied recently developed risk assessment approaches to nonindigenous freshwater molluscs at two geographic scales: the Laurentian Great Lakes basin and the 48 contiguous states of the United States. We used data on natural history and biogeography to discriminate between established freshwater molluscs that are benign and those that constitute nuisances (i.e., cause environmental and/or economic damage). Two statistical techniques, logistic regression and categorical tree analysis, showed that nuisance status was positively associated with fecundity. Other aspects of natural history and biogeography did not significantly affect likelihood of becoming a nuisance. We then used the derived statistical models to predict the chance that 15 mollusc species not yet in natural ecosystems would cause damage if they become established. We also tested whether time since establishment is related to the likelihood that nonindigenous mollusc species in the Great Lakes and United States would cause negative impacts. No significant relationship was evident at the U.S. scale, but recently established molluscs within the Great Lakes were more likely to cause negative impacts. This may reflect changing environmental conditions, changing patterns of trade, or may be an indication of “invasional meltdown.” Our quantitative analyses could be extended to other taxa and ecosystems and offer a number of improvements over the qualitative risk assessments currently used by U.S. (and other) government agencies.

Keywords: biological invasion, ecological forecasting, ecological prediction, mollusk, risk analysis, risk assess- ment

La Fecundidad como Base para la Evaluación de Riesgo de Moluscos Dulceacúıcolas No Nativos

Resumen: La prevención de la introducción de especies perjudiciales es la manera más eficiente de reducir los daños futuros ocasionados por especies no nativas. Aunque los ecólogos han buscado predecir la identidad de tales especies durante mucho tiempo, avances metodológicos actuales prometen éxito en donde han fallado intentos anteriores. Aplicamos métodos de evaluación de riesgo, desarrollados recientemente, en moluscos dul- ceacuı́colas en dos escalas regionales: la cuenca Laurentian de Grandes Lagos y los 48 estados contiguos de los Estados Unidos. Utilizamos datos de historia natural y biogeograf́ıa para discriminar moluscos dulceacuı́colas establecidos que son benéficos de los que son perjudiciales (i.e., causan daño ambiental y/o económico). Dos técnicas estadı́sticas, regresión loǵıstica y análisis de árbol categórico, mostraron que el estatus perjudicial estaba asociado positivamente con la fecundidad. Otros aspectos de la historia natural y biogeograf́ıa no alteraron la probabilidad de convertirse en perjudicial. Posteriormente utilizamos los modelos estadı́sticos derivados para predecir la probabilidad de que 15 especies de moluscos que aun no están en ecosistemas naturales pudieran causar daños en caso de establecerse. También probamos si el tiempo transcurrido desde el establecimiento está relacionado con la probabilidad de que especies de moluscos no nativos en los Grandes Lagos y en Estados Unidos pudieran causar impactos negativos. No hubo relación significativa evidente en

‡email [email protected] §Current address: Institute of Ecology, University of Georgia, Athens, GA 30602, U.S.A. ∗∗Current address: Department of Biological Sciences, University of Notre Dame, Notre Dame, IN 46556, U.S.A. Paper submitted December 5, 2005; revised manuscript accepted May 15, 2006.

191

192 Fecundity Predicts Mollusc Impacts Keller et al.

la escala de E. U. A., pero los moluscos recientemente establecidos en los Grandes Lagos tuvieron mayor prob- abilidad de provocar impactos negativos. Esto puede ser reflejo de condiciones ambientales cambiantes, de patrones de comercio cambiantes o puede ser un indicador de una “fundición invasiva.” Nuestros análisis cuantitativos podŕıan ser extendidos a otros taxa y ecosistemas y ofrecen numerosas mejoŕıas de las evalua- ciones de riesgo cualitativas que actualmente son utilizadas por agencias gubernamentales de E.U.A (y otros paı́ses).

Palabras Clave: análisis de riesgo, evaluación de riesgo, invasión biológica, molusco, predicción ecológica, pronóstico ecológico

Introduction

Nonindigenous freshwater molluscs cause decreased agri- cultural (Lach & Cowie 1999) and utility production (Mackie 2000), increased health risks to humans, live- stock, and wildlife (WHO 2002), and are an important threat to native biodiversity (Ricciardi et al. 1998). World- wide, the economic and environmental costs of non- indigenous freshwater molluscs are increasing as grow- ing numbers of species are transported beyond their native ranges (Mills et al. 1993; Cowie 1998; Levine & D’Antonio 2003). The principal pathways of this move- ment include ships’ ballast water and the pet and live-food trades (Cowie & Robinson 2003). Although the majority of introduced species never establish reproducing pop- ulations and the majority of those that establish do not cause impacts (Williamson 1996), the rate of discovery of established species is increasing, presumably due to increasing pathway traffic (Cohen & Carlton 1998) or be- cause of accumulating lags between introduction and dis- covery (Solow & Costello 2004).

Measures to prevent the spread of all freshwater molluscs—essentially ceasing trade—would bring enor- mous costs to society. An alternative, especially for inten- tional pathways such as the pet and live-food trades, is to predict which species are most likely to cause negative impacts and concentrate resources on preventing those from entering pathways. Particularly, quantitative meth- ods based on information that can be easily identified in advance should be used to estimate the probability that introduced species will become a nuisance.

Nevertheless, a long history of work, including that by Baker (1974), suggests that lists of characteristics thought to confer negative impacts do not provide a sound basis for predictions. Although some generalizations exist that apply to a wide variety of taxa (Williamson 1996; Kolar & Lodge 2001), they are not sufficiently specific and robust to serve as a basis for predictions. Thus, some ecologists conclude that the invasion process is inherently too com- plicated for future nuisance species to be predicted with worthwhile accuracy (Williamson 1999).

In contrast recent approaches recognize that risk anal- yses must be more limited geographically and taxonom- ically and should be limited to specific “invasion steps” (Kolar & Lodge 2001). This approach explicitly acknowl-

edges that to become a nuisance a species must pass through three steps: it must be transported; it must estab- lish in a new range; and it must spread and cause harm. For example, on the basis of 24 characteristics of fishes introduced to the Great Lakes, Kolar and Lodge (2002) predicted with ∼90% accuracy the identity of introduced species that became established and the identity of es- tablished species that caused negative impacts. Similar success was achieved in predicting the identity of woody plant invaders in the United States (Reichard & Hamil- ton 1997), and strong correlations have been found be- tween biological traits and likelihood of negative impacts for nonindigenous birds in New Zealand (Veltman et al. 1996), nonindigenous fishes in California (Marchetti et al. 2004), and nonindingenous conifers worldwide (Richard- son & Rejmánek 2004).

We extended this approach in three ways. First, we ap- plied the methods to molluscs, a taxonomic group for which relatively little natural history data are available. This allowed us to test whether these methods are robust when only small amounts of data are available to explain species impact. Second, we applied these methods to a smaller data set than has been done previously. Statistical methods work best with large sample sizes, and our anal- yses tested whether the risk analysis methods described are robust to low numbers of species. This is important because in many taxonomic groups only a small number of species have become established beyond their native range. If quantitative risk assessments are to be applied to these groups, then knowing the robustness of these meth- ods to small sample sizes is essential. Finally, we compared the results from logistic regression and a categorical and regression tree (CART) approach.

We applied these statistical discrimination techniques to nonindigenous freshwater molluscs in the Laurentian Great Lakes basin and in the 48 contiguous states of the United States, of which the Great Lakes is a subset. We examined the final step in the invasion process—from established to nuisance—because this is the important economic and ecological step, and because data on the earlier steps are unavailable. We tested longstanding hy- potheses about which characteristics are related to im- pact and show that risk assessment for nuisance mollusc species in the United States may require very little natural history data.

Conservation Biology Volume 21, No. 1, February 2007

Keller et al. Fecundity Predicts Mollusc Impacts 193

Methods

We classified all established nonindigenous freshwater mollusc species as either nuisance or benign. Nuisance species are those for which we could find credible evi- dence of environmental and/or economic harm. We ac- cepted published scientific literature or unpublished sci- entific data made available by other scientists as credible evidence. We classified as benign all established species for which we could find no evidence of negative impacts. Thus, we have attempted to identify species that society might have made efforts to prevent from becoming es- tablished had their eventual impacts been known. Our definition of nuisance is similar to some definitions of invasive (e.g., Kolar & Lodge 2001), but to avoid ambigu- ity we do not use the term invasive (e.g., Richardson et al. 2000; Colautti & MacIsaac 2004). From a societal per- spective, our distinction between nuisance and benign is the most relevant distinction for determining whether or not a species proposed for import should be permitted.

From a literature search we determined that 18 species of nonindigenous molluscs are established in the Great Lakes basin (Mills et al. 1993; Turgeon et al. 1998; Grig- orovich et al. 2000). Four of these are native to other North American drainages (Elimia virginica, Gillia al- tilis, Lasmigona subviridis, Viviparus georgianus, Table 1) (Mills et al. 1993). For each nonindigenous species we conducted a literature search for natural history character- istics. Because the natural history literature on molluscs is poor for most species, we occasionally asked experts to provide unpublished data. Despite these efforts lack of in- formation forced us to eliminate from our analysis several characteristics that we originally hoped to include (e.g., growth rate, tolerances for pH, temperature, oxygen, and all other metrics of environmental tolerance). Our anal- yses were thus based on eight natural history traits for which adequate data existed (Table 2). In addition, we collected data on the time since establishment for each established mollusc species (Table 2).

Two statistical discrimination techniques were used for data analysis. First, we tested for relationships between the natural history of a species and its impact with CART approach (Therneau & Atkinson 2005). Because our re- sponse data were binary (nuisance/benign), we created categorical trees. CART works by finding the split in one of the available predictor variables that maximizes the within-group homogeneity of the two groups produced (De’ath & Fabricius 2000). Any predictor variable could be used to make this split, and further splits within the resulting groups are made until the user-defined limit tree size is reached. CART is nonparametric, does not assume normality, is relatively robust to the distribution for pre- dictor variables, and operates on both categorical and continuous data (De’ath & Fabricius 2000). CART’s split- ting points are placed at the midpoint between the two cases where the best split exists. Hence, if the split occurs

Table 1. Established nonindigenous molluscs in the Laurentian Great Lakes and 48 contiguous states of the United States.

Family and species Impacts Fecunditya

Ampullariidae Marisa cornuarietis environmental 1711 Pomacea bridgesi Pomacea canaliculata economic 4355 Pomacea haustrum

Bithyniidae Bithynia tentaculatab economic 174

Corbiculidae Corbicula flumineab economic, 68,678

environmental Dreissenidae

Dreissena bugensisb economic, 960,000 environmental

Dreissena polymorphab economic, 960,000 environmental

Hydrobiidae Gillia altilisb,c

Potamopyrgus environmental 230 antipodarumb

Lymnaeidae Radix auriculariab 1300

Physidae Stenophysa marmorata Stenophysa maugeriae

Planorbidae Biomphalaria glabrata 356 Drepanotrema aeruginosus Drepanotrema cimex Drepanotrema kermatoides

Pleuroceridae Elimia virginicab,c

Sphaeriidae Pisidium amnicumb 10 Pisidium henslowanumb 5.8 Pisidium moitesserianumb 3.1 Pisidium supinumb 12 Sphaerium corneumb 62

Thiaridae Melanoides tuberculata environmental 365 Melanoides turriculus Tarebia granifera 213

Unionidae Lasmigona subviridisb,c

Valvatidae Valvata piscinalisb 150

Viviparidae Cipangopaludina chinensisb 65 Cipangopaludina japonicab 65 Viviparus georgianus b,c 39

aNumber of eggs or live offspring produced/female/year. bEstablished in Laurentian Great Lakes. cNative to the United States but not the Laurentian Great Lakes.

between cases with values of, for example, 10 and 20, the split will be at 15. The true split could be anywhere be- tween 10 and 20, however. More data in this range would be required for a better estimate.

All natural-history variables (eight variables, Table 2) were used as input data for the CART model. Each

Conservation Biology Volume 21, No. 1, February 2007

194 Fecundity Predicts Mollusc Impacts Keller et al.

Table 2. Variables, including mollusc natural-history traits and time since introduction, analyzed to determine whether differences exist between nuisance and benign, nonindigenous species in the Laurentian Great Lakes.a

Levels (categorical) or Variable Type units (continuous)

Type of reproduction categorical sexual, sequential hermaphrodite, simultaneous hermaphrodite, parthenogenetic

Egg brooding categorical yes/no Maximum size continuous mm Fecundityb continuous number of eggs or

live offspring released/female/year

Longevity continuous years Non-native elsewhere categorical yes/no Latitude range continuous max. latitude minus

min. latitude Larval stage categorical yes/no Time since continuous year of first

introduction occurrence

aFull data for all variables available on request from R.P.K. bAll individuals of parthenogenetic and simultaneous hermaphrodite species were considered females.

resulting tree was assessed for accuracy on the basis of its misclassification rate (i.e., proportion of species that the derived models would have assigned the wrong nuisance status to) and splitting was based on recursive partitioning with the Gini index (De’ath & Fabricius 2000). Because of the small sample size, and to avoid overfitting, we set the minimum number of species allowed at any tree node to three. In addition, we used jack-knife analyses to test the derived CART models by removing one species from the full list and constructing a CART model with the re- maining data. This was repeated for each species, and the jack-knife misclassification rate was the proportion of times the removed species would be incorrectly clas- sified. The effect of time since establishment on species nuisance status was assessed with the same methods.

Second, we tested for a relationship between impact status and species natural history with logistic regression, which is used to find the probability of an event occurring (in our case, a species becoming nuisance) on the basis of the level of some predictor variable(s). Input variables can be either continuous or categorical, with continuous variables assumed to have a normal distribution. Advan- tages of logistic regression are that the relative risks posed for all levels of the predictor variable can be assessed, and confidence intervals can be calculated. As above, we used eight natural-history variables as input (Table 2). Fecundity was log10 transformed to meet regression as- sumptions. Predictive power was not improved by simul- taneously including multiple explanatory variables in the model, so this approach was not pursued. Instead, we used likelihood ratio tests to assess the null hypothesis

that each natural-history character did not affect impact status (Quinn & Keough 2002).

For logistic regression we used an alpha level of 0.05, and for each significant result we calculated the receiver operating characteristic (ROC) curve to determine accu- racy (Fielding & Bell 1997). An ROC output (area under the curve) of c = 0.5 indicates that the derived logistic curve is no more accurate than tossing a coin: c = 1.0 indicates perfect accuracy, and values above c = 0.7 indi- cate a good fit between model and data. Species for which data were not available were excluded from analysis. We also used logistic regression to test the null hypothesis that time since establishment is not related to nuisance status.

To broaden our approach we expanded our species list, based on Turgeon et al. (1998), to include all 27 nonindigenous freshwater molluscs in the 48 contiguous states (Table 1). We excluded Physella acuta on the ba- sis of recent work suggesting that it is probably native to the United States (Dillon et al. 2002). Fourteen species from the Great Lakes data set described earlier (exclud- ing species native to the United States but not the Great Lakes) formed a subset of this larger group (Table 1). We searched for estimates of annual fecundity and time since establishment (the only significant variables from our first analysis, see Results) for these 27 species, but were able to obtain fecundity data for only 19 species (Table 1). For time since establishment, we obtained data for 23 species. In categorizing species as nuisance or benign and in con- ducting logistic regression and CART, we used the same methods as for the Great Lakes analysis.

Unfortunately, our overall data set was too small to con- duct phylogenetically independent contrasts (Rejmánek et al. 2005). Instead, we conducted sensitivity analyses to determine whether any one family was driving the sig- nificance of the logistic regression results by iteratively eliminating one family from the data set and recalculat- ing the logistic relationship between natural history and impact status.

Using the logistic and CART models derived from the Great Lakes and U.S. analyses, we estimated the prob- ability of becoming a nuisance for 15 species not yet present in the United States (Table 3). We selected these 15 species for predictive analysis because reliable fecun- dity data were available. Each species was assessed for likely impact status for both the Great Lakes and the 48 contiguous states.

Because the 48 contiguous states encompass a broad range of climates and habitats, we assumed that all 15 test species could become established somewhere in the United States. For each species tested we also used available information (especially the known geographical range) to predict whether it could survive in the Great Lakes (Table 3). Specifically, if a species’ distribution was limited to tropical and/or subtropical conditions, it was considered incapable of establishing in the Great Lakes.

Conservation Biology Volume 21, No. 1, February 2007

Keller et al. Fecundity Predicts Mollusc Impacts 195

Table 3. Estimated probabilities of 15 mollusc species becoming invasive in the Laurentian Great Lakes and the 48 contiguous United States should they become established.a

Great Lakes United States

Species logistic CART logistic CART

Ancylus fluviatilis 0.107 no 0.090 no Indoplanorbis exustusb 0.859 yes 0.940 yes Physa fontinalis 0.257 yes 0.280 yes Lymnaea natalensisb 0.874 yes 0.950 yes Lymnaea palustris 0.355 yes 0.414 yes Lymnaea peregra 0.650 yes 0.772 yes Planorbis contortus 0.085 no 0.067 no Biomphalaria alexandrinab 0.744 yes 0.858 yes Biomphalaria stramineab 0.687 yes 0.808 yes Biomphalaria pfeifferib 0.912 yes 0.970 yes Bulinus abyssinicusb 0.800 yes 0.900 yes Bulinus globosusb 0.688 yes 0.809 yes Bulinus tropicusb 0.917 yes 0.971 yes Bulinus truncatusb 0.657 yes 0.779 yes Pisidium hibernicum 0.061 no 0.043 no

aCART (categorical and regression tree) model gives yes/no prediction of whether a species will become invasive if established. bUnlikely (on the basis of environmental constraints not included in the statistical analysis) to become established in the Great Lakes region.

Results

Five nonindigenous molluscs were classified as nuisance in the Laurentian Great Lakes; three additional species were classified as nuisance at the level of the 48 con- tiguous states (Table 1). Of these eight nuisance species, three, two, and three species were categorized as nui- sances, respectively, on the basis of their environmen- tal and economic impacts, economic impacts only, and environmental harm only (Table 1). For the Great Lakes nuisance species were the bivalves Dreissena polymor- pha, Dreissena bugensis (see Mackie 2000 for a review of Dreissena spp. impacts), and Corbicula fluminea, which has fouled power station water intakes (W.P. Kovalak, per- sonal communication), and the gastropod Bythinia ten- taculata, which previously fouled water-supply systems throughout the Great Lakes (Baker 1898). Although C. flu- minea satisfies our definition of nuisance and is widely established in the southern Great Lakes basin, it is rarely recorded in the Great Lakes proper except at sites heated by thermal effluent. Hence, its ability to cause impact largely depends on industrial activities. Potamopyrgus antipodarum was included as a nuisance species based on its high densities (≤5653/m2, Zaranko et al. 1997), rapid spread through the Great Lakes (Zaranko et al. 1997; Grigorovich et al. 2003), and history of negative impacts in other temperate habitats in North America (Hall et al. 2003).

Three additional species were added to the list of nui- sance molluscs when the analysis was expanded to con- sider all 48 states. Marisa cornuarietis has caused large

changes in the macrophyte communities of some Texas rivers (Horne et al. 1992); Pomacea canaliculata has be- come established at a number of sites in the southern United States, including Texas rice-growing areas, where farmers are adjusting their management practices to con- trol populations (Howells & Smith 2002); and Melanoides tuberculata reduces native snail diversity in Utah streams (Rader et al. 2003).

Results from the CART analysis of natural-history char- acteristics for the Great Lakes were straightforward: species with fecundities >162 offspring/female/year were categorized as nuisances five out of six times, whereas species with lower fecundities were always cat- egorized as benign (Fig. 1a). No characteristics other than fecundity were included in the tree. This tree has an estimated misclassification rate of 1 in 15 (7%). Jack- knife analysis gave a misclassification rate of 20%, with the splitting point of fecundity ranging from a minimum of 119.5 (Valvata piscinalis removed) to a maximum of 190 (B. tentaculata removed). Although larger data sets are preferable for CART analysis, the simplicity of this re- sult (i.e., using only one predictor variable) suggests that overfitting was not a problem. Logistic regression also showed that annual fecundity was the only natural-history

Figure 1. Decision trees for determining whether a freshwater mollusc species will be a nuisance (i.e., cause economic and/or environmental harm) or benign, produced with categorical and regression tree analysis, for (a) the Laurentian Great Lakes basin and (b) the 48 contiguous states of the United States. For the Great Lakes analysis fecundity was one of eight predictor variables (Table 2) but the only one chosen by the model. For the U.S. analysis fecundity was the only predictor used. Fecundity is the annual number of eggs or live offspring released per female.

Conservation Biology Volume 21, No. 1, February 2007

196 Fecundity Predicts Mollusc Impacts Keller et al.

characteristic significantly predictive of nuisance status for the Great Lakes ( p = 0.002, c = 0.960; Fig. 2). As for CART, none of the other seven natural-history variables was significantly associated with nuisance status.

Time since establishment was significantly related to nuisance status in the Great Lakes based on logistic re- gression ( p = 0.022, c = 0.815), but in a counterintuitive direction; species that recently established were more likely to cause impacts. This result is supported by CART, which split the data once at 35.5 years ago. All four species that became established more recently than this (i.e., af- ter 1970) have become a nuisance, whereas only 1 out of 14 species established before 1970 was classified as nuisance, giving an estimated misclassification rate of 1 in 18 (6%). Under jack-knife analysis the split always oc- curred at 35.5, except when C. fluminea was removed and it became 31.5. The jack-knife misclassification rate was identical to that of the full model (6%).

The effect of fecundity on nonindigenous mollusc nui- sance status for the 48 contiguous states, estimated with CART, was identical to those for the Great Lakes and had a misclassification rate of 3 in 19 (16%) (Fig. 1b). All eight species with fecundities <162 were benign, whereas 8 of the 11 with higher fecundities were nuisance species. Jack-knife analysis gave a misclassification rate of 32%, with the splitting point ranging from a fecundity of 119.5 (Valvata piscinalis removed) to 1505.5 ( p. antipodarum or M. tuberculata removed). According to logistic regres-

Figure 2. Logistic curves (±95% CI) showing the relationship between fecundity and probability of becoming a nuisance (i.e., causing economic and/or environmental harm) for molluscs in the Laurentian Great Lakes (LGL) and 48 contiguous states of the United States. For the Great Lakes analysis eight predictor variables were used; fecundity was the only significant one. For the U.S. analysis, fecundity was the only variable used. Fecundity is the annual number of eggs or live offspring released per female.

sion fecundity was significantly related to impact for the contiguous 48 states ( p < 0.001, c = 0.923). The logis- tic relationships between probability of invasiveness and fecundity were similar for the Great Lakes and for the 48 contiguous states (Fig. 2). The close relationship between results for the two geographic regions is not surprising given that the Great Lakes data set is a subset (except for species native to the U.S. but not the Great Lakes, Ta- ble 1) of the U.S. data set. Hence, these should not be interpreted as independent analyses.

For the United States as a whole we failed to detect a significant effect of time since establishment on nuisance status (logistic regression, p = 0.238). This was consis- tent with results from CART in which species established for <47 or >78.5 years were classified as benign. Other species were classified as nuisance species and the tree had a high misclassification rate of 7 out of 23 (30%). We did not attempt further analysis of this result because it included multiple splits in a single continuous predic- tor variable, making any interpretation difficult, especially considering the small sample size (n = 23 observations). We therefore concluded that nuisance status was not sig- nificantly related to time since establishment at the scale of the 48 contiguous United States.

Sensitivity analyses based on logistic regression showed that no single family was driving the significance of the logistic relationship at either the Great Lakes or U.S. level. The greatest increase in p value for each geographic scale occurred when the family Dreissenidae was removed, but this did not affect significance at either the Great Lakes ( p = 0.025, c = 0.933) or continental U. S. ( p = 0.002, c = 0.897) scales.

According to the logistic model 14 of the 15 species not yet in the United States pose a substantial risk (>5% chance) of harm if they become established (Table 3). The CART model predicted that 12 of the 15 species would become nuisance species. Several species pose extremely high risks to the contiguous United States according to the logistic model, including Bulinus globosus, a gastro- pod species native to tropical Africa, Physa fontinalis, a gastropod native to Europe, and Indoplanorbus exus- tus, a gastropod native to the Asian subcontinent (Ta- ble 3). These species are also predicted to be nuisances by the CART model, but because CART produces sim- ple nodes we could not determine the relative likelihood of a species becoming a nuisance compared with other species within its node. For the Great Lakes the largest risks were posed by Physa fontinalis, a snail species native to Eurasia, and Planorbis contortus, a snail species native to Europe.

Nine of the 15 tested species have geographic ranges that suggested they would not establish in the Great Lakes (Table 3). Examples of these species are Biomphalaria straminea, a species native to the Caribbean (Pointier et al. 1991), and Indoplanorbus exustus, a species native to South and Southeast Asia that laboratory studies show

Conservation Biology Volume 21, No. 1, February 2007

Keller et al. Fecundity Predicts Mollusc Impacts 197

has a lower temperature tolerance of 15◦ C (Parashar & Rao 1985).

Discussion

Our results show that information about one relatively simple piece of natural history—fecundity—is sufficient to explain the likelihood of economic and/or environ- mental consequences of freshwater molluscs should they become established. The risk posed by a freshwater mol- lusc species likely to be transported in any pathway can be assessed based on its fecundity, with the clear impli- cation that high-risk species should be kept out of trade in pathways in which species are specifically marketed (e.g., aquarium trade, aquaculture trade, live-food trade). The implications for the management of risk in pathways that do not discriminate on the basis of species (e.g., bal- last, snails hitch-hiking on watergarden plants) are less straightforward. For all pathways, however, the relative risks should guide the magnitude of investment in both prevention and any post-establishment management.

The importance of fecundity as a predictor of impact is not surprising (e.g., Baker 1974; Kolar & Lodge 2001; Marchetti et al. 2004). Furthermore, McMahon (2002) suggested that invasive freshwater species are likely to have higher fecundities and lower resistance to environ- mental extremes than native species. Although data are not available to statistically test the resistance hypothesis, our results are the first to provide quantitative support that mollusc fecundity is positively linked to impact. In contrast, a number of species traits often linked to unde- sirable impacts (e.g., history of establishment beyond the native range, latitudinal range; Kolar & Lodge 2001) were not predictive for molluscs in the Great Lakes.

Our results also showed that for the Great Lakes (but not for the 48 contiguous United States) recently estab- lished molluscs are more likely to cause harm than species that have been established for a long time. This is surpris- ing because many nuisance species are known to have long lag times before population densities increase and they have impacts (Crooks & Soulé 1999). Nevertheless, we observe that recently introduced molluscs in the Great Lakes generally have high fecundities and vulnerable ju- venile stages (e.g., Dreissenid veligers). It is possible that these species have been entering pathways for a long time but have only recently been able to survive passage be- cause travel times, particularly of intercontinental ships, have been reduced. Such a pattern of increased harm from recently established species is also consistent with the “in- vasional meltdown” hypothesis (Simberloff & Von Holle 1999; Ricciardi 2001). Although time since establishment is significantly related to the likelihood that a species has undesirable impacts, this character is not useful for pre- dicting the outcome of future introductions. Our result

does, however, indicate that there are increasing benefits to be gained from preventing new species of freshwater molluscs becoming established in the Great Lakes.

Implications for Risk Assessment and Risk Management

The fact that two different statistical techniques related mollusc fecundity to nuisance status increases our confi- dence that this relationship can reasonably be applied in risk assessment. Because the two models give different types of predictions and because this will inevitably lead to some species being classified in different ways depend- ing on the model used, it is necessary to choose a pre- ferred model, or some combination of models, to apply as a predictive scheme. If one selects the logistic models, a risk threshold must be set. Determining the acceptable level of risk is a policy decision that must be informed by many considerations in addition to those presented here. If an acceptable risk of harm (should a species be- come established) is set at 1 in 20, then in the United States all species with fecundities >25/year should be ex- cluded (Fig. 2), implying prohibition of all 15 species we screened (Table 3). At the same acceptable risk of harm for the Great Lakes, the result is even more conservative: all species with fecundities >17/year should be excluded (Fig. 1). One in 20 is not a particularly conservative risk threshold compared with thresholds set for other environ- mental hazards. For example, California requires special labeling and handling of products that have a > 1:100,000 chance of causing human cancers over 70 years of expo- sure (OEHHA 2006). Finally, we reiterate that the 15 test species were chosen on the basis of availability of reliable fecundity data, and they may not be a representative sam- ple of species likely to be introduced to the United States in the future.

Although the ROC results (i.e., c values) indicate a strong relationship between fecundity and impact, the confidence intervals for these predictions were large (Fig. 2). Many, however, did not include probabilities of <0.05, which, as we suggest above, would be a reasonable up- per limit of acceptable risk. Hence, consideration of con- fidence limits on the estimated probabilities would still support unambiguous management recommendations. By contrast, application of the CART models at both spa- tial scales is straightforward: species with annual fecundi- ties >162 should be excluded and others permitted. This corresponds to a threshold of acceptable risk based on the logistic model for the Great Lakes of 25% and for the United States of 27%.

Although fecundity data are often not readily available, future risk assessments based on our results will be possi- ble. Once a threshold of acceptable risk is specified, sur- rogate data often will be sufficient to classify a species as acceptable or unacceptable. For instance, any broadcast spawners are likely to have a fecundity in the thousands

Conservation Biology Volume 21, No. 1, February 2007

198 Fecundity Predicts Mollusc Impacts Keller et al.

or greater and hence will always present a large risk. Limnoperna fortunei is an emerging invasive freshwater mollusc species in Asia and South America that is a broad- cast spawner, so our model predicts that it poses a high risk to the United States. Likewise, Anodonta woodiana is a broadcast spawner and a recent invader of Europe. Although this latter species has a glochidial juvenile stage that develops by attaching to specific fish hosts, many of these hosts are present in the United States, so we con- sider that it also poses large risks. Similarly, apple snails (family Ampullariidae) in the genus Pomacea generally have minimum clutch sizes of 100 eggs. Given that they lay many clutches in a season, any species in this genus will pose a risk far >5% in the logistic model and will likely fall above the CART model threshold of 162. Hence, a broader taxonomic approach, or in some cases records of a single clutch from the species of interest, will often be sufficient to determine whether a species poses an unacceptable risk.

Ultimately, deciding on an appropriate model or combi- nation of models for risk assessment depends on the pri- orities of the management agencies involved. Our models can be used singly or in combination, and we recognize that factors other than fecundity will need to be included in a working risk assessment. Although some of these ad- ditional factors will be subjective, the basis of the risk assessment would remain quantitative, making the entire process more defensible than many current risk assess- ments, which rely almost entirely on the judgment of the user (e.g., Orr 2003).

It is not clear how our models would perform if ap- plied to molluscs in regions other then the United States, and such a test would require a data-gathering and mod- eling effort similar to ours. The results would, however, be enlightening, first, because they would determine the generality of the models presented here. Our models were developed and validated based on U.S. data, so such ad- ditional work would test whether fecundity is related to nuisance status in other regions. In addition, if fecundity were found to be associated with nuisance status in other regions, it would provide a strong validation of the mod- els developed here. Until our models are tested for other regions it should not be assumed that they apply to mol- luscs outside the United States.

It would also be interesting to test our models on nui- sance marine molluscs. Although we do not have suf- ficient data for a formal test, we have observed that many harmful marine species have very high fecundi- ties (e.g., the gastropod Littorina littorea, an invader of the U.S. east coast that can produce up to 100,000 eggs/female/year [Jackson 2005]) and the bivalve Perna viridis, an invader of Australia, South America, and Florida that is a broadcast spawner and is therefore likely to have very high fecundity). We stress again, however, that a modeling study similar to ours would be required to quan-

titatively determine the relationship between life history and impacts for marine molluscs.

Reasons for Caution

Although our approach is consistent with many recent recommendations (Mack et al. 2000) and is a considerable advance over qualitative risk-assessment tools, there are nevertheless a number of reasons for caution. First, the number of species available to construct the models was too low to put aside a subset of species for use as an independent data set with which to test the model.

Second, because the limited number of species pre- cluded phylogenetically independent contrasts, our re- sults may have been driven by relationships at taxonomic levels above or below family. The fecundities of some families (e.g., Dreissenidae) indicate that higher-level re- lations may be influencing the results, but the fecundi- ties of Thiaridae support our methods (see Table 1). Our sensitivity analyses, however, suggest that the significant relationships found with logistic regression are phyloge- netically independent at the family level. Related to this caveat, it is possible that fecundity is correlated to an- other life-history variable for which data are not available. We consider this unlikely, however, given that fecundity has been related to impacts across a broad range of taxa, including plants (Richardson & Rejmánek 2004), fishes (Marchetti et al. 2004), and birds (Veltman et al. 1996).

Third, the water-quality characteristics of U.S. fresh- water ecosystems have changed markedly since the first introduced molluscs became established. Changing species × ecosystem interactions make it possible that different natural history characteristics will be required in the future for species to become nuisances than for the species that have already passed through this step. Given the simplicity of our result, however, we are confident that it is robust to considerable environmental change.

Fourth, it is possible that some already established species will become nuisances in the future although they are not causing perceptible harm currently. Our models do not control for time since introduction, so species whose nuisance status has not yet become appar- ent would cause noise in our regression. Nevertheless, our data suggest that for the Great Lakes recent introduc- tions are more likely to become invasive ( p = 0.022), suggesting that our results are robust to concerns about lag times.

Finally, future nuisance molluscs could have impacts of a kind not yet seen and hence might not be identified by our model. If, for example, a disease vector becomes established, then the human-health impact would be dif- ferent from all the previous invaders on which the model was built, and the predictors of that impact might also be different. Bulinus globosus and Indoplanorbus exustus are species from our test group (Table 3) that pose large

Conservation Biology Volume 21, No. 1, February 2007

Keller et al. Fecundity Predicts Mollusc Impacts 199

risks of economic and environmental impacts and are also intermediate hosts for schistosomiasis (Marti et al. 1985; Parashar & Rao 1985). Knowledge of such potential im- pacts should be cause for concern, even for species with low fecundities.

Preventing Future Invasions

Our predictions show that a number of freshwater mol- luscs not yet in the United States are likely to cause eco- nomic and environmental impacts should they become established. Given the enormous cost of nuisance mol- luscs and the fact that eradication is usually impossible, efforts to prevent the establishment of any species con- sidered likely to become a nuisance will be the most cost- effective approach to minimizing damages such as the market (O’Neill 1997; Leung et al. 2002) and nonmarket costs (Ricciardi et al. 1998) of zebra mussel (D. polymor- pha), quagga mussel (D. bugensis), and Asian clam (C. fluminea). Prevention requires that species be assessed before introduction, a difficult task given the variety of potential sources of freshwater molluscs. The problem of risk analysis is further complicated by changing pat- terns of world trade (Drake & Lodge 2004). New trading routes and products will inevitably lead to new suites of species that need to be assessed. Nevertheless, identifi- cation of possible nuisance species before introduction will usually provide a more logistically feasible and eco- nomical option than attempting to keep out all molluscs or eradicating those that become established.

The statistical approach to risk analysis of nuisance species that we have described is more rigorous and trans- parent than the methods currently used by U.S. agencies (e.g., Orr 2003). Risk assessment of species proposed for importation to the United States is rarely conducted, and the methods used are largely qualitative (Orr 2003). Al- though this allows more information to be considered, it is not transparent or repeatable, qualities that policy- determining risk assessments should have (Mack et al. 2000). We believe that statistical approaches should be used in combination with the more species-specific, qual- itative, expert analysis typically used by United States, and state agencies.

When species are found to pose an unacceptable risk according to a statistical model (e.g., Reichard & Hamil- ton 1997; Kolar & Lodge 2002; our model), additional information and expert opinion should be assembled be- fore a policy decision is made. This is the process we followed when predicting likelihood of Great Lakes in- vasion. The same applies to apparently low-risk species, especially with respect to an evaluation of potential im- pacts unlike those considered in model building. Finally, a policy decision must consider potential benefits as well as the potential harms that a species may impose on society (e.g., value to the pet industry). In practice, however, at least in the United States, these sorts of commercial ben-

efits are already given priority, whereas the probability of harms identified by the model presented here are of- ten ignored. Our hope is that quantitative risk assessment models like ours will make possible more balanced evalu- ations of costs and benefits of nonindigenous species and the pathways that deliver them.

Acknowledgments

We are indebted to R. Dillon, R. Howells, W. Kovalek, W. Lellis, R. McMahon, J-P. Pointier, D. Strayer, and the many other malacologists who helped us to find data and references. This work would have been much more diffi- cult without their willingness to share their expertise and experience. J. Murray, M. Xenopolous, J. Bossenbroek, J. Rothlisberger, and two anonymous reviewers provided valuable comments on drafts of this manuscript. This work was funded by grants from Illinois-Indiana Sea Grant (to D.M.L.), Environmental Protection Agency STAR (to D.M.L., graduate fellowship to J.M.D.) and National Sci- ence Foundation Integrated Research Challenges in Envi- ronmental Biology project on Integrated Systems for In- vasive Species (I.S.I.S.) (to D.M.L.).

Literature Cited

Baker, F. C. 1898. The mollusca of the Chicago area: the Pelecypoda. The Natural History Survey, Chicago.

Baker, H. G. 1974. The evolution of weeds. Annual Review of Ecology and Systematics 5:1–24.

Cohen, A., and J. Carlton. 1998. Accelerating invasion rate in a highly invaded estuary. Science 279:555–558.

Colautti, R. I., and H. J. MacIsaac. 2004. A neutral terminology to define ‘invasive’ species. Diversity and Distributions 10:135–141.

Cowie, R. H. 1998. Patterns of introduction of nonindigenous non- marine snails and slugs in the Hawaiian Islands. Biodiversity and Conservation 7:349–368.

Cowie, R. H., and D. G. Robinson. 2003. Pathways of introduction of nonindigenous land and freshwater snails and slugs. Pages 93–122 in G. Ruiz and J. T. Carlton, editors. Invasive species: vectors and management strategies. Island Press, Washington, D.C.

Crooks, J. A., and M. E. Soulé. 1999. Lag times in population explosions of invasive species: causes and implications. Pages 103–125 in O. T. Sandlund, P. J. Schei, and A. Viken, editors. Invasive species and biodiversity management. Kluwer Academic Publishers, Dordrecht, The Netherlands.

De’ath, G., and K. E. Fabricius. 2000. Classification and regression trees: a powerful yet simple technique for ecological data analysis. Ecology 81:3178–3192.

Dillon, R. T., A. R. Wethington, J. M. Rhett, and T. P. Smith. 2002. Pop- ulations of the European freshwater pulmonate Physa acuta are not reproductively isolated from American Physa heterostropha or Physa integra. Invertebrate Biology 12:226–234.

Drake, J. M., and D. M. Lodge. 2004. Global hot spots of biological in- vasions: evaluating options for ballast-water management. Proceed- ings: Biological Sciences Series B (print) 271:575–580.

Fielding, A. H., and J. F. Bell. 1997. A review of methods for the assessment of prediction errors in conservation presence/absence models. Environmental Conservation 24:38–49.

Conservation Biology Volume 21, No. 1, February 2007

200 Fecundity Predicts Mollusc Impacts Keller et al.

Grigorovich, I. A., A. V. Korniushin, and H. J. MacIsaac. 2000. Moitessier’s pea clam Pisidium moitessierianum (Bivalvia, Sphaeriidae): a cryptogenic mollusc in the Great Lakes. Hydrobiologia 435:153– 165.

Grigorovich, I. A., A. V. Korniushin, D. K. Gray, I. C. Duggan, R. I. Co- lautti, and H. J. MacIsaac. 2003. Lake Superior: an invasion coldspot. Hydrobiologia 199:191–210.

Hall, R., J. Tank, and M. Dybdahl. 2003. Exotic snails dominate nitro- gen and carbon cycling in a highly productive stream. Frontiers in Ecology and the Environment 1:407–411.

Horne, F., T. Arsuffi, and R. Neck. 1992. Recent introduction and poten- tial botanical impact of the giant rams-horn snail, Marisa cornuari- etis (Pilidae), in the Comal springs ecosystem of central Texas. The Southwestern Naturalist 37:194–214.

Howells, R. G., and J. W. Smith. 2002. Status of the applesnail Pomacea canaliculata in the United States. In T. Wada, Y. Yusa, R. C. Joshi, and A. I. S. Ghesquiere, editors. The seventh international congress on medical and applied malacology. Available from http://pestalert. applesnail.net (accessed July 2006).

Jackson, A. 2005. Littorina littorea. Common periwinkle. Marine life information network, biology and sensitivity key informa- tion sub-programme. The Marine Biological Association, Plymouth, Devon, United Kingdom. Available from http://www.marlin.ac. uk/species/Littorinalittorea.htm (accessed February 2006).

Kolar, C. S., and D. M. Lodge. 2001. Progress in invasion biology: pre- dicting invaders. Trends in Ecology & Evolution 16:199–204.

Kolar, C. S., and D. M. Lodge. 2002. Ecological predictions and risk assessment for alien fishes in North America. Science 298:1233– 1236.

Lach, L., and R. H. Cowie. 1999. The spread of the introduced freshwa- ter apple snail Pomacea canaliculata (Lamarck)(Gastropoda: Am- pullariidae) on O’ahu, Hawai’i. Bishop Museum Occasional Papers 58:66–71.

Leung, B., D. M. Lodge, D. Finnoff, J. F. Shogren, M. A. Lewis, and G. Lamberti. 2002. An ounce of prevention or a pound of cure: bioe- conomic risk analysis of invasive species. Proceedings: Biological Sciences 269:2407–2413.

Levine, J. M., and C. M. D’Antonio. 2003. Forecasting biological invasions with increasing international trade. Conservation Biology 17:322– 326.

Mack, R. N., D. Simberloff, W. M. Lonsdale, H. Evans, M. Clout, and F. A. Bazzaz. 2000. Biotic invasions: causes, epidemiology, global consequences, and control. Ecological Applications 10:689–710.

Mackie, G. L. 2000. Ballast water introductions of mollusca. Pages 219– 254 in R. Claudi and J. H. Leach, editors. Nonindigenous freshwater organisms: vectors, biology and impacts. Lewis Publishers, Boca Ra- ton, Florida.

Marchetti, M. P., P. B. Moyle, and R. Levine. 2004. Invasive species profil- ing? Exploring the characteristics of non-native fishes across invasion stages in California. Freshwater Biology 49:646–661.

Marti, H. P., M. Tanner, A. A. Degremont, and T. A. Freyvogel. 1985. Studies on the ecology of Bulinus globosus, the intermediate host of Schistosoma haematobium in the Ifakara area, Tanzania. Acta Tropica 42:171–187.

McMahon, R. F. 2002. Evolutionary and physiological adaptations of aquatic invasive animals: r selection versus resistance. Canadian Jour- nal of Fisheries and Aquatic Sciences 59:1235–1244.

Mills, E. L., J. H. Leach, J. T. Carlton, and C. L. Secor. 1993. Exotic species in the Great Lakes: a history of biotic crises and anthropogenic in- troductions. Journal of Great Lakes Research 19:1–54.

OEHHA (Office of Environmental Health Hazard Assessment). 2006. Proposition 65: process for developing safe harbor numbers. OE- HHA, Sacramento, California. Available from www.oehha.ca.gov/ (accessed August 2006).

O’Neill, C. R. 1997. Economic impact of zebra mussels—results from the

1995 national zebra mussel information clearinghouse study. Great Lakes Research Review 3(1):35–42.

Orr, R. 2003. Generic nonindigenous aquatic organisms risk analysis review process. Pages 415–438 in G. M. Ruiz and J. T. Carlton, editors. Invasive species: vectors and management strategies. Island Press, Washington, D.C.

Parashar, B. D., and K. M. Rao. 1985. Effect of temperature on growth, reproduction and survival of freshwater snail, Indoplanorbis exus- tus (Gastropoda: Pulmonata), vector of schistosomiasis. Archiv fur Hydrobiologie 102:379–386.

Pointier, J., J. Torraft, and M. Lefevre. 1991. Life tables of freshwater snails of the genus Biomphalaria (B. glabrata, B. alecandrina, B. straminea) and of one of its competitors Melanoides tuberculata under laboratory conditions. Malacologia 33:43–54.

Quinn, G. P., and M. J. Keough 2002. Experimental design and data anal- ysis for biologists. Cambridge University Press, Cambridge, United Kingdom.

Rader, R., M. Belk, and M. Keleher. 2003. The introduction of an in- vasive snail (Melanoides tuberculata) to spring ecosystems of the Bonneville Basin, Utah. Journal of Freshwater Ecology 18:647–657.

Reichard, S. H., and C. W. Hamilton. 1997. Predicting invasions of woody plants introduced into North America. Conservation Biology 11:193–203.

Rejmánek, M., D. M. Richardson, S. I. Higgins, M. J. Pitcairn, and E. Grotkopp. 2005. Ecology of invasive plants: state of the art. Pages 104–161 in H. A. Mooney, R. N. Mack, J. A. McNeely, L. E. Neville, P. J. Schei, and J. K. Waage, editors. Invasive alien species: a new synthesis. Island Press, Washington, D.C.

Ricciardi, A. 2001. Facilitative interaction among aquatic invaders: is an “invasional meltdown” occurring in the Great Lakes? Canadian Journal of Fisheries and Aquatic Sciences 58:2513–2525.

Ricciardi, A., R. J. Neves, and J. B. Rasmussen. 1998. Impending extinc- tions of North American freshwater mussels (Unionoida) following the zebra mussel (Dreissena polymorpha) invasion. Journal of Ani- mal Ecology 67:613–619.

Richardson, D. M., P. Pyšek, M. Rejmánek, M. G. Barbour, F. D. Panetta, and C. J. West. 2000. Naturalization and invasion of alien plants: concepts and definitions. Diversity and Distributions 6:93–107.

Richardson, D. M., and M. Rejmánek. 2004. Conifers as invasive aliens: a global survey and predictive framework. Diversity and Distributions 10:321–331.

Simberloff, D., and B. Von Holle. 1999. Positive interactions of non- indigenous species: invasional meltdown? Biological Invasions 1:21– 32.

Solow, A. R., and C. J. Costello. 2004. Estimating the rate of species introductions from the discovery record. Ecology 85:1822–1825.

Therneau, T. M., and B. Atkinson. 2005. Rpart: recursive partitioning. R package. Version 3.1–22. Mayo Foundation for Medical Education and Research, Rochester, Minnesota. Available from http://cran.r- project.org/src/contrib/Descriptions/rpart.html (accessed Novem- ber 2005).

Turgeon, D. D., et al. 1998. Common and scientific names of aquatic invertebrates from the United States and Canada: molluscs. American Fisheries Society, Bethesda, Maryland.

Veltman, C. J., S. Nee, and M. J. Crawley. 1996. Correlates of introduc- tion success in exotic New Zealand birds. The American Naturalist 147:542–557.

WHO (World Health Organization). 2002. Schistosomiasis: strategic di- rection for research. WHO, Geneva.

Williamson, M. H. 1996. Biological invasions. Chapman & Hall, London. Williamson, M. H. 1999. Invasions. Ecography 22:5–12. Zaranko, D., D. Farara, and F. Thompson. 1997. Another exotic mollusc

in the Laurentian Great Lakes: the New Zealand native Potamopyr- gus antipodarum (Gray 1843) (Gastropoda, Hydrobiidae). Canadian Journal of Fisheries and Aquatic Sciences 54:809–814.

Conservation Biology Volume 21, No. 1, February 2007

__MACOSX/._fecundity+as+basis+for+risk+assessment+in+mulloscs.pdf

phil+scrutiny+of+evidence+of+risk.pdf

Philosophical Scrutiny of Evidence of Risks: From Bioethics to Bioevidence

Deborah G. Mayo and Aris Spanost

We argue that a responsible analysis of today's evidence-based risk assessments and risk debates in biology demands a critical or metascientific scrutiny of the uncertainties, assumptions, and threats of error along the manifold steps in risk analysis. Without an accompanying methodological critique, neither sensitivity to social and ethical val- ues, nor conceptual clarification alone, suffices. In this view, restricting the invitation for philosophical involvement to those wearing a "bioethicist" label precludes the vitally important role philosophers of science may be able to play as bioevidentialists. The goal of this paper is to give a brief and partial sketch of how a metascientific scrutiny of risk evidence might work.

1. Introduction. Risk assessment controversies in biology and other sci- ences often revolve around disagreements regarding the nature, interpre- tation, and justification of methods and models used to learn from in- complete and uncertain data. While philosophers of science are ostensibly interested in helping to clarify, if not also to resolve, matters of evidence and inference, they are rarely consulted in practice for this end. Where philosophers are called on to play a role in risk debates, for example, on science panels, their input has largely been focused on the role of ethical and other value judgments in risk policy disputes. As welcome as such participation has been, our position is that issues about values in evidence- based policy call for corresponding attention to methodological issues that enter in collecting, interpreting, communicating, and evaluating the evidence. In Mayo and Hollander (1991), these were dubbed issues of "acceptable evidence" in deliberate contrast to policy questions about "acceptable risk." That risk assessment judgments intertwine with ethical and value judgments demands a greater methodological understanding that allows for a critical or metascientific scrutiny of the uncertainties,

tTo contact the authors, write to: Deborah G. Mayo, Department of Philosophy, Virginia Tech, Blacksburg, VA 24061; e-mail: [email protected]; Aris Spanos, Department of Economics, Virginia Tech, Blacksburg, VA 24061; e-mail: [email protected]. Philosophy of Science, 73 (December 2006) pp. 803-816. 0031-8248/2006/7305.0030$10.00 Copyright 2006 hy the Philosophy of Science Association. All rights teserved.

803

804 DEBORAH G. MAYO AND ARIS SPANOS

assumptions, and threats of error along the manifold steps in risk analysis. Disagreements that might be attributed to diverging policy and ethical values may actually be the result of divergent assumptions guiding the construction and use of models, and disagreements in the foundations of uncertain knowledge and statistical inference. Although these issues are generally intermingled in debates, the philosopher of science's penchant for laying bare presuppositions of claims and arguments would afford real progress in understanding. If this is correct, then restricting the in- vitation for philosophical involvement to those wearing a "bioethicist" label overlooks the most constructive ways in which philosophers of sci- ence can and should contribute to these debates. If a label is needed, perhaps bioevidentialist would do.

An important advance represented by our co-symposiasts is the rec- ognition that philosophers of science can serve an important role in de- veloping and conducting a metascientific analysis of aspects of risk de- bates. Even where there is no dispute as to whether a given harm is of concern, they rightly observe, what is often disputed is whether there is evidence of the existence of that harm or hazard, and this in turn may revolve around such choices as which end points to measure and how risks are "framed." For example, as Thompson (2006, in this issue) notes, there is a higher estimate of risks of genetically modified (GM) crops if the focus is on the initial stages where uncertainties are high rather than on a later stage after which problematic cases are likely to have been weeded out. Or, again, associating hormesis with mechanisms of natural selection, Elliott (2006, in this issue) observes, renders it of greater sci- entific respectability than when associated with homeopathic ideas. Such conceptual analysis can help shed light on the nature of risk debates, but more is required to criticize and help to adjudicate competing risk as- sessments. The question is: Why stop with conceptual analysis? Why not critique the reliability of the evidence and inferences? The goal of this paper is to give a brief and partial sketch of the kind of metastatistical scrutiny we have in mind. We will apply these ideas to two examples raised by co- symposiasts Thompson and Elliot, GM crops and hormetic effects.

2. Metastatistical Critique of Risk Inference Options. To evaluate how much of a controversy in risk assessment is due to uncertainties in data and how much to conflicting values (social, ethical, economic, religious) requires being able to critically evaluate what the evidence is; and as risk evidence invokes probabilistic and statistical methods, an adequate meta- scientific scrutiny requires coming to grips with these methods. This does not mean that philosophers of science become statisticians, toxicologists, or the like: that would be both too much and too little. Too much because it would be impractical to become experts in all the arenas involved; too

EVIDENCE OF RISKS 805

little because there is a great deal of confusion and foundational unclarity among such 'experts'. A sufficient understanding of the inference methods together with a platform for raising questions about fallacies and pitfalls, we argue, could go a long way toward developing a metascientific (and metastatistical) scrutiny with real bite.

In particular, philosophers of science can serve an important role in developing and conducting an analysis of the various judgments and de- cisions required to determine if data constitute acceptable evidence of a given risk;' we might call these risk inference option^ (e.g., choice of statistical significance levels, dose-response models). Because there is lat- itude for choice among possible inference options, and each choice infiu- ences the chance of obtaining evidence for a given risk (or benefit), much risk controversy revolves around these inference options. Notwithstanding the latitude in choosing inference options, we argue, it is possible to determine how different choices infiuence a method's ability to detect risks, that is, its risk (or benefit) detecting capacity. This would be the basis for systematically addressing the following questions:

1. How do various methodological choices made in the generating, modeling, and interpreting of data alter a test's risk detection ca- pacity? (For example. Do data-dependent searches alter risk detec- tion ability? If so, how should 'selection effects' be taken into account?)

2. What uncertainties and errors have been well ruled out? Which have been overlooked and why? (e.g., extrapolations beyond the lab). Are given policy standards met or fiouted?

3. What are the statistical and the substantive assumptions in collecting and modeling the data? How well are they satisfied,' and what are the consequences for the reliability of inference of their being vio- lated in the analysis at hand?

2.1. Beyond Dirty Hands. Failure to have a critical understanding of the (meta-) statistical issues often leads to the position that standards for estimating risks from statistical data are so bound up with subsequent policy decisions that scientists invariably (if unconsciously) introduce pol- icy bias into the interpretation of risk evidence. "While scientists use the 95% rule or confidence limits to the 95% value, they remain loyal to the

1. Risk is usually distinguished from hazard assessment in including estimates of ex- posure, but our discussion will not turn on this. 2. These may also be called risk assessment policy options, as in the NAS-NRC report 1983 (Mayo 1991).

3. For a discussion of testing the model assumptions, see Mayo and Spanos (2004).

806 DEBORAH G. MAYO AND ARIS SPANOS

conventions of their discipline . . . but they implicitly 'dirty' their hands, . . . because they risk begging important regulatory issues" (Cranor 1993, 42).

The same point is most often put in terms of type I and type II errors in testing. In the present context these two errors may be informally summarized as follows:

type I error: the data are taken as evidence of a risk (or benefit), when in fact the risk (or benefit) is absent (false positive), type II error: the data are not taken as evidence of a risk (or benefit), when in fact the risk (or benefit) is present (false negative).

The 95% rule refers to the requirement that a test have a low probability, for example, .05, of inferring a genuine effect (e.g., a genuine risk) when it would be an error to do so (commit a type I error).

The allegation is that choosing the trade-off between type I and II error rates is invariably to "dirty one's hands" with policy. This charge, however, stems from a caricature of statistical tests where a statistical report, based on arbitrary cut-offs, is taken to automatically warrant a policy decision. This would be an abuse of statistical tests. The dirty-hands allegation only underscores the need for a critical assessment of statistical tools; for it is a well-known fallacy to identify statistical significance with substantive importance (albeit still committed), how much worse to go straight from statistical significance to a policy decision. Although what counts as a risk of concern is a policy question, whether a statistically significant/ nonsignificant result warrants the presence/absence of a given risk (in- crease or decrease) is not. The same holds for the various other risk inference options needed to interpret risk data. While, in any particular case, options may be based on 'unthinking conventions' (e.g., the .05 cut- off for statistical significance), on philosophical principles of evidence, or deliberately chosen to further policy preferences, it does not follow that any criticisms of resulting inferences are themselves matters of policy and/ or value judgments. For example, one researcher may prefer a less 'pro- tective'" extrapolation model for cancer risk on grounds of policy, but such models may be evaluated on grounds of statistical adequacy or pre- dictive reliability, not on policy grounds.

Adding another level of complexity to our bioevidentialist task is the fact that the thorniest risk debates are often intermingled with founda- tional disagreements regarding methodologies of uncertain inference. Choices about which evidential methods to use may revolve around dif- ferent philosophies of statistics, quite apart from deliberate policy choices.

4. A risk assessment option that makes it more likely to regard data as evidence of a risk. For a discussion of the protectiveness of RAP options, see Mayo (1991).

EVIDENCE OF RISKS 807

For example, the same data that would lead a significance tester to infer evidence of risk may lead a Bayesian statistician to assign a fairly high posterior probability to the no-risk hypothesis (Mayo 2005); or again, a Bayesian (or follower of the likelihood principle) would not regard evi- dence as altered because of 'optional stopping', whereas a Neyman-Pear- son frequentist would (Mayo and Kruse 2001). Without taking sides, the bioevidentialist can compare the standards of protectiveness of the infer- ences licensed by different schools of inference in particular cases.

2.2. Acceptable Evidence versus Acceptable Risk Policy Decisions. In some discussions, the language of type I and II errors is taken out of the formal statistical context and exported into the arena of risk policy man- agement; and unless one is very careful, confusion ensues. Identifying the type I error with "regulating a safe technology as if it has risks" and the type II error as "implementing an unacceptably risky technology," these discussions give ethical arguments for minimizing the type II rather than the type I error probability. It is important to distinguish such discussions of 'acceptable risks' from the current discussion of 'acceptable evidence':

{a) Acceptable evidence. Given the information and data, what infer- ences about the extent of risks (or benefits) are evidentially warranted?

{b) Acceptable risk management. Given the evidence of risk, what pol- icies (or trade-offs) are acceptable?

Although questions under {a) and {b) are not always neatly distinguished, using the same terms to refer to an error in inference as an error in regulation leads to thinking that because the latter turn on ethical and policy judgments, so do the former. A question under (b) might be: should we fail to cut emissions despite the evidence of increased risks, to protect markets (Shrader-Frechette 1991, 2006, in this issue)? To address this, one may invoke general ethical principles that favor protecting the public versus protecting industry in cases with uncertainty where there is some evidence of public hazards. An opposing argument may weigh against such a precautionary stance in the face of likely economic consequences, and the debate may remain until further evidence. A question under {a) might be: do the data constitute evidence of greater increased risks than industry risk assessors allege? To suppose that disagreements about {a) also rest on ethical-policy differences is to forfeit the essential basis for charging that an assessment misinterprets the evidence.

A second consequence of using the same terms to refer to an error in inference as an error in regulation is that what is intended as advocating a protective policy stance is likely to be misunderstood as advocating the

808 DEBORAH G. MAYO AND ARIS SPANOS

minimization of the type II error in statistical testing. Taken seriously, this would allow erroneously construing data as evidence of risk with so high a probability that tests would easily become meaningless tools.

Moreover, even if one's concern is (b) (moving from risk estimates to policies), one should begin by scrutinizing the risk evidence, lest one's policy goal be inadvertently thwarted. Ironically, this often happens. For example, it may be argued that a 'precautionary' policy is warranted in cases where the evidence is ambiguous or inconclusive—a concern under (b)—but that depends on a distinct assessment of this inconclusiveness. As we will later note, current rules of thumb may allow the data to lead to 'inconclusive' reports when the data actually contain evidence of risk increases.

3. How to Find Out the Truth with Metastatistics. In a typical test of risk, we set up a null hypothesis, H^, that asserts that there is no increased risk, and an alternative hypothesis, /f,, that says there is:

HQ. a risk difference 8 is 0, //,: a risk difference 8 is nonzero.

Or // , might be that 8 is greater than 0 (one-sided test). A sample of size n is represented by a set of random variables X = {X^, . . ., XJ, and the data XQ = {x^, . . ., xj is a realization of X The experiment might involve studying and reporting the observed difference in risk among those ex- posed (or 'treated') and those unexposed (or 'untreated') denoted by d{xo).^ One uses the observed d{Xa) to learn about the underlying risk increases that gave rise to the data, as given by the parameter 8 in the statistical hypotheses.

A familiar test rule is: Reject //„ and infer that data x^ provide evidence of a risk increase if and only if d{x^ is statistically significant at the a level.^

When we speak of a test 'detecting a risk', we mean 'it reports a sta- tistically significant result' (at the chosen significance level a); with an 'insignificant result', by contrast, the null hypothesis is not rejected. Al- though this is often abbreviated as '//o is accepted', it is intended to be understood as data XQ do not provide evidence against ^o- (Subtleties between Fisherian significance tests and Neyman-Pearson tests will not alter our points, but see Mayo and Spanos 2006 and Mayo and Cox 2006).

5. These hypotheses must actually be stated in terms of parameters of a formal sta- tistical model.

6. Observed difference d{x^ is statistically significant at level a if P{d{X) > d{xo); H,) < a.

EVIDENCE OF RISKS 809

3.1. Problems with Statistically Insignificant Results. A very common worry in risk analysis is that a test fails to detect a risk not because one is absent but, rather, because the test had little chance of detecting risks even if they exist. The concern is with the type II error. Rather than construe statistical risk reports as leading to "dirtying one's hands" with policy values, our metastatistician scrutinizes such claims and avoids taking no evidence of risk as evidence of no riskl Failure to reject H^ does not license inferring that a risk increase is less than 8, if the test has very little chance of detecting an increased risk of 8, even were it present. The test, we would say, was not a very stringent or severe scru- tiny of possible risks. Since the alternative H, asserts that there is some (positive) risk 8, it is a composite hypothesis containing all the different values that the risk 8 might take. Accordingly, the probability of a type II error will vary for each value of 8, and it is not correct to speak of the type II error without specifying the alternative or discrepancy 8 for which it is being calculated.

Some apparently fear that it would be too difficult for policy makers to understand how to scrutinize insignificant results. In fact, as we will see, the reasoning required is no more complicated than the reasoning that forms the basis of the criticism of a too insensitive test.

3.2. Problems with Statistically Significant Results. In other studies a null hypothesis of zero (0) improvement is tested:

HQ: an improvement (or benefit) 8 difference is 0, //,: there is a nonzero improvement.

Data ATo may be considered to provide evidence for inferring that a treatment produces benefits when a null hypothesis of 'no benefit' is rejected at a low significance level. Here a metastatistical scrutiny would center on whether H^ was rejected too readily; high type I error prob- ability. Given that risks of interest are often of low probability, in one sense this is less often a problem than insensitive tests. However, there are aspects of the data and hypothesis generation procedure that can introduce high type I error probabilities into tests purporting to have controlled this error. One way this can occur is if the procedure searches for benefits and reports just those that are found. In a classic example of "hunting for significance," suppose one searches through 20 differences and reports the one that reaches a significance level of .05. The prob- ability of finding at least one, .05 level, statistically significant difference out of 20, even if the null hypotheses are all true, is approximately .64 (i.e., (1 - .95^")). So the type I error probability would be .64, not .05. Note that here it is the inference to the non-null alternative //, that lacks sufficient stringency or severity.

810 DEBORAH G. MAYO AND ARIS SPANOS

3.3. Two Metastatistical Tools Based on the Severity Criterion. We can systematize the above reasoning by supplementing standard statistics with metastatistical tools for interpreting (i) insignificant and (ii) significant test results. To allude to the two examples we consider in this paper, (i) is a negative result purporting to have evidence of absence of risk, for example, GM crops do not pose threats to untargeted species; while (ii) is a positive result that purports to have evidence of improvements, for example, low doses of toxins provide beneficial hormetic effects. That is,

(i) Negative result: A statistically insignificant departure from the null hypothesis of no risk is taken as evidence for H: risks do not exceed 8. (Here H corresponds to failing to reject the null hypothesis.)

A metastatistical rule must say when this is unwarranted: If there is a high probability that a test yields a statistically insignificant result, even though risk 8 is present, then x fails to provide acceptable evidence that risks 8 are absent.'

(ii) Positive result: A statistically significant departure from the null hypothesis of no risk (or benefit) is taken as evidence for H: risk (or benefits) exist. (Here H corresponds to rejecting the null hypothesis.)

Again a metastatistical rule must, at the very minimum, say when this is unwarranted: If there is a high probability that a test yields a statistically significant result, even though improvements 8 are absent, then data x^ fail to provide acceptable evidence that benefits 8 are present.

Severity Criterion (SC). Regardless of whether we have statistically significant or insignificant results, and despite the fact that there are two types of statistical errors, we are able to identify a single principle for scrutinizing the acceptability of the evidence for any given claim H. To have a unified way of speaking let us adopt testing language wherein hypothesis H 'passes a test' with Xg covers various ways in which JCQ 'fits', 'accords with', or otherwise purports to provide evidence for H. The severity criterion states:

(SC) If there is a high probability a test passes hypothesis H even though H is false, then a passing result x^ fails to provide acceptable evidence for H.

SC captures our intuition that data JCQ fail to provide good evidence for the truth of H if the test had little chance of providing evidence against H, even when H is false. Such a test, we would say, is insufficiently strin-

7. Initial discussions of this kind of metastatistical rule are in Mayo (1985, 1988, 1996, 2004, 2005).

EVIDENCE OF RISKS 811

gent or lacks severity.^ The onus is on the person claiming to have evidence for H to show that the claim is not guilty at least of egregious lack of severity, and metastatistical scrutiny can provide systematic ways to de- termine if they have succeeded.

Anticipated Objection. Given the controversies between frequentist and nonfrequentist (e.g., Bayesian) statistical approaches, do we bias things by assuming a frequentist error statistical paradigm? No, we limit our- selves here to egregious construals of evidence: any approach, frequentist or Bayesian, that is not able to mount the above criticism of inferences with high error probabilities should be seriously called into question.' The severity criterion applies also to the use of other statistical methods aside from significance tests, whether a confidence interval, Bayesian inference, or other. The standard statistical tests do not directly supply severity assessments; severity is a 'metastatistical' concept. However, error prob- abilities can be used to supplement methods of inference with a severity assessment that is sensitive to the actual outcome d{x^ from whatever procedure has been used to infer the claim in question.

4. How to Tell the Truth (about Insignificant Results) with Metastatistics: GM Crops. Consider the case of GM crops, in particular, plants genet- ically modified to have pesticidal traits (now called 'plant incorporated pesticides'), such as genes to cause crops to produce Bt {Bacillus thurin- giensis) toxin. A concern is the possible danger to nontarget species such as earthworms or monarch butterfiies. Successful EPA petitions to de- regulate a Bt crop are based on evidence of acceptable risks to nontargeted species. This evidence in turn is based on finding that the results of lab exposures are not statistically significant in tests of null hypotheses such as:

HQ. Bt crops do not adversely effect untargeted species.

The concern is that failure to reject the null may be due, not to absence of effects, but to the experiment and statistical tests not being stringent or powerful enough to detect them. The metastatistical rule for inter- preting insignificant results comes into play. For example, as discussed in

8. This notion is developed in much more detail elsewhere (Mayo 1996; Mayo and Spanos 2006). The severity function SEV(.) has three arguments: a test T, an outcome or result x, and an inference or a claim H. SEV(Test T, outcome x, claim //) , is to be read "the severity with which claim //passes test Twith outcome jc." If there is a high probability that the test would purport to have evidence for H even though H is false, then SEV (T, x, H) = low.

9. In Bayesian testing, small significance levels with large samples can lead to null hypotheses of 'no risk' receiving high posterior probabilities. In those cases, use of Bayesian posteriors to judge acceptable evidence is problematic (see Mayo 2005).

812 DEBORAH G. MAYO AND ARIS SPANOS

Marvier (2002), an experiment on four replicate batches of earthworms, 10 to a batch, were exposed to soil that included leaves from either trans- genic Bt cotton or nontransgenic cotton, and after 2 weeks (too short a time to expect differences in survival rates) the exposed worms gained 29.5% less weight on average than the others. Because this difference is not statistically significant the study concluded that this particular Bt toxin did not impair weight gain in earthworms.

However, due to the low sample size (four replicates), and the large variability among replicates, the test has low capacity to detect adverse weight effects. Given that the number of replicates the EPA requires fails to take into account within sample variation, Marvier reports, very few of the experiments that resulted in statistically insignificant results had a high (90%) probability of detecting even a 50% change (either in survival or weight decrease). Nearly all had little power to detect risks of concern; abbreviate it as 8*.

Data JCQ 'fit' the no-risk hypothesis, but the probability is high that no increased risk is detected, even if risks as high as 8* were present.

Although HQ 'passed' test T, the test it passed was not severe—it is highly probable that H^ would pass this test, even if the increased risk is actually as large as 8*. From (SC), a failure to reject //„ with test Tdoes not license inferring that the increased risk is less than 8*. What counts as a "risk of concern" reflects policy values, but this critique does not.

The severity criterion also directs us to find specific values of 8 that are large enough to be ruled out by dint of the insignificant result. The EPA test had a fairly high probability of detecting a weight change of 56.37% or more; thus the insignificant result may warrant ruling out a 56% de- creased weight (between Bt-treated and control worm groups). Generally the reports in the literature provide what is needed for a metastatistical analysis; if not, that alone is grounds for questioning.'"

There are lessons both for planning and interpretation. Pre-data: one should specify the effect size of interest (decreased survival, weight, off- spring) and calculate the sample size for reasonable power to detect it. Post-data: one should scrutinize the severity attained—based on the actual outcome, variability, and so on. For good discussions, see Marvier (2002) and Burgman (2005, Chapter 11), as well as earlier references.

Rules of thumb using confidence intervals are not immune. Appealing to

10. Reports are typically in terms of the power of a test. Although a high power to detect 5 is not necessary for a high severity that H: risk increase < 5, it is sufficient. Thus, by selecting a test with high power for detecting 5 one is assured of this much protection: a nonstatistically significant result warrants with severity a risk increase < 6.

EVIDENCE OF RISKS 813

confidence interval (CI) estimates of risk is often thought to avoid mis- interpreting insignificant results, but more care is needed. For example, a common rule of thumb is that if both the 0 effect and the risk of concern 8* are included in the interval estimate formed from data XQ, then the results are 'inconclusive' (see Burgman 2005, 341). However, the data may provide reasonably severe evidence of the presence of risk 8* (using our criterion); and thus a report of 'inconclusive' may not be warranted.

5. A Bioevidentialist Critique of Significant Effects: Hormesis. Hormesis refers to a phenomenon in which a substance that is deleterious at high doses causes a response in the opposite direction at low doses (we can call such low dose reversals 'improvements' to steer clear from calling them 'benefits'). Attention to "framing effects" in risk controversies, as usefully delineated by Elliott (2006, in this issue) for this example, reveals that the way risks are characterized can have a psychological impact in risk controversies. But applying our "bioevidential" scrutiny lets us go much further in waging an effective yet nontechnical critique.

Calabrese (2005), a leading proponent of the hormetic hypothesis, has argued that hormesis is a widespread adaptive, stimulatory response. For example, while high doses of dioxin cause increases in tumors, Calabrese cites data showing a suppression of tumors at low dose exposure to dioxin. Here we have a case where rejecting one or more null hypotheses:

Hg. no benefit (or even harms) at low doses,

is the basis for inferring evidence of improvements or decreased risk at low doses. So right away the metastatistical question directed at a positive or statistically significant result (3.3 (ii)) comes to mind: Have they properly controlled type I error probabilities (false positives)? We know from the metastatistical rule for interpreting statistically significant results that if data XQ are to provide acceptable evidence for the presence of an effect then high severity demands that it not be highly probable to have reported such evidence erroneously. The onus is on the proponents of hormesis to supply convincing evidence that they are not open to misconstruing ran- dom effects as genuine.

Before discussing this case we want to emphasize that we are not pur- porting to decide one way or another about the controversial theory of hormesis—for starters, this short discussion could not do justice to so complex an issue (Mayo and Spanos 2007). Our goal is to illustrate how a metastatistical critique can provide standard ways for nonspecialists to raise questions even in dealing with complex evidence-based risks before arguing about what policies might be warranted assuming some risk evi- dence. Considering this case also helps to illustrate the point raised in Section 3.2: it does not suffice that a low type I error is reported—the

814 DEBORAH G. MAYO AND ARIS SPANOS

actual type I error probability may be a lot higher or it may be uncon- trolled altogether, due to certain features of data-dependent selections. Finally, this case would seem to be of interest to philosophers of science both because of the relevance to evidence-based policy and the fact that it is regarded as a possibly revolutionary change in the standard models used in toxicology (Calabrese 2005).

Evidential warrant for a paradigm change? Although some hormetic effects are apparently uncontroversial, existing use of the linear threshold model in toxicology already allows taking these into account (via U or J shaped models) on a case by case basis. Calabrese and Baldwin (2003) want to go much further: they claim to have provided sufficient evidence to actually change the default assumption in toxicology: "These findings challenge the long-standing belief in the primacy of the threshold model in toxicology (and other areas of biology involving dose-response rela- tionships) and provide strong support for the hormetic-like biphasic dose- response model characterized by a low-dose stimulation and a high-dose inhibition" (ibid., 246; emphasis added). As Crump (2001) points out, however, this would demand evidence of a near universal prevalence of hormesis. So the evidential hurdle for the bioevidentialist to consider is whether there is evidence of a sufficiently general hormetic effect. Given the difficulty of detecting the low dose effects of interest, Calabrese, Bald- win, and Holland (1999) decide to obtain their evidence of hormesis through a literature search of n = 10,000 studies. By putting together those that show apparently hormetic-looking risk assessments they make a case for this "strong support." But is there acceptable evidence for this?

Among various methodological questions to which these studies give rise, we limit ourselves to a question about the effect of 'hunting for statistical significance' (Mayo 1996; Mayo and Kruse 2001; Mayo and Cox 2006). Already aware of how type I error rates increase with hunting procedures, our bioevidentialist quickly grasps the gist of Crump's con- cern: "In order to properly control for the false-positive rate one would need to know how extensive the search was that located the data set. If the data set was the most hormetic looking out of 100 examined, then to conduct a statistical test for hormesis at the standard 0.05 level one should use p = 0.0005 (the solution to 1 - (1 -/?)""' = 0.05) rather than p = 0.05" (ibid., 672). In other words, the researchers would have needed a vastly smaller significance level for each case examined in order for the overall type I error probability to be small. Notice that the task for the bioevidentialist is not to figure out precise significance levels or other error probabilities, it is to point out the kinds of fallacies that must be put to rest. It might next be noted that the data on which they base their infer- ences are not themselves a random selection from all studies but, rather, are based on a point system they devise, which itself merits scrutiny. On

EVIDENCE OF RISKS 815

this point system, data are taken as evidence of hormesis simply because a study could have shown evidence of hormesis, whether or not it actually did. ("A data set could achieve a score as high as 6 (high end of the low evidence region for hormesis) even if there was no evidence for hormesis" [Crump 2001, 675].)

An effective strategy to demonstrate lack of control of the type I error probability is to apply the test to data deliberately generated to have the null hypothesis true (no hormesis). Such a simulation allows determining the expected distribution of scores from studies in which a hormetic effect is not present (i.e., false-positive rate.) Our bioevidentialist could make use of Crump's report: "Results of this simulation . . . demonstrates the scoring system does not control the false positive rate (indication that a hormetic effect is falsely identified when none is present). . . . Using the same scoring system, between 94.9% and 99.7% of the simulated data sets showed some evidence of hormesis (score > 2), even though no hormetic effect was present" (Crump 2001, 675).

Unless the results of this simulation are themselves faulty, it appears that Calabrese et al. (1999) have not put the hormesis hypothesis to a stringent or severe test: their data collection and analysis makes it far too easy to produce apparently supporting evidence even where we know the hormetic hypothesis is false. Although philosophers of science would not be expected to run such simulations, using critical information that exists or even asking whether such a challenge could be answered are important first steps.

6. Concluding Comments. We have argued that neither sensitivity to social and ethical values, nor conceptual clarification alone, suffices for the re- sponsible analysis and understanding of today's evidence-based risk as- sessments and risk debates. Although issues of acceptable evidence are generally intermingled with those of acceptable risk management, the philosopher of science's penchant for laying bare presuppositions of claims and arguments would afford real progress in understanding. If this is correct, then restricting the invitation for philosophical involvement to those wearing a "bioethicist" label precludes the vitally important role philosophers of science may be able to play as bioevidentialists. We hope to encourage a move in that direction.

REFERENCES

Burgman, M. (2005), Risks and Decisions for Conservation and Environmental Management. Cambridge: Cambridge University Press.

Calabrese, E. J. (2005), "Hormetic Dose-Response Relationships in Immunology: Occur- rence, Qualitative Features of the Dose-Response, Mechanistic Foundations, and Clin- ical Implications," Critical Reviews in Toxicology 35: 89-295.

816 DEBORAH G. MAYO AND ARIS SPANOS

Calabrese, E. J., and L. A. Baldwin (2003), "The Hormetic Dose-Response Model Is More Common than the Threshold Model in Toxicology," Toxicological Sciences 71: 246- 250.

Calabrese, E. J., L. A. Baldwin, and C. D. Holland (1999), "Hormesis: A Highly Gener- alizable and Reproducible Phenomenon with Important Implications for Risk Assess- ment," Risk Analysis 19: 261-281.

Cranor, C. F. (1993), Regulating Toxic Substances: A Philosophy of Science and the Law. Oxford: Oxford University Press.

Crump, K. (2001), "Evaluating the Evidence for Hormesis: A Statistical Perspective," Crit- ical Reviews in Toxicology 31: 669-679.

Elliott, K. C. (2006). "A Novel Account of Scientific Anomaly: Help for the Dispute Over Low-Dose Biochemical," Philosophy of Science 73 (5), in this issue.

Marvier, M. (2002), "Improving Risk Assessment for Nontarget Safety of Transgenic Crops," Ecological Applications 12: 1119-1124.

Mayo, D. G. (1985), "Increasing Public Participation in Controversies Involving Hazards: The Values of Metastatistical Rules," Science, Technology, and Human Values 10: 55- 68.

(1988), "Toward a More Objective Understanding of the Evidence of Carcinogenic Risk," in Arthur Fine and Jarrett Leplin (eds.), PSA 1988: Proceedings of the 1988 Biennial Meeting of the Philosophy of Science Association, vol. 2. East Lansing, MI: Philosophy of Science Association, 489-503.

(1991), "Sociological vs. Metascientific Views of Risk Assessment," in Mayo and Hollander 1991, 249-279.

(1996), Error and the Growth of Experimental Knowledge. Chicago: University of Chicago Press.

(2004), "An Error-Statistical Philosophy of Evidence," in M. Taper and S. Lele (eds.). The Nature of Scientific Evidence: Statistical, Philosophical, and Empirical Con- sideration. Chicago: University of Chicago Press.

- (2005), "Evidence as Passing Severe Tests: Highly Probed vs. Highly Proved," in P. Achinstein (ed.). Scientific Evidence. Baltimore: Johns Hopkins University Press.

Mayo, D. G., and D. R. Cox (2006), "Frequentist Statistics as a Theory of Inductive Inference," in Optimality: The Second Erich L. Lehmann Symposium, vol. 49, Lecture Notes-Monograph Series. Beachwood, OH: Institute of Mathematical Statistics.

Mayo, D. G., and R. D. Hollander, eds. (1991), Acceptable Evidence: Science and Values in Risk Management. Oxford: Oxford University Press.

Mayo, D. G., and M. Kruse (2001), "Principles of Inference and Their Consequences," in D. Cornfield and J. Williamson (eds.). Foundations of Bayesianism. Dordrecht: Kluwer Academic Publishers, 381-403.

Mayo, D. G., and A. Spanos (2004), "Methodology in Practice: Statistical Misspecification Testing," Philosophy of Science 71: 1007-1025.

(2006), "Severe Testing as a Basic Concept in a Neyman-Pearson Philosophy of Induction," British Journal for the Philosophy of Science 57: 323-357.

- (2007), "Risks to Health and Risks to Science: The Need for a Responsible 'Bioev- idential' Scrutiny," Human and Experimental Toxicology, forthcoming.

NRC (National Research Council) (1983), Risk Assessment in the Federal Government. Wash- ington, DC: National Academy Press.

Shrader-Frechette, K. (1991), Risk and Rationality Berkeley: University of California Press. (2006), "Comparativist Philosophy of Science and Population Viability Assessment

in Biology: Helping Resolve Scientific Controversy," Philosophy of Science 73 (5), in this issue.

Thompson, P. B. (2006), "How Risky Are Genetically Engineered Crops? How Philosophers Can Help Answer the Question," Philosophy of Science Ti (5), in this issue.

__MACOSX/._phil+scrutiny+of+evidence+of+risk.pdf

risk+201107_OMBdraft-OzoneRIA.pdf

Regulatory Impact Analysis

Final National Ambient Air Quality Standard for Ozone

July 2011

U.S. Environmental Protection Agency Office of Air and Radiation

Office of Air Quality Planning and Standards Research Triangle Park, NC 27711

DOCKET NUMBER

TABLE OF CONTENTS

1.0 Summary

1.1 Results of Benefit‐Cost Analysis 4 1.2 Analysis of the Proposed Secondary NAAQS for Ozone 11 1.3 Baseline Emissions Inventory 12 1.4 Caveats and Conclusions 17

2.0 Re‐analysis of the Benefits of Attaining Alternative Ozone Standards to Incorporate Current Methods

2.1 Background 22 2.2 Key updates to the benefits assessment 23

2.3 Presentation of results 25 2.4 Comparison of results to previous results in 2008

Ozone NAAQS RIA 40 2.5 References 44 3.0 Secondary Ozone NAAQS Evaluation 3.1 Introduction 46 3.2 Air Quality Analysis 47 3.3 Complexities in Quantifying the Costs and Benefits of

Attaining a Secondary Ozone NAAQS 55 3.4 Pollution Control Strategies 57 3.5 Benefits of Reducing Ozone Effects on Vegetation

and Ecosystems 61 3.6 Additional Co‐benefits 79 3.7 References 82

Summary of the Supplemental Regulatory Impact Analysis (RIA) for the

Reconsideration of the 2008 Ozone National Ambient Air Quality Standard (NAAQS)

On September 16, 2009, EPA committed to reconsidering the ozone NAAQS standard

promulgated in March 2008. Today’s rule sets the ozone NAAQS at 0.070 ppm, based on this

reconsideration of the evidence available at the time the last standard was set. Today’s rule

also includes a separate secondary NAAQS, for which this RIA provides only qualitative analysis

due to the limited nature of available EPA guidance for attaining this standard

This supplement to the RIA contains an updated illustrative analysis of the potential

costs and human health and welfare benefits of nationally attaining a new primary ozone

standard of 0.070 ppm. The basis for this updated economic analysis is the RIA published in

March 2008 with changes. These changes reflect some significant methodological

improvements to air pollution benefits estimation, which EPA has adopted since the ozone

standard was last promulgated. These significant changes include the following:

 We have adopted several key methodological updates to benefits assessment since

the 2008 Ozone NAAQS RIA. These updates have already been incorporated into

previous RIAs for the Portland cement NESHAP, NO2 NAAQS RIA, and Category 3

Marine Diesel Engine Rule, and are therefore now incorporated in this analysis.

Significant updates include:

o We removed the assumption of no causality for ozone mortality, as

recommended by the National Academy of Science (NAS).

o We included two more ozone multi‐city studies, per NAS recommendation.

o We revised the Value of a Statistical Life (VSL) to be consistent with the value

used in current EPA analyses.

o We removed thresholds from the concentration‐response functions for PM2.5,

consistent with EPA’s Integrated Science Assessment for Particulate Matter.

The other elements of the illustrative analysis included in the March 2008 RIA were not

changed for this supplemental analysis. The March 2008 RIA was based on the best available

air quality modeling available and reflected emission reductions expected from federal rules

promulgated and proposed at that time. Because of the fundamental similarities between the

original and more recent air quality modeling simulations, EPA has elected not to update the

original analysis of emissions reductions needed to attain the ozone NAAQS as described in

Chapter 4 of the 2008 RIA. See section S1.3 below for discussion of the air quality baseline used

in this supplemental analysis.

Structure of this Updated RIA

As part of the ozone NAAQS reconsideration, this RIA supplement takes as its

foundation the 2008 ozone NAAQS RIA. Detailed explanation of the majority of assumptions

and methods are contained within that document and should be relied upon, except as noted

in this summary.

This supplement itself consists of four parts:

 Section 1 provides an overview of the changes to the analysis and summary tables of

the illustrative cost and benefits of obtaining a revised standard and alternatives of

0.065 ppm and 0.075 ppm.

 Section 2 contains a supplemental benefits analysis outlining the adopted changes in

the methodology, updated results for the final NAAQS of 0.070 ppm and standard

alternatives of 0.065 and 0.075 ppm using the revised methodology and

assumptions.

 Section 3 contains supplemental evaluation of a separate secondary ozone NAAQS

of 13 ppm‐hr, as well as a less stringent alternative of 15 ppm‐hr and a more

stringent alternative of 11 ppm‐hr. This supplemental includes an explanation of the

complexities associated with quantifying the costs and benefits of a secondary

standard at this time. In addition, we have incorporated an assessment of which

counties would have an additional requirement to reduce ozone concentrations to

meet a secondary standard beyond the reductions needed to meet the primary

standard, the qualitative benefits of reducing ozone exposure on vegetation, and

maps of biomass/yield loss avoided by attaining the primary and secondary ozone

standards.

S1.1 Results of Benefit‐Cost Analysis

This updated RIA consists of multiple analyses, including an assessment of the nature

and sources of ambient ozone; estimates of current and future emissions of relevant ozone

precursors; air quality analyses of baseline and alternative control strategies; illustrative control

strategies to attain the standard alternatives in future years; estimates of the incremental costs

and benefits of attaining the final standard and three alternative standards, together with an

examination of key uncertainties and limitations; and a series of conclusions and insights gained

from the analysis. It is important to recall that this RIA rests on the analysis done in 2008; no

new air quality modeling or other assessments were completed except those outlined above.

The supplement includes a presentation of the benefits and costs of attaining various

alternative ozone National Ambient Air Quality Standards in the year 2020. These estimates

only include areas assumed to meet the current standard by 2020. They do not include the

costs or benefits of attaining the alternate standards in the San Joaquin Valley and South Coast

air basins in California, because we expect that nonattainment designations under the Clean Air

Act for these areas would place them in categories afforded extra time beyond 2020 to attain

the ozone NAAQS.

[Hold for reference to Addendum]

In Table S1.1 below, the individual row estimates reflect the different studies available

to describe the relationship of ozone exposure to premature mortality. These monetized

benefits include reduced health effects from reduced exposure to ozone, reduced health

effects from reduced exposure to PM2.5, and improvements in visibility. The ranges within each

row reflect two PM mortality studies (i.e. Pope and Laden).

Ranges in the total costs column reflect different assumptions about the extrapolation

of costs as discussed in Chapter 5 of the 2008 Ozone NAAQS RIA. The low end of the range of

net benefits is constructed by subtracting the highest cost from the lowest benefit, while the

high end of the range is constructed by subtracting the lowest cost from the highest benefit.

The presentation of the net benefit estimates represents the widest possible range from this

analysis.

Table S1.2 presents the estimate of total ozone and PM2.5‐related premature mortalities

and morbidities avoided nationwide in 2020 as a result of this regulation.

Table S1. 1: Total Monetized Costs with Ozone Benefits and PM2.5 Co‐Benefits in 2020 (in Billions of 2006$) A

Ozone

Mortality

Function

Reference

Total Benefits B Total Costs C Net Benefits

3% 7% 7% 3% 7%

0 .0 7 5 p p m Multi‐city

Bell et al. 2004 $6.9 to $15 $6.4 to $13 $7.6 to $8.8 $‐1.9 to $7.4 $‐2.4 to $5.4

Schwartz 2005 $7.2 to $16 $6.8 to $13 $7.6 to $8.8 $‐1.6 to $8.4 $‐2.1 to $5.4

Huang 2005 $7.3 to $16 $6.9 to $13 $7.6 to $8.8 $‐1.5 to $8.4 $‐2.0 to $5.4

Meta‐

analysis

Bell et al. 2005 $8.3 to $17 $7.9 to $14 $7.6 to $8.8 $‐0.50 to $9.4 $‐1.0 to $6.4

Ito et al. 2005 $9.1 to $18 $8.7 to $15 $7.6 to $8.8 $0.30 to $10 $‐0.20 to $7.4

Levy et al. 2005 $9.2 to $18 $8.8 to $15 $7.6 to $8.8 $0.40 to $10 $‐0.10 to $7.4

0 .0 7 0 p p m Multi‐city

Bell et al. 2004 $13 to $29 $11 to $24 $19 to $25 $‐12 to $10 $‐14 to $5.0

Schwartz 2005 $15 to $30 $12 to $25 $19 to $25 $‐10 to $11 $‐13 to $6.0

Huang 2005 $15 to $30 $13 to $26 $19 to $25 $‐10 to $11 $‐12 to $7.0

Meta‐

analysis

Bell et al. 2005 $18 to $34 $16 to $29 $19 to $25 $‐7.0 to $15 $‐9.0 to $10

Ito et al. 2005 $21 to $37 $18 to $31 $19 to $25 $‐4.0 to $18 $‐6.0 to $12

Levy et al. 2005 $21 to $37 $18 to $31 $19 to $25 $‐4.0 to $18 $‐6.0 to $12

0 .0 6 5 p p m Multi‐city

Bell et al. 2004 $22 to $47 $19 to $40 $32 to $44 $‐22 to $15 $‐25 to $7.0

Schwartz 2005 $24 to $49 $21 to $42 $32 to $44 $‐20 to $17 $‐23 to $9.0

Huang 2005 $25 to $50 $22 to $42 $32 to $44 $‐19 to $18 $‐23 to $10

Meta‐

analysis

Bell et al. 2005 $31 to $56 $27 to $48 $32 to $44 $‐13 to $24 $‐17 to $16

Ito et al. 2005 $36 to $61 $32 to $53 $32 to $44 $‐8.0 to $29 $‐13 to $20

Levy et al. 2005 $36 to $61 $32 to $53 $32 to $44 $‐7.0 to $29 $‐12 to $20 A All estimates rounded to two significant figures. As such, they may not sum across columns. Only includes areas required to meet the current standard by 2020; does not include San Joaquin and South Coast areas in California.

B Includes ozone benefits, and PM2.5 co‐benefits. Range was developed by adding the estimate from the ozone premature mortality function to estimates from the PM2.5 premature mortality functions from Pope et al. and Laden et al. Tables exclude unquantified and nonmonetized benefits.

C Range reflects lower and upper bound cost estimates. Data for calculating costs at a 3% discount rate was not available for all sectors, and therefore total annualized costs at 3% are not presented here. Additionally, these estimates assume a particular trajectory of aggressive technological change. An alternative storyline might hypothesize a much less optimistic technological trajectory, with increased costs, or with decreased benefits in 2020 due to a later attainment date.

Table S1.2: Summary of Total Number of Ozone and PM2.5‐Related Premature Mortalities and

Premature Morbidity Avoided: 2020 National Benefits A

Combined Estimate of Mortality 0.075 ppm 0.070 ppm 0.065 ppm

NMMAPS Bell et al. (2004) 760 to 1,900 1,500 to 3,400 2,500 to 5,600

Schwartz 800 to 1,900 1,600 to 3,600 2,700 to 5,800

Huang 820 to 1,900 1,700 to 3,600 2,800 to 5,900

Meta‐analysis Bell et al. (2005) 930 to 2,000 2,000 to 4,000 3,500 to 6,600

Ito et al. 1,000 to 2,100 2,400 to 4,300 4,000 to 7,200

Levy et al. 1,000 to 2,100 2,400 to 4,300 4,100 to 7,200

Combined Estimate of Morbidity 0.075 ppm 0.070 ppm 0.065 ppm

Acute Myocardial Infarction B 1,300 2,200 3,500

Upper Respiratory Symptoms B 9,900 19,000 31,000

Lower Respiratory Symptoms B 13,000 25,000 41,000

Chronic Bronchitis B 470 880 1,400

Acute Bronchitis B 1,100 2,100 3,400

Asthma Exacerbation B 12,000 23,000 38,000

Work Loss Days B 88,000 170,000 270,000

School Loss Days C 190,000 600,000 1,100,000

Hospital and ER Visits 2,600 6,600 11,000

Minor Restricted Activity Days 1,000,000 2,600,000 4,500,000 A All estimates rounded to two significant figures. Only includes areas required to meet the current standard by 2020; does not include San Joaquin Valley and South Coast air basins in California. Includes ozone benefits, and PM2.5 co‐benefits. Mortality incidence range was developed by adding the estimate from the ozone premature mortality function to estimates from the PM2.5 premature mortality functions from Pope et al. (2002) and Laden et al. (2006).

B Estimated reduction in premature morbidity due to PM2.5 reductions only. C Estimated reduction in premature morbidity due to ozone reductions only.

The following set of graphs is included to provide the reader with a richer presentation

of the range of costs and benefits of the alternative standards. The graphs supplement the

tables by displaying all possible combinations of net benefits, utilizing the six different ozone

functions, the fourteen different PM functions, and the two cost methods. Each of the 168 bars

in each graph represents a separate point estimate of net benefits under a certain combination

of cost and benefit estimation methods. Because it is not a distribution, it is not possible to

infer the likelihood of any single net benefit estimate. The blue bars indicate combinations

where the net benefits are negative, whereas the green bars indicate combinations where net

benefits are positive. Figures S1.1 through S1.3 shows all of these combinations for all

standards analyzed. Figure S1.4 shows the comparison of total monetized benefits with costs

using the two benefits anchor points based on Pope/Bell 2004 and Laden/Levy.

Figure S1.1:

Figure S1.2:

$(100)

$(80)

$(60)

$(40)

$(20)

$‐

$20

$40

$60

$80

$100 B ill io n s o f 2 0 0 6 $

Combinations of 6 Ozone benefits estimates with 14 PM2.5 co‐benefits estimates with 2 costs estimates

Net Benefits for an Alternate Standard of 0.075 ppm (7% discount rate)

Benefits are greater than

costs

Costs are greater than

benefits

Median = $3.1b

$(100)

$(80)

$(60)

$(40)

$(20)

$‐

$20

$40

$60

$80

$100

B il li o n s o f 2 0 0 6 $

Combinations of 6 Ozone benefits estimates with 14 PM2.5 co‐benefits estimates with 2 costs estimates

Net Benefits for an Alternate Standard of 0.070 ppm (7% discount rate)

Benefits are greater than

costs

Costs are greater than

benefits

Median

= $1.4b

These graphs show all 168 combinations of the 6 different ozone mortality functions and assumptions, the 14 different PM mortality functions, and the 2 cost methods. These combinations do not represent a distribution.

Figure S1.3:

$(100)

$(80)

$(60)

$(40)

$(20)

$‐

$20

$40

$60

$80

$100 B il li o n s o f 2 0 0 6 $

Combinations of 6 Ozone benefits estimates with 14 PM2.5 co‐benefits estimates with 2 costs estimates

Net Benefits for an Alternate Standard of 0.065 ppm (7% discount rate)

Benefits are greater than costs

Costs are greater than

benefits

Median

= $0.7b

Figure S1.4:

The low benefits estimate is based on Pope/Bell 2004 and the high benefits estimate is based on Laden/Levy. The two cost estimates are based on two different extrapolated cost methodologies. These endpoints represent separate estimates based on separate methodologies. The dotted lines are a visual cue only, and these lines do not imply a uniform range between these endpoints.

$10

$20

$30

$40

$50

$60

$70

0.075 ppm 0.070 ppm 0.065 ppm

B ill io n s o f 2 0 0 6 $

Alternative Standard Level

Comparison of Total Monetized Benefits to Costs for Alternative Standard Levels in 2020 (Updated results, 7% discount rate)

S1.2 Analysis of the Proposed Secondary NAAQS for Ozone

Exposure to ozone has been associated with a wide array of vegetation and ecosystem

effects in the published literature. Sensitivity to ozone is highly variable across species, with

over 65 plan species identified as “ozone‐sensitive”, many of which occur in state and national

parks and forests. These effects include those that damage or impair the intended use of the

plant or ecosystem. Such effects are considered adverse to the public welfare and can include

reduced growth and/or biomass production in sensitive plant species, including forest trees,

reduced crop yields, visible foliar injury, reduced plant vigor (e.g., increased susceptibility to

harsh weather, disease, insect pest infestation, and competition), species composition shift,

and changes in ecosystems and associated ecosystem services.

This secondary NAAQS standard for ozone is the first secondary standard to be

promulgated with a form, averaging time, and level that is distinct from the health‐based

primary standard apart from the PM and SO2 regulations originally set in the early 1970s. The

index would be cumulated over the 12‐hour daylight window (8:00 a.m. to 8:00 p.m.) during

the consecutive 3‐month period during the ozone season with the maximum index value

(hereafter, referred to as W126). After reviewing the scientific evidence and public comments,

the Administrator selected a secondary ozone NAAQS at a level of 13 ppm‐hrs, using the W126

form, calculated as a 3‐year average of the annual sums.

Quantifying the costs and benefits of attaining a secondary NAAQS is an exceptionally

complex task, including unresolved issues related to the RIA analysis, air quality projection,

monitoring expansion, and implementation.1 Because of these complexities as well as limited

time and resources within the expedited schedule, we are limited in our ability to quantify the

costs and benefits of attaining a separate secondary NAAQS for ozone for this proposal.

However, we have incorporated an assessment of which counties would have an additional

requirement to reduce ozone concentrations to meet a secondary standard beyond the

reductions needed to meet the primary standard, the qualitative benefits of reducing ozone

exposure on vegetation, and maps of biomass/yield loss avoided by attaining the primary and

secondary ozone standards. Using a cumulative seasonal secondary standard (i.e., W126), we

evaluated alternate standard levels at 11, 13, and 15 ppm‐hours. Figure S1.5 shows the

counties projected to exceed a primary standard at 0.070 ppm and/or a secondary standard at

13 ppm‐hrs in the 2020 baseline.

1 These complexities are described in detail in Section S3.3.

Figure S1.5: Counties Projected to Exceed the Selected Primary and Secondary Standards in the Baseline in 2020*

* Many of the counties projected to exceed are in the South Coast and San Joaquin areas of California, which are not required to attain the primary standards by 2020.

S1.3 Baseline Emissions Inventory

EPA expects that the emissions reductions needed to attain the new ozone primary

standard may be less than what EPA originally predicted in the March 2008 RIA. Recent

updates to the emission and air quality modeling platform suggest that future baseline air

quality will be better than what was projected in the 2008 RIA. If the more recent projections

are better estimates of future ozone nonattainment in these areas, then the costs and benefits

of attaining the ozone NAAQS incremental to the current standard will likely be less than what

was projected as part of the 2008 RIA. However, there have also been a few rules promulgated

since the 2008 RIA baseline was developed that significantly affect ozone precursor emissions.

It is difficult to assess retroactively the net emissions impacts of these rules and how they

would likely affect total costs and benefits of the ozone NAAQS if they had been included in the

baseline. We discuss each of these baseline issues below.

(591 counties)

(70 counties)

(29 counties)

(15 counties)

Modeling Platform In March 2008, EPA completed a regulatory impacts analysis (RIA) that estimated the

potential costs and benefits of attaining a 0.075 ppm standard as well as several alternatives.

This illustrative analysis was based on the best available air quality modeling available at the

time of the original analysis. As described in Chapter 2 of the 2008 RIA, EPA used the

Community Multiscale Air Quality (CMAQ) model with inputs from a 2002 base year to project:

a) ozone concentrations in the future (i.e., 2020) and b) the amount of emissions reductions

that would be necessary to meet specified ozone targets. Since the original analysis, the CMAQ

model has been updated with several new science algorithms (Foley et al., 2010) and the base

year platform has been updated to include 2005 ambient data and model inputs. As part of this

NAAQS reconsideration, EPA completed a quick analysis to determine if the updates to the air

quality modeling system would substantially affect the original 2008 estimates of the control

costs needed to attain the new ozone standard.

One of the key elements in determining the amount of controls needed to attain a

particular ozone target is the estimate of how many areas will be above the chosen threshold in

the future and by how much they will exceed the goal. Greater amounts of residual

nonattainment will lead to greater amounts of needed emissions reductions which will lead to

higher attainment costs and benefits attributable to the new standard.

EPA compared model projections of 2020 eight‐hour ozone design values from the

original RIA (based on the 2002 platform) against the same 2020 model projections from the

latest air quality modeling simulations that use the current version of CMAQ and the more

recent base year (2005). In general, the 2020 design value estimates were very similar between

the two modeling exercises, as shown in Figure S1.6. For the 635 counties with eligible 2020

projections in both cases, the average difference between the most recent and the original

analysis was ‐0.15 ppb. That is, the updated analysis estimated slightly cleaner ozone values in

2020 than what EPA previously estimated in the original RIA. However, the two sets of

projections are very similar.

Figure S1.6: Comparison of projected 2020 eight‐hour ozone design values over 635 counties

in the U.S. from the original 2008 RIA air quality modeling (x‐axis) and a more recent

modeling analysis based on an updated model with updated model inputs (y‐axis)

The majority of the cost in attaining a new ozone standard will come from meeting the

target in the areas projected to be most polluted in the future. Limiting the analysis to only

those 61 counties where the ozone design values are projected to exceed 0.070 ppm in 2020

after the implementation of the controls in the hypothetical RIA control scenario, we see that

there is an even stronger tendency for the updated modeling to project cleaner conditions in

the future. As discussed in the original RIA, these 61 counties are primarily located in four

areas: California, Houston, the western Lake Michigan region, and the Northeast Corridor. For

this subset of locations, the average 2020 projected design value difference was ‐3.3 ppb. In

other words, the more recent EPA modeling predicts slightly cleaner ozone conditions (78.1 vs.

81.4 ppb) at the most polluted locations in the future. If the more recent projections are better

estimates of future ozone nonattainment in these areas, then the costs and benefits of

attaining the ozone NAAQS incremental to the current standard will likely be less than what

was projected as part of the 2008 RIA.

100

110

120

130

40 50 60 70 80 90 100 110 120 130

Original 2020 Ozone DV projections (ppb)

U p

d a te

d 2

0 2 0 O

zo n

e D

V p

ro je

c ti

o n

s (

p p

b )

Because of the fundamental similarities between the original and more recent air

quality modeling simulations, EPA has elected not to update the original analysis of emissions

reductions needed to attain the ozone NAAQS as described in Chapter 4 of the 2008 RIA. Based

on the latest air quality modeling information, however, it is expected that the original RIA

estimates of needed emissions reductions are greater than what is necessary to attain the new

primary standard.

Federal Rulemakings Included in the Baseline

The starting point for this analysis is the “baseline”, which represents what ambient air

quality would be nationwide in 2020 absent the revised ozone NAAQS. (2020 is when the

ozone NAAQS would be expected to be fully implemented in all areas except those with the

most significant air quality problems. Our analysis recognizes that two areas in Southern

California are not planning to meet the current standard by 2020.) The baseline for the revised

ozone standard is calculated using emissions estimates that include emission controls that will

be needed to attain the “current” standard by 2020. Since this rulemaking is a reconsideration

of the 0.080 ppm NAAQS, for this analysis the “current” standard is considered to be 0.08 ppm

(effectively 0.084 with rounding).

Two steps were used to develop the baseline for the March 2008 RIA. First, the

reductions expected nationwide in ozone concentrations from Federal rules in effect or

proposed at that time were included, as well as the controls applied as part of the PM2.5

NAAQS RIA analysis. The rules reflected in the modeling include:

 Clean Air Interstate Rule (EPA, 2005b)

 Clean Air Mercury Rule (EPA, 2005c)

 Regional Haze Regulations and Guidelines for Best Available Retrofit Technology

Determinations (EPA, 2005d)

 Clean Air Nonroad Diesel Rule (EPA, 2004)

 Light‐Duty Vehicle Tier 2 Rule (EPA, 1999)

 Heavy Duty Diesel Rule (EPA, 2000)

 Proposed rules for Locomotive and Marine Vessels (EPA, 2007a) and for Small Spark‐

Ignition Engines (EPA, 2007b)

 Proposed C3 Emission Control Area Rule (2009)

 State and local level mobile and stationary source controls identified for additional

reductions in emissions for the purpose of attaining the current PM 2.5 and Ozone

standards.

Second, since these reductions alone were not predicted to bring all areas into attainment with

the current standard, we used a hypothetical control strategy to apply additional known

controls. Additional control measures were used in four sectors to establish the baseline: Non‐

Electricity Generating Unit Point Sources (NonEGUs), Non‐Point Area Sources (Area), Onroad

Mobile Sources and Nonroad Mobile Sources.

Since the 2008 RIA was completed, a few other Federal rules significantly affecting

ozone precursor emissions have been promulgated. Also since that time, the Clean Air

Interstate Rule (CAIR) was remanded to EPA by the U.S. Court of Appeals for the D.C. Circuit and

the Clean Air Mercury Rule (CAMR) was vacated by the Court. These new developments suggest

that the baseline for this supplementary analysis does not reflect emission impacts expected

from some recent rules, and that it does reflect emission impacts from some rules that are no

longer in place.

Three major rules that were promulgated in 2010, and which affect large categories of

NOx emissions, should be represented in the baseline but are not. These are the Renewable

Fuel Standard (RFS2) and the Reciprocating Internal Combustion Engines (RICE) NESHAPs (2004

and 2010). It is difficult to assess retroactively how these rules would likely affect total costs of

the ozone NAAQS if they had been included in the baseline. NOx emissions from the two RICE

rules are estimated to have decreased by a total of about 165,000 tons per year in 2020.

However, NOx emissions are expected to increase by 247,600 tons in 2020 as a result of RFS2.

It is difficult to quantify the emission implications of having CAIR in the baseline for the

ozone analysis. In 2008, the U.S. Court of Appeals for the D.C. Circuit remanded CAIR to EPA.

(See http://www.epa.gov/CAIR/ for more background on CAIR and the Court ruling.) On July 6,

2010, EPA proposed the Transport Rule as a replacement for CAIR. For NOx, the Transport Rule

budget is lower in the near term and higher after 2015 relative to CAIR adjusting for differences

in the spatial coverage of the two rules. On net, annual NOx emissions are higher than under

CAIR once the replacement rule is in effect. Seasonal NOx emissions are lower with the

replacement rule, but this is because of differences in baseline emissions and is not attributable

to the replacement rule as emissions in the base case are lower than what is forecast with CAIR

compliance. Table S1.3 below summarizes the modeled emissions under CAIR and the

Transport Rule in various years.

Table S1.3. IPM Estimated Emissions Under CAIR and

CAIR Replacement Rule (Transport Rule)

National NOx Annual Emissions (Million Tons)

2010 2012 2015 2020 2025/2026

CAIR baseline 3.6 NA 3.7 3.7 NA

CAIR 2.4 NA 2.1 2.1 NA

Transport Rule baseline NA 3.0 3.0 3.1 3.1

Transport Rule main remedy NA 2.2 2.2 2.3 2.3

CAIR‐Region NOx Seasonal Emissions (Million Tons)

2010 2012 2015 2020 2025/2026

CAIR baseline 0.80 NA 0.80 0.80 NA

CAIR 0.70 NA 0.60 0.60 NA

Transport Rule baseline NA 0.40 0.40 0.41 0.42

Transport Rule main remedy NA 0.39 0.38 0.39 0.40

Source: CAIR results are taken from "EPA Base Case 2004” and “IPM Run CAIR 2004 Final" modeling output,

available at: http://www.epa.gov/airmarkt/progsregs/epa‐ipm/cair/index.html. Transport Rule results are taken

from “TR Base Case” and "TR SB Limited Trading" modeling output, available at:

http://www.epa.gov/airmarkets/progsregs/epa‐ipm/transport.html

At this time we are unable to assess the relative emission reductions expected from the

CAMR replacement rule relative to CAMR. In 2008, the U.S. Court of Appeals for the D.C. Circuit

vacated the Clean Air Mercury Rule (CAMR). (See http://www.epa.gov/mercuryrule/ for more

background CAMR and the Court ruling.) EPA intends to propose air toxics standards for power

plants consistent with the D.C. Circuit’s opinion regarding the CAMR by March 10, 2011 and

finalize a rule by November 16, 2011.

S1.4 Caveats and Conclusions

Of critical importance to understanding these estimates of future costs and benefits is

that they are not intended to be forecasts of the actual costs and benefits of implementing

revised standards. There are many challenges in estimating the costs and benefits of attaining a

tighter ozone standard, which are fully discussed in 2008 Ozone NAAQS RIA and the

supplement to this analysis accompanying today’s final rule.

There are significant uncertainties in both cost and benefit estimates for the full range of

standard alternatives. Below we summarize some of the more significant sources of

uncertainty common to all level analyzed in the 2008 ozone NAAQS RIA and this supplemental

analysis:

 Benefits estimates are influenced by our ability to accurately model relationships

between ozone and PM and their associated health effects (e.g., premature

mortality).

 Benefits estimates are also heavily dependent upon the choice of the statistical

model chosen for each health benefit.

 PM co‐benefits are derived primarily from reductions in nitrates (associated with

NOx controls). As such, these estimates are strongly influenced by the assumption

that all PM components are equally toxic. Co‐benefit estimates are also influenced

by the extent to which a particular area chooses to use NOx controls rather than

VOC controls.

 There are several nonquantified benefits (e.g., effects of reduced ozone on forest

health and agricultural crop production) and disbenefits (e.g., decreases in

tropospheric ozone lead to reduced screening of UV‐B rays and reduced nitrogen

fertilization of forests and cropland) discussed in this analysis in Chapter 6 of the

2008 Ozone NAAQS RIA.

 Changes in air quality as a result of controls are not expected to be uniform over the

country. In our hypothetical control scenario some increases in ozone levels occur in

areas already in attainment, though not enough to push the areas into

nonattainment.

 As explained in Chapter 5 of the 2008 Ozone NAAQS RIA, there are several

uncertainties in our cost estimates. For example, the states are likely to use different

approaches for reducing NOx and VOCs in their state implementation plans to reach

a tighter standard. In addition, since our modeling of known controls does not get all

areas into attainment, we needed to make assumptions about the costs of control

technologies that might be developed in the future and used to meet the tighter

alternative. For example, for the 21 counties (in four geographic areas) that are not

expected to attain 0.075 ppm2 in 20203, assumed costs of unspecified controls

represent a substantial fraction, of the costs estimated in this analysis ranging from

50% to 89% of total costs depending on the standard being analyzed.

 As discussed in Chapter 5 of the 2008 Ozone NAAQS RIA, advice from EPA’s

Science Advisory Board has questioned the appropriateness of an approach

similar to one of those used here for estimating extrapolated costs. For balance,

EPA also applied a methodology recommended by the Science Advisory Board in

an effort to best approximate the costs of control technologies that might be

developed in the future.

2 Areas that do not meet 0.075 ppm are Chicago, Houston, the Northeastern Corridor, and Sacramento. For more information see chapter 4 section 4.1.1 of the 2008 Ozone NAAQS RIA. 3 This list of areas does not include the San Joaquin and South Coast air basins who are not expected to attain the current 0.084 ppm standard until 2024.

 Both extrapolated costs and benefits have additional uncertainty relative to

modeled costs and benefits. The extrapolated costs and benefits will only be

realized to the extent that unknown extrapolated controls are economically

feasible and are implemented. Technological advances over time will tend to

increase the economic feasibility of reducing emissions, and will tend to reduce

the costs of reducing emissions. Our estimates of costs of attainment in 2020

assume a particular trajectory of aggressive technological change. This

trajectory leads to a particular level of emissions reductions and costs which we

have estimated based on two different approaches, the fixed cost and hybrid

approaches. An alternative storyline might hypothesize a much less optimistic

technological change path, such that emissions reductions technologies for

industrial sources would be more expensive or would be unavailable, so that

emissions reductions from many smaller sources might be required for 2020

attainment, at a potentially greater cost per ton. Under this alternative

storyline, two outcomes are hypothetically possible: Under one scenario, total

costs associated with full attainment might be substantially higher. Under the

second scenario, states may choose to take advantage of flexibility in the Clean

Air Act to adopt plan with later attainment dates to allow for additional

technologies to be developed and for existing programs like EPA’s Onroad Diesel,

Nonroad Diesel, and Locomotive and Marine rules to be fully implemented. If

states were to submit plans with attainment dates beyond our 2020 analysis

year, benefits would clearly be lower than we have estimated under our

analytical storyline. However, in this case, state decision makers seeking to

maximize economic efficiency would not impose costs, including potential

opportunity costs of not meeting their attainment date, when they exceed the

expected health benefits that states would realize from meeting their modeled

2020 attainment date. In this case, upper bound costs are difficult to estimate

because we do not have an estimate of the point where marginal costs are equal

to marginal benefits plus the costs of nonattainment. Clearly, the second stage

analysis is a highly speculative exercise, because it is based on estimating

emission reductions and air quality improvements without any information

about the specific controls that would be available to do so.

Appendix S1.A: Reductions of Criteria Air Pollutants from Travel Efficiency Strategies

The RIA contains only a minimal analysis of travel efficiency strategies to reduce vehicle

miles traveled, and thus reduce emissions of NOx and other pollutants. A recent report titled,

Moving Cooler: An Analysis of Transportation Strategies for Reducing Greenhouse Gas

Emissions4, which EPA and US DOT helped to fund, analyzed the potential levels of emissions

reductions from light‐duty travel efficiency. Moving Cooler included six different bundles of

strategies to reflect different potential groups of strategies that could be implemented. Using

data from this report, EPA conducted an analysis of the air quality benefits of a subset of the

travel efficiency strategies evaluated in the report. Below are preliminary results based on

EPA’s draft MOVES2009 Model.

For the purposes of EPA's analysis, we chose the "Low Cost" bundle because we

believed that it represented the best combination of strategies based on cost, likelihood of

success, and accuracy of the research results. This bundle included strategies like smart

growth/transit, commuter strategies, system operations (e.g., eco‐driving, ramp metering),

pricing (e.g., parking taxes, congestion pricing, intercity tolls), speed limit restrictions, and

multimodal freight strategies. Note that this bundle did not include a VMT tax or cap‐and‐trade

assumptions.

Moving Cooler made assumptions about the geographic scope for which each strategy

could be implemented, with certain strategies like transit being dependent on greater

populations, while other strategies like speed limit restrictions could be implemented in both

urban and rural areas. Adjustments were also made to operational and commuter strategies to

account for induced demand impacts. Scenarios A and B represent aggressive and maximum

deployment, respectively, of the “Low Cost” bundle of strategies in Moving Cooler.

Summary of Results

 Nationally, the modeled travel efficiency strategies would reduce exhaust PM2.5, NOx,

HC and CO from cars and light trucks by approximately 2% in 2020, to approximately 7%

in 2045, under the “aggressive” Moving Cooler assumptions.

 The modeled travel efficiency strategies would reduce these emissions by approximately

5% in 2020, to approximately 11% in 2045, under the “maximum” Moving Cooler

assumptions.

4 Cambridge Systematics, Inc. (2009). Moving Cooler: An Analysis of Transportation Strategies for Reducing Greenhouse Gas Emissions. Urban Land Institute: Washington, D.C.

 Percent reductions would be larger in urban areas, where Moving Cooler VMT

reductions are concentrated.

Detailed Results

U.S. Annual Ton Reductions from Moving Cooler Bundle 6

"Maximum" Reductions HC NOx PM2.5 CO

Tons % LD Tons % LD Tons % LD Tons % LD 2010 3,437 0.2% 6,542 0.2% 116 0.2% 63,414 0.2% 2015 21,902 1.5% 44,447 1.4% 909 1.5% 486,532 1.4% 2020 50,756 5.2% 101,773 4.8% 2,881 5.2% 1,379,197 4.9% 2025 55,130 7.4% 109,888 6.9% 4,157 7.6% 1,848,978 7.1% 2030 55,569 8.5% 109,193 7.9% 5,039 9.0% 2,157,685 8.3% 2035 56,701 9.3% 111,183 8.7% 5,794 10.0% 2,401,437 9.2% 2040 62,517 10.1% 125,017 9.5% 6,659 10.9% 2,723,906 10.1% 2045 69,934 11.0% 142,747 10.3% 7,631 11.9% 3,088,375 11.0%

"Aggressive" Reductions HC NOx PM2.5 CO

Tons % LD Tons % LD Tons % LD Tons % LD 2010 1,222 0.1% 2,325 0.1% 41 0.1% 22,536 0.1% 2015 6,345 0.4% 13,079 0.4% 265 0.4% 143,319 0.4% 2020 22,088 2.3% 44,291 2.1% 1,254 2.3% 600,223 2.1% 2025 30,592 4.1% 61,366 3.8% 2,301 4.2% 1,029,945 3.9% 2030 33,256 5.1% 65,633 4.8% 3,005 5.4% 1,292,924 5.0% 2035 35,727 5.8% 70,298 5.5% 3,638 6.3% 1,513,454 5.8% 2040 39,897 6.5% 80,020 6.1% 4,236 6.9% 1,738,288 6.5% 2045 45,181 7.1% 92,460 6.7% 4,916 7.6% 1,995,082 7.1%

1 SECTION 2: RE‐ANALYSIS OF THE BENEFITS OF ATTAINING ALTERNATIVE OZONE

STANDARDS TO INCORPORATE CURRENT METHODS Synopsis

This chapter presents a benefits analysis of three alternate ozone standards updated to reflect key methodological changes that EPA implemented after publishing the 2008 Ozone NAAQS RIA. Since the completion of this analysis EPA has introduced several methodological improvements in other RIA’s that are not incorporated in this analysis.5 In this updated analysis we re-estimate the human health benefits of reduced exposure to ambient ozone and PM2.5 co- benefits from simulated attainment with the selected daily 8hr maximum standard of 0.070 ppm and two alternate standards of 0.075 ppm and 0.065 ppm. For the selected standard of 0.070 ppm, EPA estimates the monetized benefits to be $13 to $37 billion (2006$, 3% discount rate) in 2020. For an alternative standard at 0.075 ppm, EPA estimates the monetized benefits to be $6.9 to $18 billion (2006$, 3% discount rate) in 2020.6 For the alternative standard at 0.065 ppm, EPA estimates the monetized benefits to be $22 to $61 billion (2006$, 3% discount rate) in 2020. Higher or lower estimates of benefits are possible using other assumptions. These updated estimates reflect three key methodological changes we have implemented since the publication of the 2008 RIA that reflect EPA’s most current interpretation of the scientific literature and include: (1) a no-threshold model for PM2.5 that calculates incremental benefits down to the lowest modeled air quality levels; (2) removal of the assumption of no causality for the relationship between ozone exposure and premature mortality; (3) a different Value of Statistical Life (VSL). These benefits are incremental to an air quality baseline that reflects attainment with the 1997 ozone and 2006 PM2.5 National Ambient Air Quality Standards (NAAQS). Methodological limitations prevented EPA from monetizing the benefits from several important benefit categories, including ecosystem effects. S2.1 Background

In response to the recent court vacatur of the 2008 Ozone NAAQS, EPA is reconsidering this rulemaking. Consistent with EPA’s decision to, in general, use the “existing record” for this reconsideration, we present a benefits analysis based on the same air quality modeling inputs as the 2008 analysis. However, we update this analysis to make the results consistent with an array of methodological updates that EPA has incorporated since the release of Regulatory Impact Analysis (RIA) for the 2008 Ozone NAAQS (U.S. EPA, 2008). Because the rulemaking period for the reconsideration is condensed, we only provide estimates associated with the promulgated standard level of 0.070 ppm and the two less stringent standard levels previously analysis (i.e.,

5 Such improvements include the use of more current baseline mortality and morbidity rates to calculate health impacts and the use of more recent PM health studies to calculate health impacts. The effect of these changes would be to reduce certain ozone and PM2.5-related health impacts reported in this RIA by a modest amount. 6 Results are shown as a range from Bell et al. (2004) with Pope et al. (2002) to Levy (2005) with Laden et al. (2006). PM2.5 co-benefits using a 7% discount rate would be approximately 9% lower.

0.065 ppm and 0.075 ppm). All benefits estimates in this analysis are incremental to the 1997 Ozone NAAQS standard at 0.08 ppm and the 2006 PM2.5 NAAQS standard at 15/35 µg/m3. S2.2 Key updates to the benefits assessment

In this analysis, we update several aspects of our benefits assessment for the human health benefits of reducing exposure to ozone and PM2.5.7 Both ozone benefits and PM2.5 co- benefits incorporate the updated population projections in BenMAP. In addition, both ozone benefits and PM2.5 co-benefits reflect EPA’s current interpretation of the economic literature on mortality valuation to use the value-of-a statistical life (VSL) based on meta-analysis of 26 studies.8

For ozone benefits, these updates are a response to recent recommendations from the

National Research Council (NRC, 2008). In this analysis, we have incorporated three of NRC’s recommendations:

1) We no longer include estimates of ozone benefits with an assumption of no

causal relationship between ozone exposure and premature mortality.

2) We include two additional ozone mortality estimates, one based on the National

Morbidity, Mortality and Air Pollution Study (NMMAPS) (Huang, 2005), and one

14‐city study (Schwartz, 2005), placing the greatest emphasis on the multi‐city

studies, such as NMMAPS.

3) We present additional risk metrics, including the change in the percentage of

baseline mortality attributable, and the number of life years lost due, to ozone‐

related premature mortality.

In addition to these recommendations, we modify the health functions used to estimate

the number of emergency department visits for asthma avoided by reducing exposure to

ozone. Specifically, we removed the Jaffe et al. (2003) function because the age range overlaps

partially with Wilson et al. (2005) and Peel et al. (2005) functions. This change results in a

slightly larger estimate of ozone‐related emergency department visits as compared to the 2008

analysis.

For PM2.5 co-benefits, this analysis is consistent with proposed Portland Cement NESHAP RIA (U.S. EPA, 2009a) and proposed NO2 NAAQS RIA (U.S. EPA, 2009b). In this analysis, we incorporate four updates: 7 This analysis does not attempt to describe the overall methodology for estimating the benefits of reducing ozone and PM2.5. For more information, please consult Chapter 6 of the 2008 Ozone NAAQS RIA (U.S. EPA, 2008). 8 For more information regarding mortality valuation, please consult section 5.7 of the proposed NO2 RIA (U.S. EPA, 2009b).

1) We removed assumed thresholds from the mortality and morbidity

concentration‐response functions for PM2.5. 9 Removing the assumed 10 µg/m3

threshold is a key difference between the method used in this analysis of PM2.5‐

co benefits and the methods used in RIAs prior to Portland Cement, and we now

calculate incremental benefits down to the lowest modeled PM2.5 air quality

levels. This change results in a larger estimate of PM‐related premature mortality

as compared to the 2008 analysis.

2) We now present the PM2.5 co‐benefits results using concentration‐response

functions for mortality from two cohort studies (Pope et al. (2002) and Laden et

al. (2006)) instead of range between the minimum and maximum results from an

expert elicitation of the relationship between exposure to PM2.5 and premature

mortality (Roman et al., 2008). This change produces a slightly narrower range of

PM‐related mortality estimates as compared to the 2008 analysis.

3) When adjusting the benefits of the modeled PM co‐benefits for alternate

standard levels, we apply PM2.5 benefit per ton estimates calculated using a

broader geographic area, which, when compared to the 2008 analysis, produces

more reliable and generally larger PM‐related benefits estimates.

4) We incorporated an updated methodology for quantifying the health incidences

associated with the benefit‐per‐ton estimates. This change should produce more

reliable estimates of PM‐related health impacts.

In this analysis we estimate ozone‐related premature mortality using risk coefficients

drawn from short‐term mortality studies. Two recent epidemiologic studies assessed the

relationship between long‐term exposure to ozone and premature mortality. Jerrett et al.

(2009) utilized the ACS cohort with air quality data from 1977 through 2000 (April through

September). Jarrett et al. reported a positive and statistically significant association between

ambient ozone concentration and respiratory causes of death after controlling for PM2.5 using

co‐pollutant models. Further examination of the association between ozone exposure and

respiratory‐related mortality revealed the association was increased by higher temperatures

and geographic variation. In single pollutant models, long‐term ozone exposure was also

associated with cardiopulmonary, cardiovascular, and ischemic heart disease mortality, but the

associations were not present in the co‐pollutant model. Krewski et al. (2009) also utilized data

from the ACS cohort with air quality data from 1980 (April through September) and observed a

positive association between ozone exposure and all‐cause and cardiopulmonary disease

mortality. This association was robust to control for ecologic variables, but no association was

9 For more information regarding thresholds in the PM2.5 mortality relationship, please consult the proposed Portland Cement NESHAP RIA (U.S. EPA, 2009a).

observed with ischemic heart disease or lung cancer. In addition, Krewski et al. observed no

association with year‐round ozone exposure.

S2.3 Presentation of results

Tables S2.1 through S2.6 show the results of this updated analysis. Figures S2.1 and

S2.2 show the breakdown of ozone benefits and PM2.5 co‐benefits by endpoint category using a

single mortality study as an example. Figures S2.3 and S2.4 show the ozone benefits and PM2.5

co‐benefits by mortality study. Figures S2.5 and S2.6 show the breakdown of monetized

benefits between ozone, PM, morbidity, mortality, and visibility. Figure S2.7 shows the results

of this updated analysis graphically.

Table S2.1: Summary of Total Number of Ozone and PM2.5-Related Premature Mortalities and Morbidity Incidences Avoided in 2020 A

Combined Estimate of Mortality 0.075 ppm 0.070 ppm 0.065 ppm

Multi-city Bell et al. (2004) 760 to 1,900 1,500 to 3,400 2,500 to 5,600

Schwartz 800 to 1,900 1,600 to 3,600 2,700 to 5,800

Huang 820 to 1,900 1,700 to 3,600 2,800 to 5,900

Meta-analysis Bell et al. (2005) 930 to 2,000 2,000 to 4,000 3,500 to 6,600

Ito et al. 1,000 to 2,100 2,400 to 4,300 4,000 to 7,200

Levy et al. 1,000 to 2,100 2,400 to 4,300 4,100 to 7,200

Combined Estimate of Morbidity 0.075 ppm 0.070 ppm 0.065 ppm

Acute Myocardial Infarction B 1,300 2,200 3,500

Upper Respiratory Symptoms B 9,900 19,000 31,000

Lower Respiratory Symptoms B 13,000 25,000 41,000

Chronic Bronchitis B 470 880 1,400

Acute Bronchitis B 1,100 2,100 3,400

Asthma Exacerbation B 12,000 23,000 38,000

Work Loss Days B 88,000 170,000 270,000

School Loss Days C 190,000 600,000 1,100,000

Hospital and ER Visits 2,600 6,600 11,000

B Estimated reduction in premature morbidity due to PM2.5 reductions only. C Estimated reduction in premature morbidity due to ozone reductions only.

Table S2.2: Summary of Total Monetized Benefits in 2020 (3% discount rate, in millions of

2006$)A, B, C Combined Estimate of Mortality 0.075 ppm 0.070 ppm 0.065 ppm

NMMAPS Bell et al. (2004) $6,900 to $15,000 $13,000 to $29,000 $22,000 to $47,000

Schwartz $7,200 to $16,000 $15,000 to $30,000 $24,000 to $49,000

Huang $7,300 to $16,000 $15,000 to $30,000 $25,000 to $50,000

Meta-analysis Bell et al. (2005) $8,300 to $17,000 $18,000 to $34,000 $31,000 to $56,000

Ito et al. $9,100 to $18,000 $21,000 to $37,000 $36,000 to $61,000

Levy et al. $9,200 to $18,000 $21,000 to $37,000 $36,000 to $61,000 A Does not reflect estimates for the San Joaquin and South Coast Air Basins B All estimates rounded to two significant digits c Includes Visibility benefits of $160,000 Table S2.3: Summary of Total Monetized Benefits in 2020 (7% discount rate, in millions of

2006$)A, B, C Combined Estimate of Mortality 0.075 ppm 0.070 ppm 0.065 ppm

NMMAPS Bell et al. (2004) $6,400 to $13,000 $11,000 to $24,000 $19,000 to $39,000

Schwartz $6,700 to $13,000 $12,000 to $25,000 $21,000 to $41,000

Huang $6,800 to $13,000 $13,000 to $26,000 $21,000 to $42,000

Meta-analysis Bell et al. (2005) $7,800 to $14,000 $16,000 to $29,000 $27,000 to $48,000

Ito et al. $8,600 to $15,000 $18,000 to $31,000 $31,000 to $52,000

Levy et al. $8,700 to $15,000 $18,000 to $31,000 $32,000 to $52,000 A Does not reflect estimates for the San Joaquin and South Coast Air Basins B All estimates rounded to two significant digits c Includes Visibility benefits of $160,000

Figure S2-1: Breakdown of Ozone Health Benefits (using Bell 2004)*

*This pie chart breakdown is illustrative, using the results based on Bell et al. (2004) as an example. Using the Levy et al. (2006) function for premature mortality, the percentage of total monetized benefits due to adult mortality would be 97%.

Figure S2-2: Breakdown of PM2.5 Health Benefits (using Pope)*

*This pie chart breakdown is illustrative, using the results based on Pope et al. (2002) as an example. Using the Laden et al. (2006) function for premature mortality, the percentage of total monetized benefits due to adult mortality would be 97%. This chart shows the breakdown using a 3% discount rate, and the results would be similar if a 7% discount rate was used.

Infant Hospital Admissions 1.5%

ER Visits 0.02% School Loss Days

2.3% Acute Resp Symptoms 4.1%

Adult Hospital Admissions

1.8%

Adult Mortality Bell et al. (2004) 90%

Adult Mortality ‐ Pope et al. 93%

Chronic Bronchitis 4%

AMI 2%

Acute Respiratory Symptoms 0.5%

Infant Mortality 0.4%

Work Loss Days 0.2%

Hospital Admissions, Cardio 0.2%

Hospital Admissions, Resp 0.04%

Asthma Exacerbation 0.01%

Acute Bronchitis 0.01%

Upper Resp Symp 0.00% Lower Resp Symp 0.00%

ER Visits, Resp 0.00%

Other 1%

Table S2.4: Summary of National Ozone Benefits by Standard Level with 95th percentile confidence intervals (in millions of 2006$)A, B, C

Endpoint Group Author 0.075 ppm Valuation

0.075 ppm Incidence

0.070 ppm Valuation

0.070 ppm Incidence

0.065 ppm Valuation

0.065 ppm Incidence

Infant Hospital Admissions, Respiratory $11 550 $17 1,700 $30 3,000

($5.7 -- $16) (310 -- 830) ($8.5 -- $25) (960 -- 2,600) ($15 -- $43) (1,700 -- 4,500)

Emergency Room Visits, Respiratory $0.11 290 $0.36 990 $0.66 1,800

(-$.21 -- $.35) (-310 -- 930) (-$.71 -- $1.2) (-890 -- 3,200) (-$1.3 -- $2.2) (-1,600 -- 5,800)

School Loss Days $17 190,000 $53 600,000 $96 1,100,000

($7.5 -- $24) (93,000 -- 280,000) ($23 -- $76) (300,000 -- 880,000) ($42 -- $140) (550,000 -- 1,600,000)

Acute Respiratory Symptoms $30 510,000 $96 1,600,000 $170 2,900,000

($12 -- $56) (280,000 -- 790,000) ($37 -- $180) (910,000 -- 2,500,000) ($68 -- $320) (1,700,000 -- 4,500,000)

Hospital Admissions, Respiratory $13 550 $45 1,900 $81 3,400

($1.7 -- $22) (130 -- 980) ($5.6 -- $77) (550 -- 3,400) ($11 -- $140) (1,000 -- 6,100)

Mortality Bell et al. 2004 $660 74 $2,200 250 $4,000 450

($54 -- $2,000) (36 -- 120) ($180 -- $6,600) (130 -- 410) ($330 -- $12,000) (240 -- 730)

Mortality Schwartz $1,000 110 $3,400 380 $6,200 700

($82 -- $3,000) (54 -- 190) ($270 -- $10,000) (190 -- 630) ($500 -- $19,000) (350 -- 1,100)

Mortality Huang $1,100 130 $3,800 420 $6,800 770

($95 -- $3,300) (66 -- 200) ($320 -- $11,000) (230 -- 670) ($580 -- $20,000) (420 -- 1,200)

Mortality Bell et al. 2005 $2,000 240 $7,000 800 $10,000 1,500

($190 -- $6,100) (140 -- 350) ($630 -- $21,000) (490 -- 1,200) ($1,100 -- $37,000) (910 -- 2,200)

Mortality Ito et al. $2,900 330 $9,900 1,100 $18,000 2,000

($280 -- $8,200) (230 -- 450) ($930 -- $28,000) (790 -- 1,500) ($1,700 -- $50,000) (1,400 -- 2,800)

Mortality Levy et al. $3,000 340 $10,000 1,100 $18,000 2,100

($280 -- $8,200) (260 -- 430) ($930 -- $28,000) (870 -- 1,500) ($1,700 -- $50,000) (1,600 -- 2,600) A Does not reflect estimates for the San Joaquin and South Coast Air Basins B Confidence intervals are not available for PM co-benefits because of methodological limitations when using benefit-per-ton estimates. C All estimates rounded to two significant digits

Table S2.5: Summary of National Ozone Benefits and PM2.5 Co-Benefits by Standard Level (in millions of 2006$ at a 3% discount rate)A, B, C

Endpoint Group Author 0.075 ppm Valuation

0.075 ppm Incidence

0.070 ppm Valuation

0.070 ppm Incidence

0.065 ppm Valuation

0.065 ppm Incidence

O zo

n e

Infant Hospital Admissions, Respiratory $11 550 $17 1,700 $30 3,000

Emergency Room Visits, Respiratory $0.11 290 $0.36 990 $0.66 1,800

School Loss Days $17 190,000 $53 600,000 $96 1,100,000

Acute Respiratory Symptoms $30 510,000 $96 1,600,000 $170 2,900,000

Hospital Admissions, Respiratory $13 550 $45 1,900 $81 3,400

Mortality Bell et al. (2004) $660 74 $2,200 250 $4,000 450

Mortality Schwartz $1,000 110 $3,400 380 $6,200 700

Mortality Huang $1,100 130 $3,800 420 $6,800 770

Mortality Bell et al. (2005) $2,100 240 $7,100 800 $13,000 1,500

Mortality Ito et al. $2,900 330 $9,900 1,100 $18,000 2,000

Mortality Levy et al. $3,000 340 $10,000 1,100 $18,000 2,100

P M

2. 5

Chronic Bronchitis $230 470 $430 880 $700 1,400

Acute Myocardial Infarction $140 1,300 $240 2,200 $380 3,500

Hospital Admissions, Respiratory $2.5 180 $4.3 310 $6.8 490

Hospital Admissions, Cardiovascular $11 390 $18 670 $29 1,000

Emergency Room Visits, Respiratory $0.22 590 $0.39 1,100 $0.63 1,700

Acute Bronchitis $0.08 1,100 $0.15 2,100 $0.25 3,400

Work Loss Days $11 88,000 $20 170,000 $34 270,000

Asthma Exacerbation $0.64 12,000 $1.2 23,000 $2.0 38,000

Acute Respiratory Symptoms $31 520,000 $58 980,000 $95 1,600,000

Lower Respiratory Symptoms $0.24 13,000 $0.45 25,000 $0.75 41,000

Upper Respiratory Symptoms $0.29 9,900 $0.54 19,000 $0.89 31,000

Infant Mortality $22 3 $44 5 $73 8

Mortality Pope et al $5,500 690 $10,000 1,200 $16,000 2,000

Mortality Laden et al $14,000 1,800 $26,000 3,200 $41,000 5,100

Mortality Expert K $1,900 230 $3,500 430 $5,700 700

Mortality Expert E $19,000 2,300 $34,000 4,200 $55,000 6,800 A Does not reflect estimates for the San Joaquin and South Coast Air Basins B Does not include confidence intervals

C All estimates rounded to two significant digits

Table S2.6: Summary of National Ozone Benefits and PM2.5 Co-Benefits by Standard Level (in millions of 2006$ at a 7%

Endpoint Group Author 0.075 ppm Valuation

0.075 ppm Incidence

0.070 ppm Valuation

0.070 ppm Incidence

0.065 ppm Valuation

0.065 ppm Incidence

O zo

n e

Infant Hospital Admissions, Respiratory $11 550 $17 1,700 $30 3,000

Emergency Room Visits, Respiratory $0.11 290 $0.36 990 $0.66 1,800

School Loss Days $17 190,000 $53 600,000 $96 1,100,000

Acute Respiratory Symptoms $30 510,000 $96 1,600,000 $170 2,900,000

Hospital Admissions, Respiratory $13 550 $45 1,900 $81 3,400

Mortality Bell et al. (2004) $660 74 $2,200 250 $4,000 450

Mortality Schwartz $1,000 110 $3,400 380 $6,200 700

Mortality Huang $1,100 130 $3,800 420 $6,800 770

Mortality Bell et al. (2005) $2,100 240 $7,100 800 $13,000 1,500

Mortality Ito et al. $2,900 330 $9,900 1,100 $18,000 2,000

Mortality Levy et al. $3,000 340 $10,000 1,100 $18,000 2,100

P M

2. 5

Chronic Bronchitis $230 470 $430 880 $700 1,400

Acute Myocardial Infarction $140 1,300 $240 2,200 $380 3,500

Hospital Admissions, Respiratory $2.5 180 $4.3 310 $6.8 490

Hospital Admissions, Cardiovascular $11 390 $18 670 $29 1,000

Emergency Room Visits, Respiratory $0.22 590 $0.39 1,100 $0.63 1,700

Acute Bronchitis $0.08 1,100 $0.15 2,100 $0.25 3,400

Work Loss Days $11 88,000 $20 170,000 $34 270,000

Asthma Exacerbation $0.64 12,000 $1.2 23,000 $2.0 38,000

Acute Respiratory Symptoms $31 520,000 $58 980,000 $95 1,600,000

Lower Respiratory Symptoms $0.24 13,000 $0.45 25,000 $0.75 41,000

Upper Respiratory Symptoms $0.29 9,900 $0.54 19,000 $0.89 31,000

Infant Mortality $22 3 $44 5 $73 8

Mortality Pope et al $5,000 690 $9,000 1,200 $14,000 2,000

Mortality Laden et al $13,000 1,800 $23,000 3,200 $37,000 5,100

Mortality Expert K $1,700 230 $3,100 430 $5,100 700

Mortality Expert E $17,000 2,300 $31,000 4,200 $49,000 6,800 A Does not reflect estimates for the San Joaquin and South Coast Air Basins B Does not include confidence intervals C All estimates rounded to two significant digits

Figure S2.3: Ozone benefits for Alternate Standard Levels*

*This graph shows the estimated ozone benefits in 2020 using three NMMAPS‐based epidemiology studies and three meta‐analyses. The results shown are not the direct results from the studies; rather, the estimates are based in part on the concentration‐response function provided in those studies. Because all ozone‐related health effects are short‐ term, the discount rate does not affect the results.

Figure S2.4: PM2.5 co‐benefits for Alternate Standard Levels*

*This graph shows the estimated PM2.5 co‐benefits in 2020 using the no‐threshold model at discount rates of 3% using effect coefficients using the Pope et al. study and the Laden et al study, as well as 12 effect coefficients derived from EPA’s expert elicitation on PM mortality. The results shown are not the direct results from the studies or expert elicitation; rather, the estimates are based in part on the concentration‐response function provided in those studies. Results using a 7% discount rate would be similar, but approximately 9% lower.

$10

$15

$20

Ito et al. Schwartz Bell et al. 2004 Levy et al. Bell et al. 2005 Huang

B ill io ns o f 2 00 6$

NMMAPS Epidemiology study or Meta‐Analysis

0.075 ppm 0.070 ppm 0.065 ppm

$10

$20

$30

$40

$50

$60

Pope et al.

Laden et al.

Expert A

Expert B

Expert C

Expert D

Expert E

Expert F

Expert G

Expert H

Expert I

Expert J

Expert K

Expert L

B ill io ns o f 2 00 6$

Epidemiology study or expert

0.075 ppm 0.070 ppm 0.065 ppm

Figure S2.5: Breakdown of total monetized benefits for Alternate Standard Levels (Low)

Figure S2.6: Breakdown of total monetized benefits for Alternate Standard Levels (High)

$10

$20

$30

$40

$50

$60

0.075 ppm 0.070 ppm 0.065 ppm

B ill io n s o f 2 0 0 6 $

PM Mortality ‐ Pope 2002 Ozone Mortality ‐ Bell 2004 PM Morbidity Ozone Morbidity Visibility

$10

$20

$30

$40

$50

$60

0.075 ppm 0.070 ppm 0.065 ppm

B ill io n s o f 2 0 0 6 $

PM Mortality ‐ Laden 2006 Ozone Mortality ‐ Levy PM Morbidity Ozone Morbidity Visibility

Figure S2.7: Total Monetized Benefits for Alternate Standard Levels*

*This graph shows the estimated total monetized benefits in 2020 using the no‐threshold model at discount rates of 3% using effect coefficients derived from the 6 ozone mortality studies and PM co‐benefits estimates using the Pope et al. study and the Laden et al study, as well as 12 effect coefficients derived from EPA’s expert elicitation on PM mortality. The highlighted results represent the combined estimates from Bell et al. (2004) with Pope et al. (2002) and Levy (2005) with Laden et al. (2006). The results shown are not the direct results from the studies or expert elicitation; rather, the estimates are based in part on the concentration‐response function provided in those studies. PM co‐benefit results using a 7% discount rate would be similar, but approximately 9% lower.

In 2008, the National Research Council (NRC) evaluated the EPA’s approach to estimating

ozone‐related mortality benefits. Among other recommendation, in its report the NRC indicated

that “EPA should consider placing greater emphasis on reporting decrease in age‐specific death

rates and increases in life expectancy…” (NRC, 2008). As a first step in implementing this

recommendation, below for two of the three scenarios, we present changes in the percentage of

total cause‐specific mortality attributable to ozone and the change in the number of life years.10

Table 7 summarizes the estimated number of life years gained resulting from simulated

attainment with the 0.065 ppm and 0.070 ppm standard alternatives. To simplify this presentation

10 Here we omit the results for the 0.075 ppm alternative. We estimated the benefits of attaining this alternative through an interpolation approach that made subsequent estimation of life years and changes in death rates technically challenging.

Pope et al., Bell et al. 2004

Laden et al., Levy et al.

$‐

$10

$20

$30

$40

$50

$60

$70

B ill io n s o f 2 0 0 6 $

Combinations of 6 Ozone benefits estimates with 14 PM2.5 co‐benefits estimates

0.075 ppm 0.070 ppm 0.065 ppm

we include results based on the estimates of ozone mortality reported in Levy et al. (2005) and Bell

et al. (2004), which provide upper and lower‐bound estimates, respectively.

Table S2.7: Estimated Reduction in Ozone‐Related Premature Mortality in Terms of Life Years Gained from Increases in Life Expectancy

Age Range Bell et al. (2004) mortality estimate Levy et al. (2005) mortality estimate

0.070 ppm 0.065 ppm 0.070 ppm 0.065 ppm

25‐29 75

(32—120) 130

(58—210) 660

(780—830) 1,200

(850—1,500)

30‐34 66

(28—100) 120

(51—180) 580

(420—740) 1,000

(750—1,300)

35‐44 260

(110—410) 460

(200—730) 1,600

(1,200—2,000) 2,800

(2,000—3,500)

45‐54 520

(220—830) 930

(400—1,500) 2,600

(1,900—3,300) 4,500

(3,300—5,700)

55‐64 1,000

(440—1,600) 1,800

(780—2,800) 4,600

(3,400—5,900) 8,100

(5,900—10,000)

65‐74 1,200

(500—1,900) 2,100

(900—3,300) 5,200

(3,800—6,600) 9,100

(6,700—12,000)

75‐84 810

(340—1,300) 1,400

(620—2,200) 3,500

(2,600—4,500) 6,200

(4,600—7,900)

85‐99 400

(170—630) 720

(310—1,100) 1,800

(1,300—2,200) 3,100

(2,300—4,000)

Table S2.8 summarizes the percentage of total mortality attributable to ozone. As above,

we include estimates based on the Bell et al. (2004) and Levy et al. (2005) risk coefficients.

Table S2.8: Percentage of Total Mortality Attributable to Ozone

Age Range

Bell et al. (2004) mortality estimate Levy et al. (2005) mortality estimate

0.070 ppm 0.065 ppm 0.070 ppm 0.065 ppm

25‐29 0.030% 0.054% 0.126% 0.224% 30‐34 0.029% 0.052% 0.123% 0.217% 35‐44 0.029% 0.051% 0.123% 0.217% 45‐54 0.030% 0.052% 0.127% 0.224% 55‐64 0.028% 0.050% 0.122% 0.212% 65‐74 0.027% 0.047% 0.114% 0.200% 75‐84 0.026% 0.046% 0.112% 0.197% 85‐99 0.027% 0.048% 0.115% 0.206%

Based on our review of the current body of scientific literature, EPA estimated PM-related

mortality without applying an assumed concentration threshold. EPA’s Integrated Science

Assessment for Particulate Matter (U.S. EPA, 2009c), which was recently reviewed by EPA’s Clean

Air Scientific Advisory Committee, concluded that the scientific literature consistently finds that a

no-threshold log-linear model most adequately portrays the PM-mortality concentration-response

relationship while recognizing potential uncertainty about the exact shape of the concentration-

response function. Consistent with this finding, we have conformed the threshold sensitivity

analysis to the current state of the PM science improved upon our previous approach for estimating

the sensitivity of the benefits estimates to the presence of an assumed threshold by incorporating a

new “Lowest Measured Level” (LML) assessment.

This approach summarizes the distribution of avoided PM mortality impacts according to

the baseline PM2.5 levels (i.e. those levels that exist prior to the implementation of the ozone

attainment scenario) experienced by the population receiving the PM2.5 mortality benefit (Figure

S2.8 and S2.9). We identify on this figure the lowest air quality levels measured in each of the two

primary epidemiological studies EPA uses to quantify PM-related mortality. This information

allows readers to determine the portion of PM-related mortality benefits occurring above or below

the LML of each study; in general, our confidence in the estimated PM mortality decreases as we

consider air quality levels further below the LML in the two epidemiological studies. While the

LML analysis provides some insight into the level of uncertainty in the estimated PM mortality

benefits, EPA does not view the LML as a threshold and continues to quantify PM-related mortality

impacts using a full range of modeled air quality concentrations.

The very large proportion of the avoided PM-related impacts we estimate in this illustrative

analysis occur among populations exposed at or above the LML of each study (Figures S2.8 and

S2.9), increasing our confidence in the PM mortality analysis. Approximately 62% of the avoided

impacts occur at or above an annual mean PM2.5 level of 10 µg/m3 (the LML of the Laden et al.

2006 study); about 97% occur at or above an annual mean PM2.5 level of 7.5 µg/m3 (the LML of the

Pope et al. 2002 study). As we model mortality impacts among populations exposed to levels of

PM2.5 that are successively lower than the LML of each study our confidence in the results

diminishes. However, the analysis above confirms that the great majority of the impacts occur at or

above each study’s LML.

Because time and resource limitations prevented EPA from performing air quality modeling

of the PM2.5-related co-benefits of the illustrative ozone attainment strategies, this LML analysis

considers only a single air quality modeling scenario. This single scenario represents only a portion

of PM2.5 reductions we anticipate to occur as a result of the NOx emission reductions needed to

attain a new standard of 0.065 ppm. As such, this LML analysis provides an incomplete

representation of the distribution of avoided mortality impacts and reductions in PM2.5 exposure that

might occur under a air quality modeling scenario that simulated full attainment with the 0.065 ppm

standard.

Finally, Figure S2.10 illustrates the percentage of population exposed to different levels of

annual mean PM2.5 levels in the baseline and after the implementation of the illustrative ozone

attainment strategy in 2020. This strategy achieves fairly modest reductions of PM2.5 as a co-benefit

of the ozone attainment strategy. Much of this small benefit occurs among highly exposed

populations and we find that prior to the implementation of this illustrative scenario, 83% of the

population live in areas where PM2.5 levels are projected to be above the lowest measured levels of

the Pope study. Taken together, this information increases our confidence in the estimated mortality

reductions for this rule.

While the LML of each study is important to consider when characterizing and interpreting

the overall level PM-related benefits, as discussed earlier in this chapter, EPA believes that both

cohort-based mortality estimates are suitable for use in air pollution health impact analyses. When

estimating PM mortality impacts using risk coefficients drawn from the Laden et al. analysis of the

Harvard Six Cities and the Pope et al. analysis of the American Cancer Society cohorts there are

innumerable other attributes that may affect the size of the reported risk estimates—including

differences in population demographics, the size of the cohort, activity patterns and particle

composition among others. The LML assessment presented here provides a limited representation

of one key difference between the two studies.

Figure S2.8: Percentage of PM‐related mortalities avoided by baseline PM2.5 air quality level

Figure S2.9: Cumulative percentage of total PM‐related mortalities avoided by baseline PM2.5 air quality

level

S2.10: Cumulative distribution of adult population at annual mean PM2.5 levels (pre‐ and post‐ policy

scenario)

S2.4 Comparison of results to previous results in 2008 Ozone NAAQS RIA

The overall effect of incorporating the array of methodological changes was to increase the

estimated benefits of attaining alternate ozone standards estimates presented in the 2008 Ozone NAAQS RIA. In general, the key update that had the largest effect on the valuation and the incidence results is removing the threshold from the PM concentration-response functions. Tables S2.9 and S2.10 show the total monetized benefits, costs, and net benefits for the 2008 Ozone RIA analysis and this updated analysis, respectively. Figure 6 shows a comparison of the range of net benefits estimates in this updated analysis compared to the net benefits presented in the 2008 Ozone NAAQS RIA.11

11 Net benefits are total monetized benefits minus total monetized costs. Total monetized benefits include ozone health benefits, PM2.5 health co-benefits, visibility benefits, but not other unquantified benefit categories.

Table S2.9: Total Monetized Costs with Ozone Benefits and PM2.5 Co-Benefits in 2020 (in Billions of 2006$) A 2008 RIA

Ozone Mortality Function

Reference Total Benefits B Total Costs C Net Benefits 3% 7% 7% 3% 7%

0. 07

5 pp

m NMMAPS and

Multi-city

Bell et al. 2004 $4.4 to $8.5 $4.1 to $7.7 $7.6 to $8.8 $-4.4 to $0.9 $-4.7 to $0.1 Schwartz 2005 N/A N/A N/A N/A N/A Huang 2005 N/A N/A N/A N/A N/A

Meta-analysis Bell et al. 2005 $5.6 to $9.7 $5.3 to $9.0 $7.6 to $8.8 $-3.2 to $2.1 $-3.5 to $1.4 Ito et al. 2005 $6.3 to $10 $5.9 to $9.6 $7.6 to $8.8 $-2.5 to $2.7 $-2.9 to $2.0 Levy et al. 2005 $6.3 to $10 $6.0 to $9.7 $7.6 to $8.8 $-2.5 to $2.8 $-2.8 to $2.1

0. 07

0 pp

m NMMAPS and

multi-city

Bell et al. 2004 $8.8 to $16 $8.2 to $15 $19 to $25 $-16 to $-2.8 $-17 to $4.1 Schwartz 2005 N/A N/A N/A N/A N/A Huang 2005 N/A N/A N/A N/A N/A

Meta-analysis Bell et al. 2005 $13 to $21 $13 to $19 $19 to $25 $-12 to $1.5 $-12 to $0.2 Ito et al. 2005 $15 to $23 $15 to $21 $19 to $25 $-9.6 to $3.8 $-10 to $2.5 Levy et al. 2005 $16 to $23 $15 to $22 $19 to $25 $-9.3 to 4.1 $9.9 to $2.7

0. 06

5 pp

m NMMAPS and

multi-city

Bell et al. 2004 $15 to $27 $14 to $24 $32 to $44 $-29 to $-5.4 $-30 to $-7.5 Schwartz 2005 N/A N/A N/A N/A N/A Huang 2005 N/A N/A N/A N/A N/A

Meta-analysis Bell et al. 2005 $22 to $34 $21 to $32 $32 to $44 $-22 to $2.4 $-23 to $0.3 Ito et al. 2005 $27 to $39 $26 to $36 $32 to $44 $-17 to $6.6 $-18 to $4.4 Levy et al. 2005 $27 to $39 $26 to $37 $32 to $44 $-17 to $7.0 $-18 to $4.9

A All estimates rounded to two significant figures. As such, they may not sum across columns. Only includes areas required to meet the current standard by 2020; does not include San Joaquin and South Coast areas in California.

B Includes ozone benefits, and PM2.5 co-benefits. Range was developed by adding the estimate from the ozone premature mortality function to estimates from the PM2.5 premature mortality functions from Pope et al. and Laden et al. Tables exclude unquantified and nonmonetized benefits.

Table S2.10: Total Monetized Costs with Ozone Benefits and PM2.5 Co-Benefits in 2020

(in Billions of 2006$) A Updated Analysis Ozone Mortality

Function Reference Total Benefits B Total Costs C Net Benefits

3% 7% 7% 3% 7%

0. 07

5 pp

m NMMAPS

and multi-city

Bell et al. 2004 $6.9 to $15 $6.4 to $13 $7.6 to $8.8 $-1.9 to $7.4 $-2.4 to $5.4

Schwartz 2005 $7.2 to $16 $6.8 to $13 $7.6 to $8.8 $-1.6 to $8.4 $-2.1 to $5.4

Huang 2005 $7.3 to $16 $6.9 to $13 $7.6 to $8.8 $-1.5 to $8.4 $-2.0 to $5.4

Meta-analysis

Bell et al. 2005 $8.3 to $17 $7.9 to $14 $7.6 to $8.8 $-0.50 to $9.4 $-1.0 to $6.4

Ito et al. 2005 $9.1 to $18 $8.7 to $15 $7.6 to $8.8 $0.30 to $10 $-0.20 to $7.4

Levy et al. 2005 $9.2 to $18 $8.8 to $15 $7.6 to $8.8 $0.40 to $10 $-0.10 to $7.4

0. 07

0 pp

NMMAPS and multi-city

Bell et al. 2004 $13 to $29 $11 to $24 $19 to $25 $-12 to $10 $-14 to $5.0

Schwartz 2005 $15 to $30 $12 to $25 $19 to $25 $-10 to $11 $-13 to $6.0

Huang 2005 $15 to $30 $13 to $26 $19 to $25 $-10 to $11 $-12 to $7.0

Meta-analysis

Bell et al. 2005 $18 to $34 $16 to $29 $19 to $25 $-7.0 to $15 $-9.0 to $10

Ito et al. 2005 $21 to $37 $18 to $31 $19 to $25 $-4.0 to $18 $-6.0 to $12

Levy et al. 2005 $21 to $37 $18 to $31 $19 to $25 $-4.0 to $18 $-6.0 to $12

0. 06

5 pp

NMMAPS and multi-city

Bell et al. 2004 $22 to $47 $19 to $40 $32 to $44 $-22 to $15 $-25 to $7.0

Schwartz 2005 $24 to $49 $21 to $42 $32 to $44 $-20 to $17 $-23 to $9.0

Huang 2005 $25 to $50 $22 to $42 $32 to $44 $-19 to $18 $-23 to $10

Meta-analysis

Bell et al. 2005 $31 to $56 $27 to $48 $32 to $44 $-13 to $24 $-17 to $16

Ito et al. 2005 $36 to $61 $32 to $53 $32 to $44 $-8.0 to $29 $-13 to $20

Levy et al. 2005 $36 to $61 $32 to $53 $32 to $44 $-7.0 to $29 $-12 to $20 A All estimates rounded to two significant figures. As such, they may not sum across columns. Only includes areas

required to meet the current standard by 2020; does not include San Joaquin and South Coast areas in California. B Includes ozone benefits, and PM2.5 co-benefits. Range was developed by adding the estimate from the ozone

premature mortality function to estimates from the PM2.5 premature mortality functions from Pope et al. and Laden et al. Tables exclude unquantified and nonmonetized benefits.

0.075 ppm 0.070 ppm 0.065 ppm

Figure S2.11: Comparison of Net Benefits in Updated Analysis to 2008 Ozone NAAQS RIA* 2008 RIA Updated Analysis

$(100)

$(80)

$(60)

$(40)

$(20)

$‐

$20

$40

$60

$80

$100

Median = $0.9b

Benefits are greater than costs

Costs are greater than benefits

‐$100

‐$80

‐$60

‐$40

‐$20

$20

$40

$60

$80

$100

Benefits are greater than costs

Costs are greater than benefits

Median = $3.1b

$(100)

$(80)

$(60)

$(40)

$(20)

$‐

$20

$40

$60

$80

$100

Benefits are greater than costs

Costs are greater than benefits

Median = ‐ $4.0b

$(100)

$(80)

$(60)

$(40)

$(20)

$‐

$20

$40

$60

$80

$100

Benefits are greater than costs

Costs are greater than benefits

Median = $1.4b

$(100)

$(80)

$(60)

$(40)

$(20)

$‐

$20

$40

$60

$80

$100

Costs are greater than benefits

Benefits are greater than costs

Median = ‐ $9.0b

$(100)

$(80)

$(60)

$(40)

$(20)

$‐

$20

$40

$60

$80

$100 Benefits are greater than costs

Costs are greater than benefits

Median = $0.7b

These graphs shows all combinations of the 6 different ozone mortality functions and assumptions, the 14 different PM mortality functions, and the 2 cost methods. These combinations do not represent a distribution.

S2.5 References

Bell, M.L., et al. 2004. Ozone and short-term mortality in 95 US urban communities, 1987-2000.

Journal of the American Medical Association. 292(19): p. 2372-8. Bell, M.L., F. Dominici, and J.M. Samet. 2005. A meta-analysis of time-series studies of ozone and

mortality with comparison to the national morbidity, mortality, and air pollution study. Epidemiology. 16(4): p. 436-45.

Huang Y. Dominici F. Bell M. 2005. Bayesian Hierarchical Distributed Lag Models for Summer Ozone Exposure and Cardio-Respiratory Mortality Environmetrics, 16, 547-562.

Ito, K., S.F. De Leon, and M. Lippmann. 2005. Associations between ozone and daily mortality:

analysis and meta-analysis. Epidemiology. 16(4): p. 446-57. Jaffe DH, Singer ME, Rimm AA. 2003. Air pollution and emergency department visits for asthma

among Ohio Medicaid recipients, 1991-1996. Environ Res 91(1):21-28. Jerrett M, Burnett RT, Pope CA, III, et al. 2009. Long-Term Ozone Exposure and Mortality. N

Engl J Med 360:1085-95. Krewski, D.; Jerrett, M.; Burnett, R.T.; Ma, R.; Hughes, E.; Shi, Y.; Turner, M.C.; Pope, C.A. III;

Thurston, G.; Calle, E.E.; Thun, M.J. 2009. Extended follow-up and spatial analysis of the American Cancer Society study linking particulate air pollution and mortality. HEI Research Report, 140, Health Effects Institute, Boston, MA.

Laden, F., J. Schwartz, F.E. Speizer, and D.W. Dockery. 2006. Reduction in Fine Particulate Air

Pollution and Mortality. American Journal of Respiratory and Critical Care Medicine 173:667- 672.

Levy, J.I., S.M. Chemerynski, and J.A. Sarnat. 2005. Ozone exposure and mortality: an empiric

bayes metaregression analysis. Epidemiology. 16(4): p. 458-68. National Research Council (NRC). 2008. Estimating Mortality Risk Reduction and Economic

Benefits from Controlling Ozone Air Pollution. National Academies Press. Washington, DC. Peel, J. L., P. E. Tolbert, M. Klein, et al. 2005. Ambient air pollution and respiratory emergency

department visits. Epidemiology. Vol. 16 (2): 164-74. Pope, C.A., III, R.T. Burnett, M.J. Thun, E.E. Calle, D. Krewski, K. Ito, and G.D. Thurston. 2002.

“Lung Cancer, Cardiopulmonary Mortality, and Long-term Exposure to Fine Particulate Air Pollution.” Journal of the American Medical Association 287:1132-1141.

Roman, Henry A., Katherine D. Walker, Tyra L. Walsh, Lisa Conner, Harvey M. Richmond, Bryan

J. Hubbell, and Patrick L. Kinney. 2008. Expert Judgment Assessment of the Mortality Impact of Changes

Schwartz, J. 2005. How sensitive is the association between ozone and daily deaths to control for temperature? Am J Respir Crit Care Med. Vol. 171 (6): 627-31.

U.S. Environmental Protection Agency (U.S. EPA). 2008. Regulatory Impact Analysis, 2008

National Ambient Air Quality Standards for Ground-level Ozone, Chapter 6. Office of Air Quality Planning and Standards, Research Triangle Park, NC. March. Available on the Internet at <http://www.epa.gov/ttn/ecas/regdata/RIAs/6-ozoneriachapter6.pdf>.

U.S. Environmental Protection Agency (U.S. EPA). 2009a. Regulatory Impact Analysis: National

Emission Standards for Hazardous Air Pollutants from the Portland Cement Manufacturing Industry. Office of Air Quality Planning and Standards, Research Triangle Park, NC. April. Available on the Internet at <http://www.epa.gov/ttn/ecas/regdata/RIAs/portlandcementria_4- 20-09.pdf>.

U.S. Environmental Protection Agency (U.S. EPA). 2009b. Proposed NO2 NAAQS Regulatory

Impact Analysis (RIA). Office of Air Quality Planning and Standards, Research Triangle Park, NC. July. Available on the Internet at <http://www.epa.gov/ttn/ecas/regdata/RIAs/proposedno2ria.pdf>.

U.S. Environmental Protection Agency (U.S. EPA). 2009c. Integrated Science Assessment for

Particulate Matter (Final Report). EPA-600-R-08-139F. National Center for Environmental Assessment – RTP Division. December. Available on the Internet at <http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=216546>.

Wilson, A. M., C. P. Wake, T. Kelly, et al. 2005. Air pollution, weather, and respiratory emergency room visits in two northern New England cities: an ecological time-series study. Environ Res. Vol. 97 (3): 312-21.

SECTION 3: SECONDARY OZONE NAAQS EVALUATION

1.1

1.2 Synopsis

This section contains an evaluation of the regulatory impacts associated with a distinct

secondary NAAQS for ozone. The purpose of a secondary NAAQS is to protect the public welfare

against the negative effects of criteria air pollutants, including decreased visibility, damage to

animals, crops, vegetation, and buildings. Exposure to ozone has been associated with a wide

array of vegetation and ecosystem effects, including those that damage or impair the intended use

of the plant or ecosystem. Such effects are considered adverse to the public welfare. This

secondary NAAQS standard for ozone is the first secondary standard to be promulgated with a

form, averaging time, and level that is distinct from the health‐based primary standard, apart from

the PM and SO2 regulations originally set in the early 1970s. Quantifying the costs and benefits of

attaining a secondary NAAQS is an exceptionally complex task, including unresolved issues related

to the RIA analysis, air quality projections, monitoring expansion, and implementation.12 Because

of these complexities as well as limited time and resources within the expedited schedule, we are

limited in our ability to quantify the costs and benefits of attaining a distinct secondary NAAQS for

ozone for this rule. However, we provide a semi‐quantitative assessment in this analysis, including

identifying which counties would have an additional requirement to reduce ozone concentrations

to attain a secondary standard beyond the reductions needed to attain the primary standard,

qualitative descriptions of available pollution control strategies, qualitative benefits of reducing

ozone exposure on forests, crops, and ornamental plants, and maps of avoided biomass/yield loss

for the currently monitor locations. The Administrator selected a secondary ozone NAAQS at a

level of 13 ppm‐hrs using the W126 form. Using a cumulative seasonal secondary standard (i.e.,

W126), we evaluated alternate standard levels at 11, 13, and 15 ppm‐hours.

S2.6 Introduction

As defined by section 109(b)(2) of the Clean Air Act (CAA), the purpose of a secondary

NAAQS is to protect the public welfare against any known or anticipated negative effects

associated with criteria air pollutants. These welfare effects include, but are not limited to,

‘‘effects on soils, water, crops, vegetation, man‐made materials, animals, wildlife, weather,

visibility, and climate, damage to and deterioration of property, and hazards to transportation, as

well as effects on economic values and on personal comfort and wellbeing.’’

The secondary NAAQS for ozone is focused on the negative effects on vegetation

associated with direct ozone exposure. Exposure to ozone has been associated with a wide array

12 These complexities are described in detail in Section S3.3.

of vegetation and ecosystem effects in the published literature (U.S. EPA, 2006). Sensitivity to

ozone is highly variable across plant species, with over 65 plant species identified as “ozone‐

sensitive”, many of which occur in state and national parks and forests. 13 These effects include

those that damage or impair the intended use of the plant or ecosystem. Such effects are

considered adverse to the public welfare and can include reduced growth and/or biomass

production in sensitive plant species, including forest trees, reduced crop yields, visible foliar

injury, reduced plant vigor (e.g., increased susceptibility to harsh weather, disease, insect pest

infestation, and competition), species composition shift, and changes in ecosystems and

associated ecosystem services.

Vegetation effects research has shown that seasonal air quality indices that cumulate peak‐

weighted hourly ozone concentrations are the best candidates for relating exposure to plant

growth effects (U.S. EPA, 2006). Based on this research, the 2007 Ozone Staff Paper (hereafter,

“the Staff Paper”) concluded that the cumulative, seasonal index referred to as “W126” is the

most appropriate index for relating vegetation response to ambient ozone exposures (U.S. EPA,

2007b). Based on additional conclusions regarding appropriate diurnal and seasonal exposure

windows, the Staff Paper recommended a cumulative seasonal secondary standard, expressed as

an index of the annual sum of weighted hourly concentrations (using the W126 form), set at a

level in the range of 7 to 21 ppm‐hours. The index would be cumulated over the 12‐hour daylight

window (8:00 a.m. to 8:00 p.m.) during the consecutive 3‐month period during the ozone season

with the maximum index value (hereafter, referred to as W126). After reviewing the

recommendations in the Staff Paper, EPA’s Clean Air Scientific Advisory committee (CASAC) agreed

with the form of the secondary standard, but instead recommended a range of 7 to 15 ppm‐hours

(U.S. EPA‐SAB, 2007). In January 2010, EPA’s Administrator proposed a range of secondary

standards based on the W126 index between 7 and 15 ppm‐hrs (U.S. EPA, 2010). After reviewing

the scientific evidence and public comments, the Administrator selected a secondary ozone

NAAQS at a level of 13 ppm‐hrs, using the W126 form, calculated as a 3‐year average of annual

sums.

To comply with Circular A‐4 (OMB, 2003), this analysis includes the selected standard level

as well as one more stringent and one less stringent alternative. Therefore, this analysis focuses

on secondary standards at 13 ppm‐hrs, as well as 15 ppm‐hrs and 11 ppm‐hrs.

S2.7 Air Quality Analysis

Ozone is a secondary pollutant formed by atmospheric reactions involving two classes of

precursor compounds: nitrogen oxides (NOx) and volatile organic compounds (VOCs) (U.S. EPA,

13 Appendix S3A contains a list of plant species identified as “ozone‐sensitive”.

2007b). The W126 standard is a specific peak‐weighted index that is summed over 12 hours per

day during the maximum 3‐month period within the ozone season and calculated as the 3‐year

average of the annual sums. An example of this calculation is described in more detail in Appendix

S3‐B of this RIA. The 3‐year average provides increased stability due to large year‐to‐year

variability. As described in the Staff Paper, using the highest PRB estimate of 0.035 ppm from

Fiore et al. (2003) as a constant value would only add up to a 3‐month 12‐hr W126 of less than 1

ppm‐hr (U.S. EPA, 2007b).

a. Ambient Monitoring Data (2007 – 2009)

The monitoring data for this analysis has been updated since the proposal. In addition to

incorporating more recent monitoring data, we have also excluded monitoring data from CASTNET

that cannot be used for nonattainment designations. Ozone concentrations were generally lower

in 2009, and thus the 2007‐2009 design values indicate fewer counties would violate the

secondary standard compared to the counties shown in the proposal analysis. These monitoring

data are limited to the existing monitoring network. It is important to note that nonattainment

designations are likely to be based on 2008‐2010 data, not 2007‐2009 data. 14

In this analysis, we considered the extent to which there is overlap between county‐level

air quality measured in terms of the 8‐hour average form of the current standard and that

measured in terms of the cumulative W126, seasonal form. Using monitoring data collected from

2007 to 2009, Table S3‐1 shows the number of counties that exceed the alternate secondary

standard levels in comparison to the number of counties that exceed the selected primary

standard at 0.070 ppm. Figure S3‐1 maps the counties that correspond with Table S3‐1.

Table S3‐1: Number of Counties Exceeding Alternate Secondary Standards

(2007‐2009 monitoring data)

Monitor Baseline 15 ppm‐hrs 13 ppm‐hrs 11 ppm‐hrs

Attain primary (0.070 ppm) and secondary 270 262 257

Exceed only primary (0.070 ppm) 335 268 194

Exceed primary (0.070 ppm) and secondary 85 152 226

Exceed only secondary 3 11 16

* As these estimates are limited to existing ozone monitoring data, there might be other non‐monitored areas after the monitoring network is expanded that would exceed the secondary standard. There are 693 currently monitored counties with sufficient data for this analysis.

14 Monitoring data for 2010 is not yet available.

Figure S3‐1: Counties exceeding Primary Standard at 0.070 ppm or Secondary Standard at 13 ppm‐hours (based on 2007–2009 monitoring data)

b. Modeling Projection Data (2020)

In this analysis, we also projected W126 levels for two scenarios in 2020 developed as part

of the 2008 analysis of the primary standard: the baseline scenario and the after hypothetical RIA

controls scenario. The modeling methodology used to project W126 levels into the future utilizes

the same approach as used to project design values of the primary standard, as described in EPA

modeling guidance (U.S. EPA, 2007a). The 2020 baseline and hypothetical RIA control scenario are

fully described in Chapter 3 of the 2008 Ozone NAAQS RIA (U.S. EPA, 2008a). The baseline

includes current state and federal programs plus additional controls EPA estimated would be

necessary to attain the previous ozone and PM2.5 standards. For the hypothetical RIA control

scenario, EPA applied additional known NOx and VOC controls in those specific geographic

areas that were predicted to exceed an 0.070 ppm primary standard in 2020.15

Additionally, EPA estimated the counties that are projected to attain the primary standard

in 2020 but would still exceed the alternate secondary standards. These data are listed in Table

S3‐2, and mapped in Figures S3‐2 through S3‐5. Because this projection approach is prefaced on

15 It is important to note that the modeled hypothetical RIA controls did not fully attain the primary standard of 0.070 ppm, especially in Southern California, Houston, Eastern Lake Michigan, and the Northeast corridor.

(262 counties)

(268 counties)

(152 counties)

(11 counties)

ambient data, projections can only be made for counties with ozone monitoring data for the base

period. As a result, Table S3‐2 and the associated figures may not capture other, currently

unmonitored, locations.

Table S3‐2: Number of Counties Projected to Exceed Alternate Secondary Standards in 2020*

2020 Baseline 15 ppm‐hrs 13 ppm‐hrs 11 ppm‐hrs

Attain primary (0.070 ppm) and secondary 599 591 580

Exceed only primary (0.070 ppm) 79 70 55

Exceed primary (0.070 ppm) and secondary 20 29 44

Exceed only secondary 7 15 26

After Hypothetical RIA controls 15 ppm‐hrs 13 ppm‐hrs 11 ppm‐hrs

Attain primary (0.070 ppm) and secondary 633 624 613

Exceed only primary (0.070 ppm) 48 41 36

Exceed primary (0.070 ppm) and secondary 17 24 29

Exceed only secondary 7 16 27

* As these projections are limited to counties with existing ozone monitoring data, there might be other non‐ monitored areas that would exceed the secondary standard while attaining the primary standard. There are 705 currently monitored counties with sufficient data for this analysis. It is important to note that the modeled hypothetical RIA controls did not fully attain the primary standard of 0.070 ppm, especially in Southern California, Houston, Eastern Lake Michigan, and the Northeast corridor. The number of counties that exceed only the secondary standard increase after the hypothetical RIA controls because those counties now attain the primary standard.

Figure S3‐2: Projected W126 Levels in the Baseline in 2020*

* Many of the counties projected to exceed the alternate secondary standard levels are in the South Coast and San Joaquin areas of California, which are not required to attain the primary standards by 2020.

(635 counties)

(26 counties)

(17 counties)

(27 counties)

Figure S3‐3: Change in Projected W126 Levels from the Hypothetical RIA controls in 2020*

*All of the counties projected to experience minor or moderate worsening due to the hypothetical RIA controls in 2020 are located in areas well below the alternate secondary standard levels. Because the hypothetical RIA controls were designed to reduce ozone concentrations in areas that exceeded the primary standard, those areas are also projected to experience minor to major improvements in W126 levels in 2020. It is important to note that the modeled hypothetical RIA controls did not fully attain the primary standard of 0.070 ppm, especially Southern California, Houston, Eastern Lake Michigan, and the Northeast corridor.

Figure S3‐4: Counties Projected to Exceed the Selected Primary and Secondary Standards in the Baseline in 2020*

Figure S3‐5: Counties Projected to Exceed the Selected Primary and Secondary Standards after Hypothetical RIA Controls in 2020*

(591 counties)

(70 counties)

(29 counties)

(15 counties)

(624counties)

(41 counties)

(24counties)

(16 counties)

* Many of the counties projected to exceed the secondary standard are in the South Coast and San Joaquin areas of California, which are not required to attain the primary standards by 2020. The number of counties that exceed only the secondary standard increase after the hypothetical RIA controls because those counties now attain the primary standard.

As noted above, this analysis only projected W126 levels in 2020 where ozone monitors

currently exist. Due to the lack of more complete monitor coverage in many rural areas, this

analysis might not be an accurate reflection of ozone concentrations in non‐monitored, rural

counties where sensitive vegetation, important ecosystems, or other areas of national public

interest could be located. Many counties that contain high elevation, rural or remote sites tend to

have flatter ozone concentration distributions. These areas may not reflect the typical urban and

near‐urban pattern of low morning and evening ozone concentrations with a high mid‐day peak,

but instead maintain relatively flat patterns with many concentrations in the mid‐range (e.g., 0.05‐

0.09 ppm) for extended periods. Therefore, the potential for disconnect between 8‐hour average

and cumulative, seasonal form is greater. Additional rural, high elevation areas important for

vegetation that are not currently monitored would likely experience similar ozone exposure

patterns (U.S. EPA, 2007b). This is an important caveat because: (1) the biological database

stresses the importance of cumulative, seasonal exposures in determining plant response; (2)

plants have not been specifically tested for the importance of daily maximum 8‐hour ozone

concentrations in relation to plant response; and (3) the effects of attainment of a 8‐hour standard

in upwind urban areas on rural air quality distributions cannot be characterized with confidence

due to the lack of monitoring data in rural and remote areas (U.S. EPA, 2007b).

Thus far, we have not expressly considered the question of whether it would be more

difficult to attain the secondary standard than the primary or what levels of controls would be

required to attain the secondary standard. Based on the existing air quality modeling from the

2008 Ozone NAAQS RIA, we have examined how W126 values might change in response to the

hypothetical RIA control strategy designed to attain the primary standard. Based on projected

W126 ozone levels before and after the implementation of the hypothetical RIA control strategy in

2020, there is some evidence that it may indeed be harder to attain the secondary standard in

some areas. As an example, the hypothetical RIA control scenario reduces the number of counties

exceeding a primary NAAQS of 0.070 ppm by about 34%; whereas the same control

scenario reduces the number of counties exceeding a secondary NAAQS of 13 ppm‐hours by only

9%.

The air quality modeling for the 2008 RIA focused on quantifying the impacts and costs of

attaining the primary standard. Because the form of the secondary standard is calculated by

summing the daily ozone concentrations over a three‐month period, it is possible that mitigation

strategies may be different for a secondary ozone standard than for the primary ozone standard.

Initial ambient data analyses and future‐year modeling suggest that it may be more difficult to

attain the secondary standard in the western U.S. than in the eastern U.S for several

reasons. First, ozone concentrations have less variability across days in the western U.S. Second,

the meteorological parameters that generally result in lower daily ozone peaks (e.g., clouds,

precipitation, frontal passages) occur less frequently in the western States. Lastly, the secondary

standard may have larger implications for rural areas currently without monitors as opposed to

the urban areas where the primary ozone standard is already a concern. Attainment of the

secondary standard may involve more regional and national scale controls than the current local

efforts to reduce peak concentrations.

S2.8 Complexities in Quantifying the Costs and Benefits of Attaining a Secondary Ozone

NAAQS

Despite recent proposals, EPA has not promulgated a secondary NAAQS with a form,

averaging time, and level that is distinct from the health‐based primary standard, apart from the

secondary NAAQS for PM and SO2 originally set in the early 1970s. Therefore, prior to this rule,

EPA has not conducted a regulatory analysis of a secondary NAAQS. Quantifying the costs and

benefits associated with attaining a distinct secondary standard is an exceptionally complex task.

We describe these complexities in detail below.

Because of these complexities as well as limited time, resources, and available data within

the expedited schedule, we are limited in our ability to quantify the costs and benefits of attaining

a distinct secondary NAAQS for ozone. However, we recognize that the regulatory impacts

associated with this standard are of interest to many. Therefore, we provide a semi‐quantitative

assessment in this analysis, including identifying which counties would have an additional

requirement to reduce ozone concentrations to attain a secondary standard beyond the

reductions needed to attain the primary standard, qualitative descriptions of available pollution

control strategies, qualitative benefits of reducing ozone exposure on forests, crops, and

ornamental plants, and maps of avoided biomass/yield loss for the currently monitor locations.

S4.3.1 RIA complexities

There are two unresolved RIA issues that complicate a fully quantitative analysis of a

secondary standard for ozone. First, it is unclear when an area would need to attain a secondary

standard, which makes it difficult to choose an appropriate analysis year for the RIA. Whereas

attainment dates for the primary NAAQS are explicitly designated in the CAA, the attainment dates

for the secondary NAAQS are required “as expeditiously as practicable” after the nonattainment

designation (42 USC §7502(a)(2)). As air quality improves over time from regulations already

promulgated, an area would not need as many emission reductions for a later analysis year as the

area would need for an earlier analysis year. Assuming an analysis year of 2020 as was assumed

for the primary standard would substantially overestimate the costs and benefits associated with

attaining the secondary standard. Even if we determined that it was most appropriate to choose

an analysis year of 2030, 2040, or even 2050, we are limited to the available modeling data for

2020. Therefore, the choice of an analysis year has a significant effect on the magnitude of the

costs and benefits of attaining a secondary standard.

Second, it is unclear whether it is appropriate to include emission reductions that occur as

a result of implementing the primary standard in the baseline for the analysis of the secondary

standard. This is a critical decision, as it would either improperly ascribe the costs and benefits of

the primary NAAQS to the secondary NAAQS or it would violate the requirements of OMB’s

Circular A‐4 to only include promulgated rules in the regulatory baseline. Most of the areas that

exceed the secondary standard also exceed the primary standard. As shown in Table S3‐2, the

hypothetical RIA controls designed to attain the primary standard also reduce the number of

counties that exceed the secondary standard. Furthermore, it is likely that full attainment of the

primary standard in areas like Southern California or Eastern Lake Michigan would further reduce

the number of counties that exceed the secondary standard.

S4.3.2 Air quality data complexities

In addition to unresolved RIA issues, we have limited information available from the

available air quality modeling data to inform a secondary standard analysis. As shown in Table S3‐

2, several counties are projected to not to attain the alternate secondary standard levels in 2020

even after applying controls for the hypothetical RIA control scenario. Estimating the amount of

additional reductions (extrapolated tons) needed to attain a secondary standard would require a

better understanding of the relationship between emissions reductions and the W126 metric. Our

long experience with the primary standard allows us to use simple impact ratios with some

confidence in the extrapolated cost analysis for the primary standard. At present, it is not possible

to reproduce a similar analysis for the secondary standard. Without the amount of emission

reductions required to attain, it is not possible to identify the pollution control measures or the

associated costs.

S4.3.3 Monitoring complexities

As described in Section S3.2, the current monitoring network was not designed to

adequately reflect W126 levels in many areas of the country, especially the rural west. Therefore,

we cannot extrapolate the concentrations beyond the currently monitored counties, and we

cannot quantify the potential ozone vegetation impacts in many areas of high ecological value,

such as National Parks, wilderness areas, or other areas of sensitive national vegetation and

ecosystems. We note, however, that even if additional monitors were deployed, it may prove

challenging to completely characterize ozone concentrations in some locations that have not

traditionally been areas of focus for ozone network deployment.

S4.3.4 Implementation complexities

Other complexities related to implementation have yet to be resolved. For example, EPA

has not yet issued guidance for States to recommend boundaries of nonattainment areas for a

seasonal secondary ozone standard. The CAA requires that nonattainment areas include areas

that violate the standard as well as nearby areas that contribute to a violation. Based on modeled

projections of W126 levels in 2020, many of the areas that would exceed the secondary standard

without exceeding the primary standard are located in rural areas. Many of those areas lack

significant emission sources of ozone precursors within the area, so the cause of the violation is

likely due to longer‐range transport of ozone and precursors. Analyses of the origin of the

contributing emissions in such areas are unavailable. It is unclear what the appropriate

boundaries for these projected nonattainment areas would need to be such that the nearby

sources that are contributing to the violation are included but the contributing sources that are

not “nearby” are excluded. It is important to note that EPA intends to designate nonattainment

areas for the 2011 secondary NAAQS for ozone in 2013 based on the recent air quality monitoring

data at that time, not on the 2020 projected levels.

In addition, EPA is in the process of developing rules on how States should implement the

secondary ozone standard. One issue that must be addressed from a legal stand point is whether

planning for nonattainment areas must be done under the more prescriptive subpart 2

requirements of the CAA, which would require classification (as marginal, moderate, serious, etc)

or under the less prescriptive subpart 1 of the CAA. For areas classified under subpart 2, there are

certain specific control measures that States must adopt. The CAA language is unclear as to

whether subpart 2 applies to nonattainment areas under a secondary standard (although it

appears to be clear that the maximum statutory attainment dates in the classification table only

apply to the “primary” standard). Therefore, it is unclear whether it is appropriate to include the

subpart 2 mandatory measures in this analysis. The agency has never faced this issue in the past

for ozone, so this will be addressed in the upcoming rules. Since most, if not all, of the areas that

might be designated as nonattainment for the secondary standard would also be in nonattainment

for the primary standard, it is unclear whether States would need to adopt additional control

measures to attain the secondary standard.

S2.9 Pollution Control Strategies

The pollution control measures that might be adopted to attain the secondary standard

overlap substantially with the control measures used to attain the primary standard. The air

quality analysis showed that most areas that exceed the secondary standard would also exceed

the primary standard. If there are areas that would need additional emission reductions to attain

the secondary standard, we have included brief descriptions of some available NOx and VOC

controls below.

S3.4.1 Point Source Control Measures

For electrical generating units (EGUs), the primary measures for controlling NOx emissions

are selective catalytic reduction (SCR), selective noncatalytic reduction (SNCR), and low‐NOx

burners (LNB). SCR or SNCR can be applied along with a combustion control to further reduce NOx

emissions.

Several types of NOx control technologies exist for nonEGU point sources: SCR, SNCR,

natural gas reburn (NGR), coal reburn, and LNB. In some cases, LNB accompanied by flue gas

recirculation (FGR) is applicable, such as when fuel‐borne NOx emissions are expected to be of

greater importance than thermal NOx emissions. When circumstances suggest that combustion

controls do not make sense as a control technology (e.g., sintering processes, coke oven batteries,

sulfur recovery plants), SNCR or SCR may be an appropriate choice. Finally, SCR can be applied

along with a combustion control such as LNB with overfire air (OFA) to further reduce NOx

emissions. All of these control measures are available for application on industrial boilers and

other non‐EGU point sources.

Besides industrial boilers, other nonEGU point source categories that could install controls

include petroleum refineries, kraft pulp mills, cement kilns, stationary internal combustion

engines, glass manufacturing, combustion turbines, and incinerators. NOx control measures

available for petroleum refineries, particularly process heaters at these plants, include LNB, SNCR,

FGR, and SCR along with combinations of these technologies. NOx control measures available for

kraft pulp mills include those available to industrial boilers, namely LNB, SCR, SNCR, along with

water injection (WI). NOx control measures available for cement kilns include those available to

industrial boilers, namely LNB, SCR, and SNCR. Non‐selective catalytic reduction (NSCR) can be

used on stationary internal combustion engines. OXY‐firing, a technique to modify combustion at

glass manufacturing plants, can be used to reduce NOx at such plants. LNB, SCR, and SCR + steam

injection (SI) are available measures for combustion turbines. Finally, SNCR is an available control

technology at incinerators.

VOC controls include a variety of nonEGU point sources as defined in the emissions

inventory. The first control is permanent total enclosure (PTE) applied to paper and web coating

operations and fabric operations, and incinerators or thermal oxidizers applied to wood products

and marine surface coating operations. A PTE confines VOC emissions to a particular area where

can be destroyed or used in a way that limits emissions to the outside atmosphere, and an

incinerator or thermal oxidizer destroys VOC emissions through exposure to high temperatures

(2,000 degrees Fahrenheit or higher). The second control is petroleum and solvent evaporation

applied to printing and publishing sources as well as to surface coating operations.

S3.4.2 Area Source Control Measures

There are three control measures available for NOx emissions from area sources. The first

is RACT (reasonably available control technology) to 25 tpy (LNB). This control is the addition of a

low NOx burner to reduce NOx emissions. This control applies to industrial oil, natural gas, and

coal combustion sources. The second control is water heaters plus LNB space heaters. This control

is based on the installation of low‐NOx space heaters and water heaters in commercial and

institutional sources for the reduction of NOx emissions. The third control is switching to low sulfur

fuel for residential home heating. This control is primarily designed to reduce sulfur dioxide, but

has a co‐benefit of reducing NOx.

An available control to reduce VOC emissions from area sources is CARB Long‐Term Limits.

This control, which represents controls available in VOC rules promulgated by the California Air

Resources Board, applies to commercial solvents and commercial adhesives, and depends on

future technological innovation and market incentive methods to achieve emission reductions.

The next most frequently applied control was the use of low or no VOC materials for graphic art

source categories. The South Coast Air District’s SCAQMD Rule 1168 control applies to wood

furniture and solvent source categories sets limits for adhesive and sealant VOC content. The OTC

solvent cleaning rule control establishes hardware and operating requirements for specified vapor

cleaning machines, as well as solvent volatility limits and operating practices for cold cleaners. The

Low Pressure/Vacuum Relief Valve control measure is the addition of low pressure/vacuum (LP/V)

relief valves to gasoline storage tanks at service stations with Stage II control systems. LP/V relief

valves prevent breathing emissions from gasoline storage tank vent pipes. SCAQMD Limits control

establishes VOC content limits for metal coatings along with application procedures and

equipment requirements. Switching to Emulsified Asphalts control is a generic control measure

replacing VOC‐containing cutback asphalt with VOC‐free emulsified asphalt. The equipment and

maintenance control measure applies to oil and natural gas production. The Reformulation—FIP

Rule control measure intends to reach the VOC limits by switching to and/or encouraging the use

of low‐VOC pesticides and better Integrated Pest Management (IPM) practices.

S3.4.3 Mobile Source Control Measures

The NOx control measures available to onroad mobile sources include retrofits of diesel

engines, reduction of long duration heavy duty truck idling, continuous inspection and

maintenance programs and commuter programs. For nonroad sources, retrofits of diesel engines

and engine rebuilds are available. The VOC control measures available to onroad and nonroad

mobile sources include the listed controls for NOx plus reduction of Reid vapor pressure in

gasoline engines.

S3.4.4 Control Measures beyond the Identified Control Measures Database

Below is a list of controls beyond those in our identified control measures database that

are under development and not widely available as yet. There are major uncertainties associated

with each of these measures.

 Enhanced LDAR for Fugitive Leaks: This control measure is a more stringent program to

reduce leaks of fugitive VOC emissions from chemical plants and refineries that presumes

that an existing LDAR program already is in operation.

 Flare Gas Recovery: This control measure is a condenser that can recover 98 percent of the

VOC emitted by flares that emit 20 tons per year or more of the pollutant.

 Cooling Towers: This control measure is continuous monitoring of VOC from the cooling

water return to a level of 10 ppb. This monitoring is accomplished by using a continuous

flow monitor at the inlet to each cooling tower. There is not a general estimate of CE for

this measure; one is to apply a continuous flow monitor until VOC emissions have reached a

level of 1.7 tons/year for a given cooling tower.16

 Wastewater Drains and Separators: This control measure includes an inspection and

maintenance program to reduce VOC emissions from wastewater drains and water seals on

drains. This measure is a more stringent version of measures that underlie existing NESHAP

requirements for such sources.

 Work Practices or Use of Low VOC Coatings: The control measure is either application of

work practices (e.g., storing VOC‐containing cleaning materials in closed containers,

minimizing spills) or using coatings that have much lower VOC content. These measures,

which are of relatively low cost compared to other VOC area source controls, can apply to a

variety of processes, both for non‐EGU point and area sources, in different industries and is

defined in the proposed control techniques guidelines (CTG) for paper, film and foil

coatings, metal furniture coatings, and large appliance coatings published by the US EPA in

July 2007.17 The estimated CE expected to be achieved by either of these control measures is

90 percent.

16 Bay Area Air Quality Management District (BAAQMD). Proposed Revision of Regulation 8, Rule 8: Wastewater Collection Systems. Staff Report, March 17, 2004.

17 U.S. Environmental Protection Agency. Consumer and Commercial Products: Control Techniques Guidelines in Lieu of Regulations for Paper, Film, and Foil Coatings; Metal Furniture Coatings; and Large Appliance Coatings. 40 CFR 59. July 10, 2007. Available on the Internet at http://www.epa.gov/ttncaaa1/t1/fr_notices/ctg_ccp092807.pdf. It should be noted that this CTG became final in October 2007.

S2.10 Benefits of Reducing Ozone Effects on Vegetation and Ecosystems18

Air pollution can affect the environment and affect ecological systems, leading to changes in the

ecological community and influencing the diversity, health, and vigor of individual species (U.S. EPA, 2006).

Ozone causes discernible injury to a wide array of vegetation (U.S. EPA, 2006; Fox and Mickler,

1996). Sensitivity to ozone is highly variable across plant species, with over 65 plant species

identified as “ozone‐sensitive”, many of which occur in state and national parks and forests. 19 In

terms of forest productivity and ecosystem diversity, ozone may be the pollutant with the greatest

potential for regional‐scale forest impacts (U.S. EPA, 2006). Studies have demonstrated

repeatedly that ozone concentrations commonly observed in polluted areas can have substantial

impacts on plant function (De Steiguer et al., 1990; Pye, 1988).

When ozone is present in the air, it can enter the leaves of plants, where it can cause

significant cellular damage. Like carbon dioxide (CO2) and other gaseous substances, ozone enters

plant tissues primarily through the stomata in leaves in a process called “uptake” (Winner and

Atkinson, 1986). Once sufficient levels of ozone (a highly reactive substance), or its reaction

products, reaches the interior of plant cells, it can inhibit or damage essential cellular components

and functions, including enzyme activities, lipids, and cellular membranes, disrupting the plant's

osmotic (i.e., water) balance and energy utilization patterns (U.S. EPA, 2006; Tingey and Taylor,

1982). With fewer resources available, the plant reallocates existing resources away from root

growth and storage, above ground growth or yield, and reproductive processes, toward leaf repair

and maintenance, leading to reduced growth and/or reproduction. Studies have shown that

plants stressed in these ways may exhibit a general loss of vigor, which can lead to secondary

impacts that modify plants' responses to other environmental factors. Specifically, plants may

become more sensitive to other air pollutants, or more susceptible to disease, pest infestation,

harsh weather (e.g., drought, frost) and other environmental stresses, which can all produce a loss

in plant vigor in ozone‐sensitive species that over time may lead to premature plant death.

Furthermore, there is evidence that ozone can interfere with the formation of mycorrhiza,

essential symbiotic fungi associated with the roots of most terrestrial plants, by reducing the

amount of carbon available for transfer from the host to the symbiont (U.S. EPA, 2006).

This ozone damage may or may not be accompanied by visible injury on leaves, and

likewise, visible foliar injury may or may not be a symptom of the other types of plant damage

described above. Foliar injury is usually the first visible sign of injury to plants from ozone

exposure and indicates impaired physiological processes in the leaves (Grulke, 2003). When visible

18 It is important to note that these vegetation benefits are contingent upon the secondary standard being the controlling standard. In other words, if the primary standard is controlling in all areas, there would not be any additional vegetation benefits beyond those due to the primary standard.

19 Appendix S3A contains a list of plant species identified as “ozone‐sensitive”.

injury is present, it is commonly manifested as chlorotic or necrotic spots, and/or increased leaf

senescence (accelerated leaf aging). Visible foliar injury reduces the aesthetic value of ornamental

vegetation and trees in urban landscapes and negatively affects scenic vistas in protected natural

areas.

Ozone can produce both acute and chronic injury in sensitive species depending on the

concentration level and the duration of the exposure. Ozone effects also tend to accumulate over

the growing season of the plant, so that even lower concentrations experienced for a longer

duration have the potential to create chronic stress on sensitive vegetation. Not all plants,

however, are equally sensitive to ozone. Much of the variation in sensitivity between individual

plants or whole species is related to the plant’s ability to regulate the extent of gas exchange via

leaf stomata (e.g., avoidance of ozone uptake through closure of stomata) and the relative ability

of species to detoxify ozone‐generated reactive oxygen free radicals (U.S. EPA, 2006; Winner,

1994). After injuries have occurred, plants may be capable of repairing the damage to a limited

extent (U.S. EPA, 2006). Because of the differing sensitivities among plants to ozone, ozone

pollution can also exert a selective pressure that leads to changes in plant community

composition. Given the range of plant sensitivities and the fact that numerous other

environmental factors modify plant uptake and response to ozone, it is not possible to identify

threshold values above which ozone is consistently toxic for all plants.

Because plants are at the base of the food web in many ecosystems, changes to the plant

community can affect associated organisms and ecosystems (including the suitability of habitats

that support threatened or endangered species and below ground organisms living in the root

zone). Ozone impacts at the community and ecosystem level vary widely depending upon

numerous factors, including concentration and temporal variation of tropospheric ozone, species

composition, soil properties and climatic factors (U.S. EPA, 2006). In most instances, responses to

chronic or recurrent exposure in forested ecosystems are subtle and not observable for many

years. These injuries can cause stand‐level forest decline in sensitive ecosystems (U.S. EPA, 2006,

McBride et al., 1985; Miller et al., 1982). It is not yet possible to predict ecosystem responses to

ozone with certainty; however, considerable knowledge of potential ecosystem responses is

available through long‐term observations in highly damaged forests in the U.S. (U.S EPA, 2006).

a. Ozone Effects on Forests

Ozone has been shown in numerous studies to have a strong, negative effect on the health of a variety

of commercial and ecologically important forest tree species throughout the U.S. (U.S. EPA, 2007b). In the U.S.,

this data comes from the U.S. Department of Agriculture (USDA) Forest Service Forest Inventory

and Analysis (FIA) program. As part of its Phase 3 program (formerly known as Forest Health

Monitoring), FIA looks for visible foliar injury of ozone‐sensitive forest plant species at each ground

monitoring site across the country (excluding woodlots and urban trees) that meets certain

minimum criteria. Because ozone injury is cumulative over the course of the growing season,

examinations are conducted in July and August, when ozone concentrations and associated injury

are typically highest.

Monitoring of ozone injury to plants by the U.S. Forest Service has expanded over the last

15 years from monitoring sites in 10 states in 1994 to nearly 1,000 monitoring sites in 41 states in

2002. Since 2002, the monitoring program has further expanded to 1,130 monitoring sites in 45

states. Figure S3‐6 shows the results of this monitoring program for the year 2002 broken down

by U.S. EPA Regions.20 Figure S3‐7 identifies the counties that were included in Figure S3‐6, and

provides the county‐level data regarding the presence or absence of ozone‐related injury. As

shown in Figure S3‐7, large geographic areas of EPA Regions 6, 8, and 10 were not included in the

assessment. Ozone damage to forest plants is classified using a subjective five‐category biosite

index based on expert opinion, but designed to be equivalent from site to site. Ranges of biosite

values translate to no injury, low or moderate foliar injury (visible foliar injury to highly sensitive or

moderately sensitive plants, respectively), and high or severe foliar injury, which would be

expected to result in tree‐level or ecosystem‐level responses, respectively (U.S. EPA, 2006;

Coulston, 2004). The highest percentages of observed high and severe foliar injury, which are

most likely to be associated with tree or ecosystem‐level responses, are primarily found in the

Mid‐Atlantic and Southeast regions. While the assessment showed considerable regional variation

in ozone injury, this assessment targeted different ozone‐sensitive species in different parts of the

country with varying ozone sensitivity, which contributes to the apparent regional differences. It is

important to note that ozone can have other, more significant impacts on forest plants (e.g.

reduced biomass growth in trees) prior to showing signs of visible foliar injury (U.S. EPA, 2006).

20 The data are based on averages of all observations collected in 2002, which is the last year for which data are publicly available. For more information, please consult EPA’s 2008 Report on the Environment (U.S. EPA, 2008d).

Figure S3‐6: Visible Foliar Injury to Forest Plants from Ozone in U.S. by EPA Regions, 2002a, b, c

c Degree of Injury: These categories reflect a subjective index based on expert opinion. Ozone can have other, more significant impacts on forest plants (e.g. reduced biomass growth in trees) prior to showing signs of visible foliar injury.

Figure S3‐7: Presence and Absence of Visible Foliar Injury, as measured by U.S. Forest Service, 2002 (U.S. EPA, 2007)

Assessing the impact of ground‐level ozone on forests in the U.S involves understanding

the risks to sensitive tree species from ambient ozone concentrations and accounting for the

prevalence of those species within the forest. As a way to quantify the risks to particular plants

from ground‐level ozone, scientists have developed ozone‐exposure/tree‐response functions by

exposing tree seedlings to different ozone levels and measuring reductions in growth as “biomass

loss.” Typically, seedlings are used because they are easy to manipulate and measure their growth

loss from ozone pollution. The mechanisms of susceptibility to ozone within the leaves of

seedlings and mature trees are identical, and the decreases predicted using the seedlings should

be related to the decrease in overall plant fitness for mature trees, but the magnitude of the effect

may be higher or lower depending on the tree species (Chappelka and Samuelson, 1998). In areas

where certain ozone‐sensitive species dominate the forest community, the biomass loss from ozone can be sig‐

nificant. Experts have identified 2% annual biomass loss as a level of concern, which would cause

long term ecological harm as the short‐term negative effects on seedlings compound to affect

long‐term forest health (Heck and Cowling, 1997).

Ozone damage to the plants including the trees and understory in a forest can affect the

ability of the forest to sustain suitable habitat for associated species particularly threatened and

endangered species that have existence value – a nonuse ecosystem service ‐ for the public.

Similarly, damage to trees and the loss of biomass can affect the forest’s provisioning services in

the form of timber for various commercial uses. In addition, ozone can cause discoloration of

leaves and more rapid senescence (early shedding of leaves), which could negatively affect fall‐

color tourism because the fall foliage would be less available or less attractive. Beyond the

aesthetic damage to fall color vistas, forests provide the public with many other recreational and

educational services that may be impacted by reduced forest health including hiking, wildlife

viewing (including bird watching), camping, picnicking, and hunting. Another potential effect of

biomass loss in forests is the subsequent loss of climate regulation service in the form of reduced

ability to sequester carbon and alteration of hydrologic cycles.

Some of the common tree species in the United States that are sensitive to ozone are black

cherry (Prunus serotina), tulip‐poplar (Liriodendron tulipifera), and eastern white pine (Pinus

strobus). Ozone‐exposure/tree‐response functions have been developed for each of these tree

species, as well as for aspen (Populus tremuliodes), and ponderosa pine (Pinus ponderosa) (U.S.

EPA, 2007b). Other common tree species, such as oak (Quercus spp.) and hickory (Carya spp.),

have not been studied for ozone sensitivity. Consequently, with knowledge of the range of

sensitive species and the level of ozone at particular locations, it is possible to estimate the

percentage of biomass loss for each species across their range. As shown in Figure S3‐8, current

ambient levels of ozone are associated with significant biomass loss across large geographic areas

(U.S. EPA, 2009b). However, this information is unavailable for a future analysis year or

incremental to a specified control strategy.

Figure S3‐8: Estimated Biomass Loss for Black Cherry, Yellow Poplar, Sugar Maple, Eastern White Pine, Virginia Pine, Red Maple, and Quaking Aspen due to Ozone Exposure, 2006‐2008

(U.S. EPA, 2009b)*

*This map does not include other tree species that are potentially sensitive to ozone.

According to the Staff Paper, the scientific consensus is that there is no threshold for

exposures that cause effects on vegetation (Heck and Cowling 1997, U.S. EPA 2006). It is

important to note that biomass loss in tree seedlings is not intended to be a surrogate for

expected biomass loss in mature trees of the same species. Studies indicate that mature trees can

be more or less sensitive than seedlings depending on the species. Sources of uncertainty include

the ozone‐exposure/plant‐response functions, the tree abundance, and other factors (e.g., soil

moisture). Although these factors were not considered in this assessment, they can affect ozone

damage (Chappelka and Samuelson, 1998). EPA concluded in the Ozone Criteria Document that significant

interactions with acid rain are unlikely (U.S. EPA, 2006).

Since the proposal, we have expanded the analysis of qualitative assessment of ozone

impacts on forests. In this analysis, we include quantitative estimates of the tree biomass loss avoided by

the primary and secondary standards across the range of the species. In this analysis, we estimate the

biomass loss avoided for 6 tree species (i.e., ponderosa pine, red alder, black cherry, quaking

aspen, yellow (tulip) poplar, and Virginia pine) in the continental U.S. These species were selected

because they met two criteria: (1) the Staff Paper provided a W126‐derived exposure‐response

function, and (2) the Staff Paper listed the species as an ozone‐sensitive plant species (U.S. EPA,

2007b). To estimate the biomass loss avoided, we simply used the projected W126 design values in

the exposure‐response functions and subtracted the difference in biomass loss between the

baseline and hypothetical RIA control scenarios. For mapping purposes, we assume that the W126

design value is representative of the W126 levels in the county. We then overlaid a map of the

species range to focus on those areas where the species is likely to grow.21 Though each map

shows the geographical range for a species, it does not presume that an individual of that species

would be found at every point within its range. Due to uncertainties in extrapolating W126 values,

we have confined this analysis to the currently monitored counties. To calculated biomass loss

associated with the secondary standard, we simply rolled back the W126 value in only the

violating county to just attain the selected secondary standard.

Table S3‐6 shows the exposure‐response functions used to generate the tree maps. A full

list of ozone‐sensitive plant species from the Staff Paper is provided in Appendix S3A of this RIA.

Figures S3‐9 through S3‐20 map the biomass loss avoided for each of the selected tree species by

the hypothetical RIA controls for the primary standard and by the rollback to the secondary

standard. It is important to note that the modeled hypothetical RIA controls did not fully attain

the primary standard of 0.070 ppm, so this map underestimates the biomass loss avoided in

several areas, especially Southern California, Houston, Eastern Lake Michigan, and the Northeast

corridor. It is also important to note that the control strategy is likely to reduce W126 levels over a

broader geographic area than just the violating county, so this map underestimates regional

biomass loss avoided. Because we deliberately chose assumptions that underestimate tree

biomass loss avoided, we have minimized potential uncertainty, and we have high confidence that

the benefits are at least as high as those shown in the maps. Due to time and resource limitations,

we were unable to monetize the benefits associated with avoiding tree biomass loss in this

analysis. As mentioned above, these tree species provide several valuable ecosystem services,

including timber, recreational/tourism, existence value, and climate and hydrologic regulation.

Table S3‐6: Biomass Loss Functions for Trees

Species Exposure‐Response Function

Ponderosa Pine 1‐exp(‐1*(W126/159.63)^1.190)

Red Alder 1‐exp(‐1*(W126/179.06)^1.2377)

Black Cherry 1‐exp(‐1*(W126/38.92)^0.9921)

Quaking Aspen 1‐exp(‐1*(W126/109.81)^1.2198)

Virginia Pine 1‐exp(‐1*('W126/1714.64)^1)

Yellow (Tulip) Poplar 1‐exp(‐1*(W126/51.38)^2.0889)

*All functions are from Table 7F‐3 of the Staff Paper (U.S. EPA, 2007b). Each function represents the median composite function for tree seedlings.

1.1 21 The species geographic ranges are identical to those in the Staff Paper (U.S. EPA, 2007b)., and are from "Atlas of United States Trees" by Elbert L. Little, Jr, available on the Internet at http://esp.cr.usgs.gov/data/atlas/little/.

Figure S3‐9: Biomass Loss Avoided by Primary Standard in 2020 for Quaking Aspen*

* It is important to note that the modeled hypothetical RIA controls did not fully attain the primary standard of 0.070 ppm, so this map underestimates the biomass loss avoided in several areas, especially Southern California, Houston, Eastern Lake Michigan, and the Northeast corridor. Experts have identified 2% annual biomass loss as a level of concern, which would cause long term ecological harm as the short‐term negative effects on seedlings compound to affect long‐term forest health. Though each map shows the geographical range for a species, it does not presume that an individual of that species would be found at every point within its range.

Figure S3‐10: Additional Biomass Loss Avoided by Secondary Standard of 13 ppm‐hrs in 2020 for Quaking

Aspen*

* It is important to note that the control strategy is likely to reduce W126 levels over a broader geographic area than just the violating county, so this map underestimates regional biomass loss avoided. Experts have identified 2% annual biomass loss as a level of concern, which would cause long term ecological harm as the short‐term negative effects on seedlings compound to affect long‐term forest health. Though each map shows the geographical range for a species, it does not presume that an individual of that species would be found at every point within its range.

(9 counties)

Figure S3‐11: Biomass Loss Avoided by Primary Standard in 2020 for Black Cherry*

Figure S3‐12: Additional Biomass Loss Avoided by Secondary Standard of 13 ppm‐hrs in 2020 for Black Cherry*

(2 counties)

(97 counties)

Figure S3‐13: Biomass Loss Avoided by Primary Standard in 2020 for Ponderosa Pine*

Figure S3‐14: Additional Biomass Loss Avoided by Secondary Standard of 13 ppm‐hrs in 2020 for Ponderosa Pine*

(6 counties)

Figure S3‐15: Biomass Loss Avoided by Primary Standard in 2020 for Red Alder*

Figure S3‐16: Additional Biomass Loss Avoided by Secondary Standard of 13 ppm‐hrs in 2020 for Red Alder*

Figure S3‐17: Biomass Loss Avoided by Primary Standard in 2020 for Virginia Pine*

Figure S3‐18: Additional Biomass Loss Avoided by Secondary Standard of 13 ppm‐hrs in 2020 for Virginia Pine*

Figure S3‐19: Biomass Loss Avoided by Primary Standard in 2020 for Yellow (Tulip) Poplar*

Figure S3‐20: Additional Biomass Loss Avoided by Secondary Standard of 13 ppm‐hrs in 2020 for Yellow (Tulip) Poplar*

(1 county)

b. Ozone Effects on Crops

Laboratory and field experiments have shown reductions in yields for agronomic crops

exposed to ozone, including vegetables (e.g., lettuce) and field crops (e.g., cotton and wheat).

Damage to crops from ozone exposures includes yield losses (i.e., in terms of weight, number,

or size of the plant part that is harvested), as well as changes in crop quality (i.e., physical

appearance, chemical composition, or the ability to withstand storage) (U.S. EPA, 2007b). The

most extensive field experiments, conducted under the National Crop Loss Assessment

Network (NCLAN) examined 15 species and numerous cultivars. The NCLAN results show that

“several economically important crop species are sensitive to ozone levels typical of those

found in the United States” (U.S. EPA, 2006). In addition, economic studies have shown

reduced economic benefits as a result of predicted reductions in crop yields, directly affecting

the amount and quality of the provisioning service provided by the crops in question,

associated with observed ozone levels (Kopp et al, 1985; Adams et al., 1986; Adams et al.,

1989). In addition, visible foliar injury by itself can reduce the market value of certain leafy

crops (such as spinach, lettuce). According to the Staff Paper, there has been no evidence that

crops are becoming more tolerant of ozone (U.S. EPA, 2007b). Using the Agriculture Simulation

Model (AGSIM) (Taylor, 1994) to calculate the agricultural benefits of reductions in ozone

exposure, U.S. EPA estimated that attaining a W126 standard of 13 ppm‐hr would produce

monetized benefits of approximately $400 million to $620 million (inflated to 2006 dollars) (U.S.

EPA, 2007b).

According to the Staff Paper, the scientific consensus is that there is no threshold for

exposures that cause effects on vegetation (Heck and Cowling 1997, U.S. EPA 2006). Sources of

uncertainty include the ozone‐exposure/plant‐response functions, soil moisture/irrigation,

fertilization, and other factors. Agricultural systems are heavily managed and vulnerable to

adverse impacts from a variety of other factors (e.g., weather, insects, disease), which can

overshadow the ozone‐related effects. Additional research is needed to better understand the

nature and significance of interactive effects of ozone with other plant stressors (U.S. EPA,

2007b).

Since the proposal, we have expanded the analysis of qualitative assessment of ozone

impacts on crops. In this analysis, we include quantitative estimates of the crop yield loss avoided by the

primary and secondary standards across the crop production areas for 3 crops (i.e., cotton, soybean,

and winter wheat) in the continental U.S. These crops were selected because they met three

criteria: (1) the Staff Paper provided a W126‐derived exposure‐response function, (2) the Staff

Paper listed the crops as an ozone‐sensitive plant species (U.S. EPA, 2007b), and (3) the Staff

paper included maps of the crop production areas. To estimate the biomass loss avoided, we

simply used the projected W126 design values in the exposure‐response functions and

subtracted the difference in yield loss between the two scenarios. For mapping purposes, we

assume that the W126 design value is representative of the W126 levels in the county. We

then overlaid a map of the crop production area to focus on those areas where the species is

likely to be grown.22 Due to uncertainties in extrapolating W126 values, we have confined this

analysis to the currently monitored counties. To calculated biomass loss associated with the

secondary standard, we simply rolled back the W126 value in only the violating county to just

attain the selected secondary standard.

Table S3‐6 shows the exposure‐response functions used to generate the crop maps. A

full list of ozone‐sensitive crops from the Staff Paper is provided in Appendix S3A of this RIA.

Figures S3‐21 through S3‐26 map the crop yield loss avoided for each of the selected crops by

hypothetical RIA controls for the primary standard and by the rollback to the secondary

standard. It is important to note that the modeled hypothetical RIA controls did not fully attain

the primary standard of 0.070 ppm, so this map underestimates the crop yield loss avoided in

several areas, especially Southern California, Houston, Eastern Lake Michigan, and the

Northeast corridor. It is also important to note that the control strategy is likely to reduce

W126 levels over a broader geographic area than just the violating county, so this map

underestimates regional crop yield loss. Because we deliberately chose assumptions that

underestimate crop yield loss, we have minimized potential uncertainty, and we have high

confidence that the benefits are at least as high as those shown in the maps. Due to time and

resource limitations, we were unable to monetize the benefits associated with avoiding crop

yield loss in this analysis. As mentioned above, these crop species provide several valuable

ecosystem services, including especially food and fiber production.

Table S3‐6: Yield Loss Functions for Selected Crops

Crop Exposure‐Response Function

Cotton 1‐exp(‐1*(W126/96.1)^1.482)

Soybean 1‐exp(‐1*(W126/110.2)^1.359)

Winter Wheat 1‐exp(‐1*(W126/53.4)^2.367)

*All functions are from Table 7F‐1 of the Staff Paper (U.S. EPA, 2007b). Each function represents the median function.

22 Crop production areas are identical to those in the Staff Paper (U.S. EPA, 2007b) and were derived from the 2002 Census of Agriculture and from NASS 2001 County Crop Data. For more details on the crop production areas, please consult U.S. EPA (2007c).

Figure S3‐21: Yield Loss Avoided by Primary Standard in 2020 for Cotton*

*It is important to note that the modeled hypothetical RIA controls did not fully attain the primary standard of 0.070 ppm, so this map underestimates the yield loss avoided in several areas, especially Southern California, Houston, Eastern Lake Michigan, and the Northeast Corridor.

Figure S3‐22: Yield Loss Avoided by Secondary Standard of 13 ppm‐hrs in 2020 for Cotton*

*It is important to note that the control strategy is likely to reduce W126 levels over a broader geographic area than just the violating county, so this map underestimates the regional yield loss avoided.

(9 counties)

Figure S3‐23: Yield Loss Avoided by Primary Standard in 2020 for Soybean*

Figure S3‐24: Yield Loss Avoided by Secondary Standard of 13 ppm‐hrs in 2020 for Soybean*

Figure S3‐25: Yield Loss Avoided by Primary Standard in 2020 for Winter Wheat*

Figure S3‐26: Yield Loss Avoided by Secondary Standard of 13 ppm‐hrs in 2020 for Winter Wheat*

(16 counties)

c. Ozone Effects on Ornamental Plants

Urban ornamental plants are an additional vegetation category likely to experience some

degree of negative effects associated with exposure to ambient ozone levels. A variety of

ornamental species have been listed as sensitive to ozone (Abt Associates, 1995). Because ozone

causes visible foliar injury, the aesthetic value of ornamental plants (such as petunia, geranium,

and poinsettia) in urban landscapes would be reduced (U.S. EPA, 2007b). Sensitive ornamental

species would require more frequent replacement and/or increased maintenance (fertilizer or

pesticide application) to maintain the desired appearance because of exposure to ambient ozone

(U.S. EPA, 2007b). In addition, many businesses rely on healthy‐looking vegetation for their

livelihoods (e.g., horticulturalists, landscapers, Christmas tree growers, farmers of leafy crops,

etc.). The ornamental landscaping industry is a multi‐billion dollar industry that affects both

private property owners/tenants and governmental units responsible for public areas (Abt

Associates, 1995). Preliminary data from the 2007 Economic Census indicate that the landscaping

services industry, which is primarily engaged in providing landscape care and maintenance services

and installing trees, shrubs, plants, lawns, or gardens, was valued at $53 billion (U.S. Census

Bureau, 2010). Therefore, urban ornamentals represent a potentially large unquantified benefit

category. This aesthetic damage may affect the enjoyment of urban parks by the public and

homeowners’ enjoyment of their landscaping and gardening activities. In addition, homeowners

may experience a reduction in home value or a home may linger on the market longer due to

decreased aesthetic appeal. In the absence of adequate exposure‐response functions and

economic damage functions for the potential range of effects relevant to ornamental plants, we

cannot conduct a quantitative analysis to estimate these effects.

S2.11 Additional Co‐benefits

1.3

1.4 In addition to the direct benefits on vegetation that the secondary ozone NAAQS is

intended to produce, there are other co‐benefits associated with reducing ambient ozone

concentrations and ozone precursor pollutants. It is important to note that these additional

benefits are contingent upon the secondary standard being the controlling standard. In other

words, if the primary standard is controlling in all areas, there would not be any additional benefits

beyond those attributable to implementation of the primary standard. For areas where additional

control measures are needed to attain the secondary standard beyond those needed to attain the

primary standard, there would be additional benefits associated with those emission reductions.

These additional benefits are described below.

S4.6.1 Qualitative Human Health Co‐benefits

1.4.1.1.1 Reducing ozone concentrations is associated with significant human

health benefits, including avoiding mortality and respiratory morbidity. Researchers

have associated ozone exposure with adverse health effects in numerous toxicological,

clinical and epidemiological studies (U.S. EPA, 2006a). These health effects include

respiratory morbidity such as fewer asthma attacks, hospital and ER visits, school loss

days, as well as premature mortality. 23

NOx is an ozone precursor, and reducing NOx emissions would also reduce health effects

associated with NO2 exposure. Following an extensive evaluation of health evidence from

epidemiologic and laboratory studies, the Integrated Science Assessment (ISA) for Nitrogen

Dioxide concluded that there is a likely causal relationship between respiratory health effects and

short‐term exposure to NO2 (U.S. EPA, 2008b). Persons with preexisting respiratory disease,

children, and older adults may be more susceptible to the effects of NO2 exposure. The NO2 ISA

identified four short‐term morbidity endpoints as a “likely causal relationship”: asthma

exacerbation, respiratory‐related emergency department visits, and respiratory‐related

hospitalizations. The NO2 ISA also concluded that the relationship between short‐term NO2

exposure and premature mortality was “suggestive but not sufficient to infer a causal relationship”

because it is difficult to attribute the mortality risk effects to NO2 alone. Although the NO2 ISA

stated that studies consistently reported a relationship between NO2 exposure and mortality, the

effect was generally smaller than that for other pollutants such as PM. The differing evidence and

associated strength of the evidence for these different effects is described in detail in the NO2 ISA.

1.4.1.1.2 Furthermore, NOX and VOCs are precursors to PM2.5 as well as ozone.

Reducing exposure to PM2.5 is associated with significant human health benefits,

including avoiding mortality and respiratory morbidity.24 Researchers have associated

PM2.5‐ exposure with adverse health effects in numerous toxicological, clinical and

epidemiological studies (U.S. EPA, 2009). These health effects include premature

mortality for adults and infants, cardiovascular morbidity such as heart attacks, hospital

admissions, and respiratory morbidity such as fewer asthma attacks, bronchitis, hospital

and ER visits, work loss days, restricted activity days, and respiratory symptoms.25

23 See Chapter 6 of the 2008 Ozone RIA, the updated benefits analysis in Section 3 of this supplemental for additional information on the ozone‐related health effects associated with attaining the primary standard.

24 See Chapter 6 of the 2008 Ozone RIA, the updated benefits analysis in Section 3 of this supplemental for additional information on the PM2.5‐related health effects associated with attaining the primary standard.

25 See Chapter 6 of the 2008 Ozone RIA, the updated benefits analysis in Section 3 of this supplemental for additional information on the ozone‐related health effects associated with attaining the primary standard.

S4.6.2 Qualitative Welfare Co‐benefits

In addition to impacts on vegetation, ozone can also impact other welfare categories,

including damage to certain manmade materials (e.g., elastomers, textile fibers, dyes, paints, and

pigments) and climate interactions. The amount of damage to actual in‐use materials and the

economic consequences of that damage are poorly characterized, however, and the scientific

literature contains very little new information to adequately quantify estimates of materials

damage from photochemical oxidants (U.S. EPA, 2007b). Ozone is a well‐known greenhouse gas,

and the overall body of scientific evidence suggests that high concentrations of ozone on the

regional scale could have a discernable influence on climate, leading to surface temperature and

hydrological cycle changes (U.S. EPA, 2006).

1.4.1.1.3

1.4.1.1.4 NOx is an ozone precursor, and reducing NOx emissions would also

reduce adverse welfare effects from acidic deposition, nutrient enrichment, and

visibility impairment. Deposition of nitrogen causes acidification, which can cause a loss

of biodiversity of fishes, zooplankton, and macro invertebrates in aquatic ecosystems, as

well as a decline in sensitive tree species, such as red spruce (Picea rubens) and sugar

maple (Acer saccharum) in terrestrial ecosystems. In the northeastern United States, the

surface waters affected by acidification are a source of food for some recreational and

subsistence fishermen and for other consumers and support several cultural services,

including aesthetic and educational services and recreational fishing. Biological effects

of acidification in terrestrial ecosystems are generally linked to aluminum toxicity, which

can cause reduced root growth, which restricts the ability of the plant to take up water

and nutrients. These direct effects can, in turn, increase the sensitivity of these plants to

stresses, such as droughts, cold temperatures, insect pests, and disease leading to

increased mortality of canopy trees. Terrestrial acidification affects several important

ecological services, including declines in habitat for threatened and endangered species

(cultural), declines in forest aesthetics (cultural), declines in forest productivity

(provisioning), and increases in forest soil erosion and reductions in water retention

(cultural and regulating). (U.S. EPA, 2008c)

Deposition of nitrogen is also associated with aquatic and terrestrial nutrient enrichment.

In estuarine waters, excess nutrient enrichment can lead to eutrophication. Eutrophication of

estuaries can disrupt an important source of food production, particularly fish and shellfish

production, and a variety of cultural ecosystem services, including water‐based recreational and

aesthetic services. Terrestrial nutrient enrichment is associated with changes in the types and

number of species and biodiversity in terrestrial systems. Excessive nitrogen deposition upsets the

balance between native and nonnative plants, changing the ability of an area to support

biodiversity. When the composition of species changes, nonnative grasses can fuel more frequent

and more intense wildfires. (U.S. EPA, 2008c)

Reducing NOx and the secondary formation of PM2.5 would reduce visibility impairment

throughout the U.S. Fine particles with significant light‐extinction efficiencies include sulfates,

nitrates, organic carbon, elemental carbon, and soil (Sisler, 1996). These suspended particles and

gases degrade visibility by scattering and absorbing light. Higher visibility impairment levels in the

East are due to generally higher concentrations of fine particles, particularly sulfates, and higher

average relative humidity levels. Visibility has direct significance to people’s enjoyment of daily

activities and their overall sense of wellbeing. Good visibility increases the quality of life where

individuals live and work, and where they engage in recreational activities.

S2.12 References

Abt Associates, Inc. 1995. Urban ornamental plants: sensitivity to ozone and potential economic losses. U.S. EPA, Office of Air Quality Planning and Standards, Research Triangle Park. Under contract to RADIAN Corporation, contract no. 68‐D3‐0033, WA no. 6. pp. 9‐10.

Adams, R. M., Glyer, J. D., Johnson, S. L., McCarl, B. A. 1989. A reassessment of the economic effects of ozone on U.S. agriculture. Journal of the Air Pollution Control Association, 39, 960‐968.

Adams, R. M., Hamilton, S. A., McCarl, B. A. 1986. The benefits of pollution control: the case of ozone and U.S. agriculture. American Journal of Agricultural Economics, 34, 3‐19.

Chappelka, A.H., Samuelson, L.J. 1998. Ambient ozone effects on forest trees of the eastern United States: a review. New Phytologist, 139, 91‐108.

Coulston, J.W., Riitters, K.H., Smith, G.C. 2004. A preliminary assessment of the Montreal process indicators of air pollution for the United States. Environmental Monitoring and Assessment, 95, 57‐74.

De Steiguer, J., Pye, J., Love, C. 1990. Air Pollution Damage to U.S. Forests. Journal of Forestry, 88(8), 17‐22.

Fox, S., Mickler, R. A. (Eds.). 1996. Impact of Air Pollutants on Southern Pine Forests, Ecological Studies. (Vol. 118, 513 pp.) New York: Springer‐Verlag.

Grulke, N.E. 2003. The physiological basis of ozone injury assessment attributes in Sierran conifers. In A. Bytnerowicz, M.J. Arbaugh, & R. Alonso (Eds.), Ozone air pollution in the Sierra Nevada: Distribution and effects on forests. (pp. 55‐81). New York, NY: Elsevier Science, Ltd.

Heck, W.W. &Cowling E.B. 1997. The need for a long term cumulative secondary ozone standard – an ecological perspective. Environmental Management, January, 23‐33.

Kopp, R. J., Vaughn, W. J., Hazilla, M., Carson, R. 1985. Implications of environmental policy for U.S. agriculture: the case of ambient ozone standards. Journal of Environmental Management, 20, 321‐331.

McBride, J.R., Miller, P.R., Laven, R.D. 1985. Effects of oxidant air pollutants on forest succession in the mixed conifer forest type of southern California. In: Air Pollutants Effects on Forest Ecosystems, Symposium Proceedings, St. P, 1985, p. 157‐167.

Miller, P.R., O.C. Taylor, R.G. Wilhour. 1982. Oxidant air pollution effects on a western coniferous forest ecosystem. Corvallis, OR: U.S. Environmental Protection Agency, Environmental Research Laboratory (EPA600‐D‐82‐276).

Office of Management and Budget (OMB). 2003. Circular A‐4: Regulatory Analysis. Washington, DC. Available on the internet at http://www.whitehouse.gov/omb/circulars/a004/a‐4.html

Prasad, A.M. and Iverson, L.R. 2003. Little’s range and FIA importance value database for 135 eastern U.S. tree species. Northeastern Research Station, USDA Forest Service. Available on the Internet at

http://www.fs.fed.us/ne/delaware/4153/global/littlefia/index.html Pye, J.M. 1988. Impact of ozone on the growth and yield of trees: A review. Journal of Environmental Quality, 17, 347‐360.

Smith, G., Coulston, J., Jepsen, E., Prichard, T. 2003. A national ozone biomonitoring program— results from field surveys of ozone sensitive plants in Northeastern forests (1994‐2000). Environmental Monitoring and Assessment, 87, 271‐291.

Taylor R. 1994. Deterministic versus stochastic evaluation of the aggregate economic effects of price support programs. Agricultural Systems 44: 461‐473.

Tingey, D.T., and Taylor, G.E. 1982. Variation in plant response to ozone: a conceptual model of physiological events. In M.H. Unsworth & D.P. Omrod (Eds.), Effects of Gaseous Air Pollution in Agriculture and Horticulture. (pp.113‐138). London, UK: Butterworth Scientific.

U.S. Census Bureau. 2010. American FactFinder: “Sector 56: EC075612: Administrative Support and Waste Management and Remediation Services: Industry Series: Preliminary Comparative Statistics for the United States (2002 NAICS Basis): 2007 and 2002.” Available on the Internet at http://factfinder.census.gov (accessed August 25, 2010)

U.S. Environmental Protection Agency (U.S. EPA). 1999. The Benefits and Costs of the Clean Air Act, 1990‐2010. Prepared for U.S. Congress by U.S. EPA, Office of Air and Radiation, Office of Policy Analysis and Review, Washington, DC, November; EPA report no. EPA410‐R‐99‐001. Available on the Internet at http://www.epa.gov/oar/sect812/1990‐2010/chap1130.pdf

U.S. Environmental Protection Agency (U.S. EPA). 2006. Air Quality Criteria for Ozone and Related Photochemical Oxidants (Final). EPA/600/R‐05/004aF‐cF. Washington, DC: U.S. EPA. February. Available on the Internet at http://cfpub.epa.gov/ncea/CFM/recordisplay.cfm?deid=149923.

U.S. Environmental Protection Agency (U.S. EPA). 2007a. Guidance on the Use of Models and Other Analyses for Demonstrating Attainment of Air Quality Goals for Ozone, PM2.5, and Regional Haze. Office of Air Quality Planning and Standards. EPA‐454/B‐07‐002. April. Available on the Internet at http://www.epa.gov/scram001/guidance/guide/final‐03‐pm‐rh‐guidance.pdf

U.S. Environmental Protection Agency (U.S. EPA). 2007b. Review of the National Ambient Air Quality Standards for Ozone: Policy assessment of scientific and technical information. Staff Paper. Office of Air Quality Planning and Standards. EPA‐452/R‐07‐007a. July. Available on the Internet at http://www.epa.gov/ttn/naaqs/standards/ozone/data/2007_07_ozone_staff_paper.pdf

U.S. Environmental Protection Agency (U.S. EPA). 2007c. Technical Report on Ozone Exposure, Risk, and Impact Assessments for Vegetation. Office of Air Quality Planning and Standards. EPA‐ 452/R‐07‐002. January. Available on the Internet at http://www.epa.gov/ttn/naaqs/standards/ozone/s_o3_pr.html

U.S. Environmental Protection Agency (U.S. EPA). 2008a. Regulatory Impact Analysis, 2008 National Ambient Air Quality Standards for Ground‐level Ozone, Chapter 6. Office of Air Quality Planning and Standards, Research Triangle Park, NC. March. Available at http://www.epa.gov/ttn/ecas/regdata/RIAs/6‐ozoneriachapter6.pdf

U.S. Environmental Protection Agency (U.S. EPA). 2008b. Integrated Science Assessment for Oxides of Nitrogen ‐ Health Criteria (Final Report). National Center for Environmental Assessment, Research Triangle Park, NC. July. Available on the Internet at http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=194645

U.S. Environmental Protection Agency (U.S. EPA). 2008c. Integrated Science Assessment for Oxides of Nitrogen and Sulfur – Ecological Criteria National (Final Report). National Center for Environmental Assessment, Research Triangle Park, NC. EPA/600/R‐08/139. December. Available on the Internet at http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=201485

U.S. Environmental Protection Agency (EPA). 2008d. EPA’s 2008 Report on the Environment. National Center for Environmental Assessment, Washington, DC; EPA/600/R‐07/045F. May Available from the National Technical Information Service, Springfield, VA. Available on the Internet at http://www.epa.gov/roe

U.S. Environmental Protection Agency (U.S. EPA). 2009a. Integrated Science Assessment for Particulate Matter (Second External Review Draft). National Center for Environmental Assessment, Research Triangle Park, NC. EPA/600/R‐08/139B. July. Available on the Internet at http://cfint.rtpnc.epa.gov/ncea/prod/recordisplay.cfm?deid=210586

U.S. Environmental Protection Agency (U.S. EPA). 2009b. The NOx Budget Trading Program: 2008 Environmental Results. Clean Air Markets Division. September. Available on the Internet at http://www.epa.gov/airmarkt/progress/NBP_3/NBP_2008_Environmental_Results.pdf

U.S. Environmental Protection Agency (U.S. EPA). 2010. updated Regulatory Impact Analysis (RIA) for the Reconsideration of the 2008 Ozone National Ambient Air Quality Standard (NAAQS). Office of Air Quality Planning and Standards, Research Triangle Park, NC. January. Available on the Internet at http://www.epa.gov/ttn/ecas/regdata/RIAs/s1‐supplemental_analysis_full.pdf

U.S. Environmental Protection Agency ‐ Science Advisory Board (U.S. EPA‐SAB). 2007. Clean Air Scientific Advisory Committee’s (CASAC) Review of the Agency’s Final Ozone Staff Paper. EPA‐ CASAC‐07‐002. Available on the Internet at http://yosemite.epa.gov/sab/sabproduct.nsf/FE915E916333D776852572AC007397B5/$File/casa c‐07‐002.pdf

Winner, W.E. 1994. Mechanistic analysis of plant responses to air pollution. Ecological Applications, 4(4), 651‐661.

Winner, W.E., and C.J. Atkinson. 1986. Absorption of air pollution by plants, and consequences for growth. Trends in Ecology and Evolution 1:15‐18.

2 APPENDIX S3A: OZONE SENSITIVE PLANTS (FROM U.S. EPA, 2007)

Allegheny blackberry Rubus allegheniensis American elder Sambucus canadensis American hazelnut Corylus americana American sycamore Platanus occidentalis Basswood Tilia Americana Big‐leaf aster Aster macrophyllus Black cherry Prunus serotina Black huckleberry Gaylussacia baccata Black locust Robinia pseudoacacia Black poplar Populus balsamifera trichocarpa Blue elderberry Sambucus mexicana Box elder Acer negundo California black oak Quercus kelloggii Chokecherry Prunus virginiana Common milkweed Asclepias syriaca Cottonwood Populus deltoids Crown‐beard Verbesina occidentalis Cutleaf coneflower Rudbeckia laciniata Dogbane, Indian hemp Apocynum cannibinum Evening primrose Oenothera elata Goldenrod Solidago altissima Gooding’s willow Salix goodingii Green ash Fraxinus pennsylvanica Groundnut Apios americana Huckleberry Vaccinium membranaceum Jack pine Pinus banksiana Jeffrey pine Pinus jeffreyi Loblolly pine Pinus taeda Maleberry Lyonia ligustrina Monterey pine Pinus radiata Mountain dandelion Krigia montana Mugwort Artemisia douglasiana Ninebark Physocarpus capitatus Northern fox grape Vitis labrusca Ohio Buckeye, Horse chestnut Aesculus glabra Pacific ninebark Physocarpus malvaceum Paper birch Betula papyrifera Pinus ponderosa Pinus ponderosa Pitch pine Pinus rigida Poke milkweed Asclepias exaltata Ponderosa pine Pinus ponderosa Quaking aspen Populus tremuloides Red alder Alnus rubra Red elderberry Sambucus racemosa Redbud Cercis Canadensis

Saskatoon serviceberry Amelanchier alnifolia Sassafras Sassafras albidum Scouler’s willow Salix scouleriana Serviceberry Amelanchier alnifolia Silver wormwood Artemisia ludoviciana Single‐leaf ash Fraxinus anomala Skunkbush Rhus trilobata Smooth cordgrass Spartina alterniflora Snowberry Symphoricarpos albus Speckled alder Alnus rugosa Spreading dogbane Apocynum androsaemifolium Swamp milkweed Asclepias incarnata Sweet mock orange Philadelphus coronarius Sweetgum Liquadambar styraciflua Table‐mountain pine Pinus pungens Tall milkweed Asclepias exaltata Thimbleberry Rubus parviflorus Thornless blackberry Rubus canadensis Tree‐of‐heaven Ailanthus altissima Twinberry Lonicera involucrata Virgin’s bower Clematis virginiana Virginia creeper Parthenocissus quinquefolia Virginia pine Prunus virginiana White ash Fraxinus americana White snakeroot Eupatorium rugosum White stem blazingstar Mentzelia albicaulis Whorled aster Aster acuminatus Winged sumac Rhus copallina Yellow‐poplar Liriodendron tulipifera Ozone Sensitive Crops Cotton Peanuts Potatoes Soybeans Tobacco Winter Wheat

Sand blackberry Rubus cuneifolius

3 APPENDIX S3B: CALCULATING THE W126 INDEX

Steps in calculating W126 value for a particular site:

1. Measure O3 concentrations for each hour within 12‐hour daylight period (8 am to 8 pm)

2. Weight each hourly O3 concentration to get a W126 value: lower concentrations receive less

weight than higher concentrations

3. Add the 12 weighted hourly W126 values to calculate daily W126 value for each day

4. Sum daily W126 values within each month to get a monthly W126 value

5. Identify the consecutive 3‐month period whose sum of monthly W126 values produces the highest

W126 index value. This maximum consecutive 3‐month sum = seasonal W126 value for that site (in

ppm‐hrs)

Example of weighting over 5‐hour period:

Daily value = sum of values over 12 daylight

hours

Hourly O3 (primary) Weight W126

(ppm‐hrs)

0.03 0.01 0.00

0.05 0.11 0.01

0.06 0.30 0.02

0.08 0.84 0.07

0.10 1.0 0.10

SUM: 0.20

__MACOSX/._risk+201107_OMBdraft-OzoneRIA.pdf

risk+fear+blame+public+safety.pdf

RISK, FEAR, BLAME, SHAME AND THE REGULATION OF PUBLIC SAFETY

JONATHAN WOLFF∗

University College London

The question of when people may impose risks on each other is of funda- mental moral importance. Forms of “quantified risk assessment,” especially risk cost-benefit analysis, provide one powerful approach to providing a systematic answer. It is also well known that such techniques can show that existing resources could be used more effectively to reduce risk overall. Thus it is often argued that some current practices are irrational. On the other hand critics of quantified risk assessment argue that it cannot adequately capture all relevant features, such as “societal concern” and so should be abandoned. In this paper I argue that current forms of quantified risk assessment are inadequate, and in themselves, therefore, insufficient to demonstrate that current practices are irrational. In particular, I will argue that insufficient attention has been given to the cause of a hazard, which needs to be treated as a primary variable in its own right. However rather than reject quantified risk assessment I wish to supplement it by proposing a framework to make explicit the role causation plays in the understanding of risk, and how it interacts with factors which influence perception of risks and other attitudes to risk control. Once an improved description of risk perception is available

∗ This paper was drafted as part of the project “Philosophical Foundations of Public Policy: Rethinking Cost-Benefit Analysis” funded by the Arts and Humanities Research Board under its Innovations Award Scheme. The final version was written up as part of an extended leave grant also from the AHRB. I am very grateful for the Board’s assistance. Earlier drafts of this paper were presented to the Society of Applied Philosophy in Manchester, and at Lancaster University, and I am very grateful to the audiences at both occasions for their discussion and comments. I am also pleased to thank Andrew Sharpe at the Rail Safety and Standards Board for bringing this topic to my attention, and his encouragement and useful comments on earlier drafts of this work. Finally I am very grateful to the referees and editors of Economics and Philosophy, and especially Luc Bovens, for suggestions which have led, I hope, to significant improvements in this paper.

409

410 JONATHAN WOLFF

it will become possible to have a more informed debate about the normative question: how safety should be regulated.

1. INTRODUCTION

When can people knowingly impose risks on each other? This question is central to such issues as transport safety policy, environmental risk, and health and safety at work. In many cases decisions will be taken by politicians and company directors, advised by engineers and economists. Different understandings and different approaches may well be followed in different cases. This gives rise to at least two causes for concern. First, as is well known, an uncoordinated approach may give rise to serious inefficiencies, in that it may be possible to use existing resources in far more effective ways. The second is also well known, if less often discussed. The question of what risks we can impose on each other is fundamentally a moral one: the proper aims of safety regulation and the proper means of achieving these aims are moral issues. It is unclear that between them politicians, company directors, economists, and engineers will be properly attuned to all the relevant moral issues. These two concerns come together in the following way: unless we have a firm understanding of the morally justified aims of safety regulation we cannot say whether any particular measure is rational or irrational, efficient or inefficient.

The purpose of this paper is to lay the groundwork so that the moral questions appear in clear focus. I will not, however, attempt to settle these questions here. Rather, in order to understand the proper aims of safety regulation we need to have a good understanding of its possible aims, or, better, the plausible aims. Eventually this should allow an evaluation of such possibilities. My strategy will be to look at the way in which safety decisions are influenced, made, and perceived in practice, in order to bring out the diversity of factors which come into play. Accordingly, the remainder of this paper falls into three main parts. Section 2 lays out the background to the issue, looking at the standard paradigm of quantified risk assessment, and the argument that some current practices can be seen to be irrational in such terms. Section 3 introduces and explores a number of complicating factors – summarized as fear, blame, and shame – while Section 4 produces an overall framework in which it can be seen how these differing factors relate to each other. The main theoretical proposal is to add the concept of “cause” to that of hazard and probability, as primary variables in the analysis, and to show how this inter-acts with fear, blame, and shame which are here treated as “secondary” variables. One conclusion is, perhaps, a predictable one: that greater clarification of the proper aims of safety regulation is necessary before we can argue that any particular safety measure should be denounced as irrational or inefficient. But the main pay-off of the analysis is the beginnings of a

RISK, FEAR, BLAME, SHAME AND THE REGULATION OF PUBLIC SAFETY 411

model of the “anatomy of risk” which sets the groundwork for informed normative discussion of the proper aims of safety control and regulation.

2. QUANTIFIED RISK ASSESSMENT

2.1 Safety has a price

Life is a risky business. We all face threats to life and safety every day. Some circumstances, and especially some working environments, seem especially risky. What should be done about this? What should be done, for example, about hazardous working conditions?

As a first thought, it might be proposed that it is always wrong knowingly to inflict risks on others, and so there should be an absolute duty on factory owners and others to eliminate all known risks to their workers. But a moment’s reflection shows that this is an impossible aim. Virtually any human activity involves some risk. Even a perfectly maintained and serviced machine might malfunction with unpredictable effects. Even very sensible workers can trip on an even floor. These are risks we know about, yet cannot eliminate entirely.

However, even though we cannot eliminate all risks, we might be able to reduce many of them. So perhaps the goal should be to reduce risks in so far as this is technically possible. But this again seems to have some absurd consequences. We might virtually eliminate fatal road accidents by lowering the speed limit to 10 miles an hour. We can end injuries to coal miners by closing the mines. Neither suggestion would be treated as a serious contribution to safety policy debate. Safety – and therefore life and limb – is not the only thing we value, nor, it seems, is it always the highest value.

The lesson is that safety has a price, in terms of its impact on other things we want or value, and there are limits to what we are prepared to pay. It seems that in generating policy we are forced to put a value on life – and this inevitably seems to mean a financial value – which helps us generate rules about how much firms, and in some cases the government, can reasonably be expected to pay for safety improvements.

This may seem callous or inhuman. Don’t we know that life has infinite value? But what is the alternative? Not putting a price on safety? Allowing companies to operate with dangerous machinery, because we can’t put an infinite value on life and any finite value is arbitrary and demeaning? This hardly seems an improvement.

2.2 The standard paradigm

Safety is regulated in somewhat different ways in different jurisdictions, although the differences in detail need not detain us. As it is the example I know best I shall take the UK as my main example. Work-related safety

412 JONATHAN WOLFF

is regulated by the Health and Safety Executive, and its general approach is explained in a publication called Reducing Risk, Protecting People (Health and Safety Executive 2001). Here, for simplicity, we will concentrate on risks of death, although other risks are covered too. In standard cases the basic approach is to divide risks into three categories. Some risks of death are too high, in probabilistic terms, to allow and must be reduced (unless there are special circumstances). Some are so low (in the sense that the probabilities are minute) that they do not require any special measures. In the large middle ground are risks which although in some sense are “broadly tolerable,” should be reduced “as far as is reasonably practicable” (44ff.).

If a risk falls within the broadly tolerable region those in control of the risk are required to perform a risk cost-benefit analysis. To carry this out risk assessors must calculate the probability of death for the risk under consideration. Suppose, for example, a piece of machinery could trap and kill a careless and negligent worker and that machines of this type kill 1 in 10,000 of their operators every year. And, as ease of calculation would have it, in your large factory you have 1,000 operators. Hence you should expect a death every 10 years or 0.1 death a year, assuming that there is nothing special about your factory.1

Let us suppose there is a possible modification to the machine which could reasonably be predicted to eliminate half the deaths in your factory, thus saving 0.05 lives a year. Should you introduce the modification or not?

For simplicity let us assume that you believe that the machines will be in use for another 10 years, and let us also apply no discount rate for future deaths. To know whether to introduce the modification two further pieces of information are required. First, how much the modification will cost, and second, what financial value should be placed on preventing a fatality (VPF). Currently the UK operates with a figure of a little over £1 million (£1 million at 2001 prices, and hence a few thousand more now, allowing for inflation). How to calculate such a figure is a matter of some controversy, which I shall not enter into here, but for the purposes of this paper it makes no difference what figure is selected.2 So let us round down to 1 million. Consequently as the modification will save 0.5 of a life over ten years you would be required to introduce it if (and only if) it will cost less than £500,000.

1 Where there are frequencies of this nature probabilities are relatively easy to estimate. Of course this is an unusual case, and especially in the case of new risks there can be great controversy about actual probabilities.

2 For an illustration of one important method for determining such a value see Jones-Lee et al. (1999). In the US different figures are used. According to Richard Posner, current estimates range from $4 million to $9 million with a mean of $7 million. Posner (2002: 166). For further discussion of the US approach see Sunstein (2002: 153–90).

RISK, FEAR, BLAME, SHAME AND THE REGULATION OF PUBLIC SAFETY 413

This general approach is intended to apply to all work-related risks in the UK. It was first devised as a response to the risks of nuclear power stations. It also, for some reason, applies to the risks of railway travel, and the location of airports, although not to road travel, and not for product safety. However, it could be easily be adapted to these cases too.

2.3 The irrationality argument

What I have described is one application of risk cost-benefit analysis, which in turn is a form of quantified risk analysis. Many people find these types of approaches rather chilling, at least at first. However risk cost- benefit analysis can be used as a very powerful tool for examining current practice. One familiar debate concerns the contrasting situations of rail and road safety.

In recent years the UK public has become very concerned about railway safety, and in particular about train crashes. How many passengers die in railway accidents in the UK each year? In a recent, as yet unpublished, study around 1000 people were asked this question. Their answers lay in the range 10 to 2,000, with a mean, excluding outliers, of 99. In the last decade the actual average number was about six deaths of passengers per year. Passenger deaths, in fact, are only a small proportion of deaths on the railways in the UK. Over recent years the annual average number of deaths on the railways as a whole is about 275, with the vast majority being suicides and trespassers. (Others include members of the workforce, people at stations, and occupants of vehicles on the line.)

Recently technology has been introduced, at a cost of about £585 million, to make it less likely that a train will run into another train if it runs through a red light. (Commission for Integrated Transport 2004: section 3) Arguably this is already saving lives, albeit at a cost of somewhere between £5 million and £20 million a life saved, depending on what is counted (whether or not injuries are counted as fractions of death) and how long a time period is taken. Further technological innovation – Automatic Train Protection – is being discussed which would make it theoretically impossible – i.e. impossible if the system works – for a train to run a red light. The system currently under active discussion, on the lowest estimate I have seen, would cost £3.6 billion. Even its defenders admit that it will cost close to £100 million for each rail passenger life saved.3

3 Commission for Integrated Transport, 2004. Some non-passenger lives, it has been claimed, would also be saved, although some workers may die installing it. There are further arguments on both sides claiming that the cost per life saved would be higher or lower. It is worth noting that currently most of those who defend the introduction of ATP do so on the basis of the commercial and performance benefits it is calculated to bring, and not on its contribution to safety.

414 JONATHAN WOLFF

Turning now to the roads, in recent years about 3,500 people die in the UK each year (the figures are remarkably stable), of whom around 2,000 are occupants of cars. By every measure travelling by road is much more dangerous than travelling by rail. While there are always reasons to be skeptical about any particular calculation, there are indications that the hundreds of millions recently spent on the railways could have saved perhaps ten times as many people on the roads. And for £3.6 billion miracles could happen, at least if road safety campaigners are to be believed.

To bring out the disparity between our attitude to road and rail safety consider the arguments of commentators who have discussed the aftermath of the Hatfield rail accidents. At Hatfield a high-speed train was derailed when the track it was travelling on shattered. Four people died, and others were injured. A very cautious response followed, and speed restrictions were enforced throughout the network so that all relevant track could be checked for similar faults. The resulting memorable chaos meant that train travel was unreliable to an unprecedented degree; it was as if there was no timetable. Frustrated passengers took to their cars. It has been estimated that there may have been as many as five extra road deaths in the first month as a result.4 Although the comparison may seem rather mischievous, it appears that in some sense we would have been better off with no speed restrictions and a Hatfield sized crash every second month, compared to what, it is claimed, actually happened.

These examples bring out a stark general message. We can easily save more lives by spending our resources in different ways. Indeed some analysts are raising the possibility (behind closed doors) that we should significantly reduce the amount of money we already spend on railway safety, diverting the resources to road safety, public health, or even foreign aid. Essentially the same argument is made in the US concerning consumer protection and environmental protection. Huge sums are being spent to mitigate tiny risks, while much larger risks go ignored (Sunstein 2002). It seems hard to avoid the conclusion that this is an irrational state of affairs. For this reason I shall call it the irrationality argument, and it is growing in popularity. So, for example, Bjorn Lomborg has notoriously argued that instead of spending trillions of dollars slowing down global warming by a few years, we would do better to spend a fraction of that money helping developing countries build the level of infrastructure that will

4 This figure is reported as an estimate, but without attribution, by Sunstein (2002: 2). However Sunstein’s diagnosis of the change in behavior as resulting from individual over-reaction to small risks seems mistaken. Rather it was a perfectly rational response to massive disruption to the service, caused by the industry’s highly cautious response to the incident. Further, his comment that these five road deaths is “nearly equal the total number of deaths from train accidents in the previous thirty years” is extraordinary. The true figure, according to Evans (2005), is close to 200.

RISK, FEAR, BLAME, SHAME AND THE REGULATION OF PUBLIC SAFETY 415

allow them permanently to cope with effects of global warming (Lomborg 2003, drawing on Lomborg 2001). Although, no doubt, the science and economics are contestable, this line of argument – another application of the irrationality argument – can be made to seem quite compelling. Those who oppose it are portrayed as supporters of the politics of gesture and a dangerous menace to rational thought and even to life on earth.

I hold no brief for current practices, and I accept that thinking through the consequences of the irrationality argument can be liberating. Yet I will argue that we should not be quite so quick to think that the irrationality argument settles anything. There are subtleties which we need to investigate first. This is another way of saying that it is possible that risk cost-benefit analysis, as currently used, does not capture all the information that is needed in order to make the best decisions.

3. COMPLICATING FACTORS

3.1 Fear reduction

Each of us is afraid of some risks, but less so of others. What is the relation between risk and fear? To make progress we must distinguish objective risk and subjective risk, or in other words belief in risk. For there need be no relation between objective risk and fear. How can you fear an unknown risk? Well, of course you can fear the unknown, but there is no reason to think that anyone’s fear will be related to the actual risk. Furthermore, it is well known that fear may be out of all proportion to the objective risk; this is the central finding of those who argue for the “social amplification of risk”: essentially the idea that there are numerous social mechanisms which can make people feel that risks are much higher, or lower, than in fact they are (Pidgeon et al. 2003). But, more pertinently, is there a clear correlation between subjective risk and fear? Obviously they are not the same thing as one is a belief and the other an emotion. But there could be a causal, or even a partially constitutive, relationship. Let us, in the first instance, assume there is some sort of direct connection, although we will examine this shortly.

It seems clear that people may pay their own money, or agree to spend taxpayers’ money, to reduce risks in order increase their sense of safety.5

Or, to put this another way round, it is reasonable to believe that people are prepared to pay more to reduce those risks they fear most. But notice if the point of risk regulation is to reduce each individual’s subjective sense of being at risk then already we can see that the irrationality argument may not go through in such straightforward fashion. For the irrationality argument concerned objective risk. It argues that particular risk reduction

5 The distinction between death reducing policies and anxiety reducing policies is explained well in Schelling (1968).

416 JONATHAN WOLFF

policies are irrational because they are inefficient means of achieving the goal of reducing objective risk. The point about inefficiency relative to the goal of risk reduction can be granted, but irrationality need not follow. For if the point of safety policies is, or includes, the reduction of subjective risk – to increase feelings of security – then the failure to reduce objective risk is no longer decisive. Indeed, we can generalize this point. If the goal of safety policy includes anything other than objective risk reduction, then it is moot whether the irrationality argument goes through. Everything needs to be recalculated in the light of the new goals of policy.

Here, though, defenders of the irrationality argument may well change tack. The point, they will say, of safety regulation ought to be reduction of objective risk. How plausible is this? Note that this response need not downplay the importance of fear, anxiety, and insecurity in people’s lives. Such emotions, it can be conceded, are terrible things to suffer. Perhaps they are much worse than the presence of small risks in one’s life. After all, small risks rarely lead to actual harm, whereas fear and so on can have a constant dampening effect on one’s spirits. But, so the argument goes, the way to respond to this is not to introduce expensive means of reducing what may already be barely significant risks. Instead, education is needed so that public fears track the real risks, and people worry about only what they ought to be worried about. False fears should be calmed by good information and the same means should be used to ensure that people come to fear the objective risks they face.

While this appears very attractive it nevertheless relies on some assumptions which may well be false. In particular, it relies on an intuitive assessment of the effectiveness – and hence the costs and benefits – of alternative policies. Changing public attitudes and emotions through provision of information is very difficult. Or rather, it is difficult to change public attitudes in a positive direction. It is expensive to attempt, and rarely more than marginally effective. Who can we rely on to provide accurate information? In the current climate people profess to distrust scientists, doctors, the government, bankers, big business, the police, the media, civil servants, lawyers, educationalists, anyone in the employ of the government and, indeed, anyone on a decent salary. In the light of this it is rather hard to see how anyone comes to any beliefs about anything. More pertinently, the prospects for a public education strategy which would bring subjective and objective risk into step seem pretty bleak. I am not proud of humanity for this, but it may turn out that once we do the sums, the most cost-effective way of reducing public anxiety could be to spend huge amounts of money on almost useless safety devices. Certainly anyone who has traveled by air lately, and seen what is being done in the name of reducing risks of terrorism may well have had the thought: obviously not much better than useless, but nevertheless somehow strangely reassuring, at least for some people. But in fact the practice of symbolic safety measures

RISK, FEAR, BLAME, SHAME AND THE REGULATION OF PUBLIC SAFETY 417

to reduce fears is much older. Has, in recent times, anyone’s life been saved on a standard commercial aircraft by a life-jacket? Or by that little whistle?

The distinction between objective risk, and fear, is hardly news. It is becoming common in the UK to make a distinction between reduction in crime, and reduction in the fear of crime. This is clearly inspired by the recognition that fear of crime can have a deeper impact on people’s lives than crime itself, coupled with the thought that one way of reducing crime is to make people hyper-vigilant, which may make them hyper-scared too. So the two goals have a complex relation. Yet was it right to assume that fear and anxiety are so directly correlated with subjective risk, which I understand as belief in risk? This is not entirely clear. Studies show that women are more fearful of crime than men, even though they know full well that men are more often the victims of crime than women (Burgess- Jackson 1994). However, this too is complex. Consider an example from John Adams. Are roads safer now for children pedestrians than they were in the 1950s? Statistically the result is surprising. Fewer child pedestrians are killed on the roads now than for decades. But this, he argues, is because we believe that roads are so dangerous that we keep our children away from them (Adams 2001: 10–14). What this shows is that we need to be very careful in how we collect and present our statistics. If women do not go out on their own late at night it isn’t surprising that men are more likely to be victims of street crime at night. But if we were to measure “crime per risk taken” then the figures may be very different. Or they may not. We need careful studies by people who are not setting out to confirm a particular hypothesis.

But nevertheless although for a given individual whether or not there may be good reason to believe that there is a positive correlation between subjective belief in risk and fear, there is no reason for postulating an interpersonal correlation. People just have different personalities. Some are neurotic, some are oblivious. And there are many shades in between. On an aggregative ethic of fear reduction we may have to reduce small subjective risks for one group before addressing larger subjective risks for another. And all of this is independent of objective risk. But the main lesson is that this reinforces the claim that the irrationality argument fails to take into account that saving lives is not the only possible point of risk reduction policy.

3.2 Causation and blame

We have sketched out so far two main theories of risk regulation: risk reduction and fear reduction. This distinction is well known. Yet we are far from finished. There is another dimension to which we should pay attention.

418 JONATHAN WOLFF

Standard risk analysis begins with two concepts: hazard and probability. The only hazard we are concerned with here is death, and so the risks we have been concerned with are probabilities of death from particular causes. However it is vital to recognize that essentially the same type of hazard can have more than one possible cause. Take the example of death in a house fire. Some people die in house fires caused by electrical faults, caused, in turn through freak accidents – rodents gnawing through wires, for example. Or the fault could have been caused by negligent workmanship. Or through deliberate arson. The death in each case is equally gruesome, although not equally morally culpable.

Imagine that as a society we take a decision to reduce the number of deaths in house fires. Policies are proposed to combat each of the three causes: by means of regular safety checks; better training of workers; and better policing to track down and punish arsonists. In deciding which policy to adopt one possibility would be to carry out a cost-effectiveness analysis, working out which policy saves most lives for a given budget. Yet an alternative approach would be to argue that it is more important to eliminate some causes of house fires than others. This would be to make the judgment that some processes by which risks are created and sustained are worse, in some non-statistical sense, than others, and so should be a priority to eliminate even if this does not lead to the most cost-effective way of eliminating risk. Thus the hazard/probability analysis is too superficial. We must also take into account the process by which the hazard comes into being, or is sustained, or perhaps, permitted.6

It seems that, in general, people worry about some processes more than others (Baier 1986). For example it is widely documented that “man- made” hazards are regarded as in some sense “worse” than “natural” hazards. However there is more than one sense in which a hazard can be worse than another of the same objective magnitude. One is that it makes people more fearful. Another is that it generates greater moral concern or outrage. While these may often go together they need not. For example, I could be outraged at the existence of a risk that I do not even face.

In the light of this it might not be surprising if a society chose to eliminate morally blameworthy culpable behavior first. Morally culpable behavior comes in various forms. Roughly we can distinguish malice, recklessness, negligence, and incompetence. Malice is to set out a course of action with the deliberate aim of imposing harm or risks to people. Recklessness is to act knowing that it could cause harm or risk, but not taking this properly into account in deciding whether to act. Negligence is to fail to consider whether or not your action carries risks to others, when such risks were reasonably foreseeable. Incompetence, in this context, is to carry out a proper risk assessment and decide to take appropriate

6 For another perspective on the importance of cause in risk analysis, see Hopkins (2004).

RISK, FEAR, BLAME, SHAME AND THE REGULATION OF PUBLIC SAFETY 419

action, but fail to do so. We blame people and organizations where we feel they have violated some moral norm, and an extreme form of blame is outrage.

There are important distinctions here, and we may well feel differently about different types and levels of culpable behaviour, but to keep the discussion within manageable bounds I will consider only malice here. And it seems that we do have policies which give a priority to reducing hazards brought about by malice. After all, it is not obvious that the resources put into deterring, detecting, and punishing murderers, or preventing terrorist attacks, can be justified on a risk cost-benefit analysis valuing each saved life at £1 million.

Is it plausible that we should, as a priority, eliminate hazards caused by malice? I think that this would be a common view. But what explains it? One possibility is simply that we think very badly of malice, and take particular satisfaction in eliminating its effects, or, to put it differently, we find some actions moral outrageous and we find ourselves giving a high priority to removing sources, or potential sources, of outrage. I’m sure that this is at least part of the story, although not all of it. An alternative explanation appeals to the vital distinction between risk and uncertainty. Risk involves known hazards and probabilities, whereas uncertainty involves lack of knowledge, either of the precise nature of the hazard, or the probability of its occurrence, or both. For most people, in most of their life, they are faced with uncertainty, at least within a range, rather than risk in the technical sense. This puts us in quite a different situation, both practically and technically. For risk cost-benefit analysis assumes that we know the hazards and probabilities, or, at least, have a good basis for estimating them, or at least enough stability to apply some other methodological approach. Without this the analysis can’t even get started.

The relevance of this distinction is that it is not implausible that once malicious human beings threaten, we are moved into a world of uncertainty, not risk. And perhaps what in part explains any belief that we should give the rooting out of bad behavior special attention is the further belief that bad behavior places us under conditions not of known probability but uncertainty, and eliminating this uncertainty is the priority. With a few arsonists running around we cannot predict what is going to happen. And it is the same with killer sharks lurking in shallow waters, even though we don’t tend to hold them morally to account. This may indicate that part of the problem does indeed lie in uncertainty.

However, it seems highly likely that we find ourselves with two converging explanations in these cases: root out bad behavior and control uncertainty. Both are distinct from risk reduction, and may lead to irrational results in such terms. But, many will argue, so much the worse for risk reduction, as the sole aim of safety policy.

420 JONATHAN WOLFF

3.3 Reputation and shame

Alleged bad behavior, though, comes into the picture another way too. Often a firm will pay more for safety than is mandated by the regulations. Why? Sometimes a firm may want to take expensive safety measures because it thinks it is the right thing to do. Sometimes there is a commercial advantage in having a squeaky clean reputation for safety. But for either reason a firm may decide to spend more than the regulations require. Is this irrational? The irrationality argument is that the redistribution of resources across sectors may reduce overall risk. However, it is rare that it would be within the power of one firm to do this, as firms operate only within a restricted domain. All they can do is regulate their own area. Their own budgeting trades safety against other aspects of the quality of the product, together with prices and profits. Their choice is simply to spend more money or less on safety, knowing that they could well suffer commercial damage from any accident involving their own goods or services. Hence it is very likely that a “not on my watch” phenomenon will sometimes operate; we know that there are going to be accidents but we don’t want them here. Individual decisions may then lead to a very uneven provision of safety. Some firms may overspend while others underspend, depending on how prepared they are to risk harm and consequent reputational damage, as well as legal liability and possible bankruptcy. That is, different attitudes to being blamed for causing harm will lead firms to different attitudes to safety. And it could be that in a given industry no one can afford to be singled out as relatively dangerous, even when general standards are very high. Conceivably this is true of air transport, where, we already saw, it could be argued that far too much is spent on safety measures which have either only a negligible, or perhaps symbolic, effect. But an operator may feel compelled to match “best practice” since no operator can afford a reputation for being less safe than the competitors. This we could call the problem of clean hands – no one wants to be the site of where the harm takes place.

This problem goes all the way up. A safety regulator cannot, for example, tell firms to stop spending money on safety improvements but pass the money to the health service instead. And, understandably, it will want as few deaths as possible in the areas it regulates. Excessive media attention, and a reputation for poor safety, however undeserved, follows accidents. A concern for reputation may lead to apparently irrational over- provision and even over-regulation, in particular areas, relative to the goal of risk-reduction. Yet, once again, if the aimed-for goal is “reputational damage reduction” this turns out to be rational after all.

And it does not stop with the reputation of the safety regulator. In the aftermath of a train crash people often feel ashamed to be identified with a country where this sort of thing can happen. It is shocking to find

RISK, FEAR, BLAME, SHAME AND THE REGULATION OF PUBLIC SAFETY 421

that even in a modern, industrialized economy at many junctions the only thing stopping our trains crashing is the driver giving the correct response to a trackside light, and a bell. Admittedly the system is set up so that if a driver loses concentration the train will stop automatically, but there is still room for driver error. And if a train crashes through driver error, our shame that this can still happen in our country may overwhelm the thought that these things happen very rarely, and that for the last 50 years in the UK there probably has not been a week where more people died on the railways than on the roads.

This “shame” perspective can be based on presumed international comparison and a thought about how we must look to others. Once more, when one has got “into the skin” of this approach it may seem quite reasonable to want to take steps to reduce the potential for shame, even if they are expensive. Yet from a risk reduction perspective it is an absurd waste of money.

4. MODIFYING THE RISK ASSESSMENT PARADIGM

4.1 Perception of risk and societal concern

This issues I have mentioned – anxiety, malice, recklessness, negligence, incompetence, reputation, and shame – are not unknown to those who theorize and regulate risk. However risk management has struggled to work out how to incorporate them. Two leading approaches are what we can call the “perception of risk” framework and the “societal concern” framework.

The perception of risk framework, drawing especially on the work of Paul Slovic and associates (Slovic 2000), pays attention to how individuals perceive the seriousness of risks. So, for example, it is commonly noted that some risks seem to give special concern. Particularly important categories are those that are “dreaded,” such as the fear of cancer, and those that are outside the control of individuals, either in the sense that individuals have no influence over whether they are exposed to the risk or that there are no strategies they can personally adopt to mitigate their risk. Here traveling by car and traveling by air are an interesting comparison. Of course, whether or not one is exposed to the risk at all is generally a matter involving choice. But once the journey is under way a car driver has a measure of influence over subsequent events in a way that no air passenger has. Such utter reliance on others seems to create special concern.

Risk management needs to decide what to do about risks that give rise to special concern. One possibility, of course, would be simply to ignore special concern, which implicitly is what the irrationality argument recommends. But if one is unhappy with this approach something else is needed. The UK regulations deal with dread risks of cancer by doubling the

422 JONATHAN WOLFF

value of preventing a fatality from cancer.7 We can see how this modifies the standard risk cost-benefit analysis (RCBA). As we saw, in its simplest form RCBA derives an appropriate spending figure for risk reduction by means of a two-value formula of hazard (number of statistical deaths, each valued at £1 million) multiplied by probability reduction. The “dread” factor can then used as a multiplier, so, in effect, a third variable. Hence anything that the perception of risk framework wishes to include can be added as a multiplier (or indeed divider) of the result that would otherwise be derived from the RCBA. In this manner the standard paradigm can be used in a much more flexible way.

“Societal concern” is a different matter. This term answers to the need to generate a concept to capture the idea that society may have concerns which go beyond the sum of concern each individual has for his or her own life. How to specify this is fraught with difficulties, given that its key defining feature is a negative: not the sum of individual concern. Therefore what is to be included can be contested. The Health and Safety Executive say the following:

Societal concerns [are] the risks or threats from hazards which impact on society and which, if realised, could have adverse repercussions for the institutions responsible for putting in place the provisions and arrangements for protecting people, e.g. Parliament or the Government of the day. This type of concern is often associated with hazards that give rise to risks which, were they to materialise, could provoke a socio-political response, e.g. risk of events causing widespread or large scale detriment or the occurrence of multiple fatalities in a single event. Typical examples relate to nuclear power generation, railway travel, or the genetic modification of organisms. Societal concerns due to the occurrence of multiple fatalities in a single event is known as societal risk. Societal risk is therefore a subset of societal concerns. (12)

Clearly a number of different issues are brought in here. For example, societal concern seems to include both direct costs – “large-scale detriment” – and indirect costs such as loss of confidence in government and presumably the safety regulators too. This may have further direct costs – loss of business to better regulated countries, for example – or the costs may be less tangible, such as scorn or being the butt of jokes. While many analysts seem to agree that there is such a thing as societal concern, and that it should be taken into account, there seems a great puzzle about what it is, precisely, and how it could be taken into account. One attempt at least to do something has been the railway industry which has used a

7 “HSE takes the view that it is only in the case where death is caused by cancer that people are prepared to pay a premium for the benefit of preventing a fatality and has accordingly adopted a VPF twice that of the roads benchmark figure” (Health and Safety Executive 2001: 65).

RISK, FEAR, BLAME, SHAME AND THE REGULATION OF PUBLIC SAFETY 423

higher VPF for multiple fatality accidents than those that involve a single death. So, for example, one accident that kills four people is considered significantly worse than four that each kills one person. Again, this is to use societal concern as applying a multiplier to the standard VPF in the basic formula.8

Is it right or wrong to use an enhanced (or reduced) VPF to reflect varying individual concern (anxiety, dread) or varying societal concern? Here, I think, the correct thing to say that it is right to wish to modify the standard “hazard multiplied by probability” paradigm, but this is a crude way of doing so, especially as we do not have an agreed understanding of what the nature of societal concern is, and how it is engaged. Can anything better be envisaged?

Before we can settle this normative question – which is not in any case the task of this paper – we need first to understand how all the various attempts to understand and modify the analysis of risk can be brought together. Until we have an accurate understanding of the various factors in play it is hard to know what should properly be taken into account and how. So first we have to provide a descriptive model before a complete and compelling set of normative recommendations can be made. Yet before that we need to be as clear as we can about what the model is a model of: what precisely are we attempting to model? Here the best thing I can say is that we need an outline model of all the factors that affect human attitudes to risk, in the sense of influencing beliefs about how much we should do, pay, or sacrifice to mitigate risks. Many of these factors affect public perception to risk. Others, such as the “not on my watch” phenomenon – influence the attitude of those subjecting others to risk, or those who regulate risk. These can also influence public perception when there is an identification of some sort between members of the public and the organization in question, and so any failure of that organization can cause shame or embarrassment for the public. In the next section I will introduce a model of (part of) public concern, showing how fear, blame, and shame operate as factors in the public perception of risk, and in calls for its mitigation. For shorthand, I will call this a model of the “anatomy of risk.”

4.2 The anatomy of risk

The main lesson so far is, I believe, that the standard paradigm that risk is to be understood as hazard multiplied by probability is inadequate as

8 There is also the issue that some groups are especially vulnerable and so some weighting needs to be given to this. This may or may not fall under the heading of societal concern. Clearly it is a serious and important point. However it is beyond the scope of the current enquiry to consider this, as is the further, fundamental question of the degree to which public consultation is integral to the risk assessment and management process. I hope to address these issues in future work.

424 JONATHAN WOLFF

a way of modeling the actual concerns people have about risk, and the way in which those concerns translate into pressure to mitigate one risk rather than another. The “perception of risk” literature provides a way of incorporating some further features by means of allowing that fear may be out of proportion to objective risk, and this can be included as a multiplier or divider of risk. This is already reasonably well understood in the literature.9 However, as noted above “outrage,” or alternatively “blame” is distinct from “fear” and so needs to be incorporated as a second perception factor.

This does not yet incorporate all the issues included under the heading of “societal risk”, which I have assumed stands for those factors which can influence attitudes to risk, but go beyond each individual’s concern for the risks that he or she faces. Drawing on the examples and the analysis of Section 3 of this paper it appears that for analytical purposes the concept ‘societal concern’ is too vague to do any work, encompassing too many factors of quite different types. We do better to approach the analysis a different way. Let us begin by contrasting two rail accidents, the first caused by faulty track maintenance, the second by a car which bursts a tyre at speed, skids, and breaks through a fence onto the track, into the path of a train. The hazard in the two cases is for the purposes of analysis identical: loss of lives through derailment of a train. The probabilities may be hard to assess, but there is no reason to treat one as more probable than the other. If they are believed equally probable then again there is no reason to think that one will be more feared than the other. Yet the blame that will attach to the rail industry is likely to be very much higher in the case of faulty maintenance than in the case of the careless car driver. Indeed in the latter case there might even be sympathy for the rail industry.

If this is so then it seems that in this case at least blame attaches itself not to the hazard or the probability but to the cause of the hazard. Hence, it appears, the cause of the hazard must appear as an independent variable if we are to model public concerns about risk. Cause concerns how a hazard is created or sustained, and in consequence whether it can be viewed as a matter of culpable human action or inaction, especially the culpable action of those supposed to have a special responsibility.

One convenient way of understanding the various factors in play is to divide them into “primary variables” and “secondary variables.” On the present analysis the primary variables are three: cause, hazard, and probability. The secondary variables so far introduced are fear/dread and blame/outrage. They are called secondary because they take the primary variables as their object. In the standard cases fear attaches to hazard and probability – the “bigger” the risk, the more it is feared – whereas

9 For a review of some of the relevant literature, and an initial sketch of some of the ideas developed here, see Wolff (2002).

RISK, FEAR, BLAME, SHAME AND THE REGULATION OF PUBLIC SAFETY 425

blame or outrage attaches to cause, as illustrated. But note that each of fear and blame/outrage can take as their object each of cause, hazard, and probability. Outrage can attach to the hazard, independently of cause. If a hazard will involve many deaths, or deaths in a particularly frightening manner, this may create pressure to do more to mitigate the hazard, even from those who are not personally at risk and so have no fear for themselves or on behalf of family and friends. Here, then, it must be a sense of moral concern, rather than personal dread, that moves such people. Finally, outrage can attach to probability. Even if the cause and the hazard do not generate outrage, the fact that something is happening “too often” may do so. Note that it is not claimed here that everyone will react the same way, but rather that broad trends may be detectable.

It may be that the secondary variables are indeterminate in number, in that a wide variety of further responses may be possible. Importantly, though, they must include shame, as introduced in Section 3. Shame can only be present for those who feel either identified with, or partly responsible for, the regrettable event. This is the sense in which ordinary citizens can feel ashamed of the loutish behavior of the football hooligans who follow the national team, or ashamed at the racism of other members of their own family. Shame, then, seems to pre-suppose some form of identification. Shame, also, is likely to lead to the worry that reputation will also be damaged, which can then have further damaging effects.

Consequently, while those directly involved in the organization imposing risks may well feel shame if those risks are culpably imposed, it does appear that this shame can also spread to the public. It could do so because the organization at fault is something with which there is strong identification – perhaps if it is a nationalized industry. Or it may be because a regulator, who acts on our behalf, has not done a good job. Or it may simply be because we as citizens have voted in a government which has allowed certain things to happen, or prevented other things from happening. Hence even “natural disasters” can give rise to shame, where the thought may be “how did we allow things to get to a state where this could have happened?” Furthermore, like blame, shame can attach itself to each of cause, hazard, and probability.

Shame, then, may affect public attitudes to risk, and may intensify public calls that a scandalous situation should be redressed. But, as indicated, shame can be more private than this, affecting only those who are directly involved in allowing a risky process or event to take place. And, as also noted, there can be reputational and commercial factors also to include, and these will influence how an organization wishes to manage risk. It is easy to see how the model can be developed to include more secondary variables, such as these, and perhaps even tertiary variables, such as regret that an organization is the cause of shame in its citizens. The central point of this paper, however, is to argue for the inclusion of

426 JONATHAN WOLFF

cause, alongside hazard and probability, as a primary variable, if we are to understand how various attitudes to risk are formed and sustained.

One further remark about cause is needed. In so far as we are concerned with strategies to mitigate risk, cause appears relevant only when it engages the secondary variables of fear, and especially blame and shame. Where there are no such factors then there is no role for cause to play and the standard paradigm of hazard and probability appears sufficient to capture general concern, subject to issues about distribution, consultation, and radical uncertainty, which I have not discussed in this paper. In this respect cause differs from the other primary variables. However this is not an argument that cause is generally redundant, but rather an argument that other effects cannot be properly understood without paying attention to cause.

Some readers may be surprised that I have attempted this analysis without appeal to the idea of “altruistic preferences” or to the distinction between a “consumer perspective” and a “citizens perspective.” Both ideas are a useful reminder that distinctions need to be made but neither, in the end, takes us far enough. The idea of altruistic preferences has been used to model the thought that people care about other things than their own self- interest (see, e.g. Jones-Lee 1991, 1992). However using a single concept to attempt to capture the whole range of issues that fall under the heading of “not-entirely-self-interested concern” is problematic given their variety. A similar problem afflicts the idea of a “citizen” perspective.10 Insofar as a citizen is someone with moral beliefs, this concern is better captured under the idea of outrage or blame. Insofar as a citizen is a member of a collective entity which in some, possibly very indirect, way is responsible for the policies, including safety policies, adopted by his or her country, the idea of “shame” captures the relevant aspects. Hence the model adopted here is intended to provide a deeper analysis.

The model, then, points to a range of factors that currently influence judgment. However it is not obvious that there are no others, or that they should all be taken account of in the risk management process. For example, as we have seen, it is sometimes thought that perception and reputational effects should be managed by better public relations rather than by taking “skewed” safety decisions. Clearly when it is appropriate to attempt this is a vital question and it could be that practice ought properly to vary. So I should make clear that by setting out this model I am not assuming that the only way of managing the secondary variables is by spending money on safety measures. My point is only that we need a recognition of these factors and that it may turn out that the best way of dealing with them is, indeed, by spending money on safety measures.

10 The distinction between a “citizen” perspective and a “consumer” perspective is discussed in Wolff (2002).

RISK, FEAR, BLAME, SHAME AND THE REGULATION OF PUBLIC SAFETY 427

5. CONCLUSION

My argument, then, is that to arrive at a model of the anatomy of risk – an account of the factors we need to include in order to decide how to manage particular risks – attention must be given to cause, hazard, probability, fear, blame and shame, and we must acknowledge that this list is unlikely to be complete. However the model provides considerable insight into the factors which influence public perception. Once we have this model we might decide that normatively it is nonsense, and all that we should be concerned with is saving lives in a cost-effective fashion, which is what the irrationality argument assumes. Here I take no view on this question. The preliminary task is to understand the factors which affect how generally we think about risk management. With a descriptively more accurate picture of the factors we actually take into account we can sensibly debate how much should be included in the normative framework of risk management, and why.

REFERENCES

Adams, J. 2001. Risk. Routledge Baier, A. 1986. Poisoning the wells. In Values at Risk, ed. D. MacLean. Rowman and Allenheld.

49–74 Burgess-Jackson, K. 1994. Justice and the distribution of fear. Southern Journal of Philosophy

32: 367–91 Commission for Integrated Transport. 2004. Rail safety: revision of Factsheet 10.

URL=http://www.cfit.gov.uk/research/railsafety/03.htm Evans, A. W. 2005. Fatal train accidents on Britain’s main line railways: end of 2004 analysis.

URL=http://www.cts.cv.ic.ac.uk/html/ResearchActivities/publicationDetails.asp? PublicationID=465

Health and Safety Executive. 2001. Reducing risks protecting people. HSE Books. URL=http://www.hse.gov.uk/risk/theory/r2p2.htm

Hopkins, A. 2004. Safety, culture and risk. CCH Books Jones-Lee, M. W. 1991. Altruism and the value of other people’s safety. Journal of Risk and

Uncertainty 4: 213–19 Jones-Lee, M. W. 1992. Paternalistic altruism and the value of statistical life. Economic Journal

102: 80–90 Jones-Lee, M. et al. 1999. On the contingent valuation of safety and the safety of contingent

valuation: Part 2 – The CV/SG “chained” approach. Journal of Risk and Uncertainty 17: 187–213

Lomborg, B. 2001. The skeptical environmentalist. Cambridge University Press Lomborg, B. 2003. Paper given to Spiked conference: panic attack, London, 9 May Pidgeon, N., Kasperson, R. E., and Slovic, P. (eds.) 2003. The social amplification of risk.

Cambridge University Press Posner, R. 2004. Catastrophe, risk and response. Oxford University Press Schelling, T. C. 1984. The life you save may be your own. In his Choice and consequence, 113–46.

Harvard University Press Slovic, P. 2000. The perception of risk. Earthscan Sunstein, C. 2002. Risk and reason. Cambridge University Press Wolff, J. 2002. Railway safety and the ethics of the tolerability of risk. Rail Standards and

Safety Board. URL=http://www.rssb.co.uk/pdf/reports/research/railway safety and the ethics of the tolerability of risk.pdf

__MACOSX/._risk+fear+blame+public+safety.pdf

risk+precautionary+princ.pdf

Journal of Medicine and Philosophy 2004, Vol. 29, No. 3, pp. 301–312

The Precautionary Principle: A Dialectical Reconsideration

H. Tristram Engelhardt, Jr., and Fabrice Jotterand Rice University, Houston, TX, USA

ABSTRACT

This essay examines an overlooked element of the precautionary principle: a prudent assessment of the long-range or remote catastrophes possibly associated with technological development must include the catastrophes that may take place because of the absence of such technologies. In short, this brief essay attempts to turn the precautionary principle on its head by arguing that, (1) if the long-term survival of any life form is precarious, and if the survival of the current human population is particularly precarious, especially given contemporary urban population densities, and (2) if technological innovation and progress are necessary in order rapidly to adapt humans to meet environmental threats that would otherwise be catastrophic on a large scale (e.g., pandemics of highly lethal diseases), then (3) the development of biomedical technologies in many forms, but in particular including human germ-line genetic engineering, may be required by the precautionary principle, given the prospect of the obliteration of humans in the absence of such enhanced biotechnology. The precautionary principle thus properly understood requires an ethos that should generally support technological innovation, at least in particular areas of biotechnology.

Keywords: precautionary principle, public health, risk assessment, technological development

I. PRUDENT RISK-TAKING

The so-called precautionary principle raises a cluster of questions about how

prudently to engage in risk-taking. All human activities involve risks. The

development of new technologies is no exception. However, given a not-

implausible account of the human situation, the unavailability of at least some

biomedical technologies may itself count as a risk to continued human

survival. This essay will examine an overlooked element of the precautionary

Address correspondence to: H. Tristram Engelhardt, Jr., Ph.D., M.D., Department of Philosophy, MS 14, Rice University, Houston, TX 77005, USA. E-mail: [email protected]

10.1080/03605310490500518$16.00 # Taylor & Francis Ltd.

principle: a prudent assessment of the long-range or remote catastrophes

possibly associated with technological development must include the

catastrophes that may take place because of the absence of such technologies.

In short, this brief essay will attempt to turn the precautionary principle on its

head by arguing that, (1) if the long-term survival of any life form is

precarious, and if the survival of the current human population is particularly

precarious, especially given contemporary urban population densities, and (2)

if technological innovation and progress are necessary in order rapidly to

adapt humans to meet environmental threats that would otherwise be

catastrophic on a large scale (e.g., pandemics of highly lethal diseases), then

(3) the development of biomedical technologies in many forms (some, such as

human reproductive cloning or embryo research may be prohibited on moral

grounds [Engelhardt, 2000]), but in particular including human germ-line

genetic engineering, may be required by the precautionary principle, given the

prospect of the obliteration of humans in the absence of such enhanced

biotechnology. The precautionary principle thus properly understood requires

an ethos that should generally support technological innovation, at least in

particular areas of biotechnology.

II. PUTTING THE PRECAUTIONARY PRINCIPLE

IN CONTEXT

There are a number of difficulties in making prudent assessments of risk. To

begin with, intuitions vary widely about how to compare risks appropriately.

There are some who regard a one-hour commercial flight with greater

apprehension than a four-hour automobile journey to the same city. This is the

case even though in general the risks are greater from the latter than the

former. In part, this perception is grounded in the difference in psycho-social

impact of learning about the death of 200 passengers in an airline crash versus

learning of 200 automobile accidents in a year’s time, each involving one

fatality. The same number of people dying or being disabled at the same time

usually has a more dramatic psycho-social impact than the same number being

disabled or dying over a more extended period of time. So, too, were it the

case that the number of individuals likely to be disabled or killed by nuclear

power plant accidents were no greater than the number of persons likely to be

disabled or killed by the generation of the same amount of electric power from

the use of fossil fuels, the socially disruptive character of all of the deaths

302 H. TRISTRAM ENGELHARDT, JR., & FABRICE JOTTERAND

happening at once appears to give this mass tragedy a weight greater than the

sum of all the individual tragedies.1 Intuitions that favor giving greater weight

to concerns regarding airplane crashes and nuclear power plant accidents may

contribute to the intuitions that are invoked to support the precautionary

principle.2

In this essay, the precautionary principle is understood as the rule that one

should never engage in a technological development or application unless it

can be shown that this will not lead to large-scale disasters or catastrophes.3

The possibility of a large-scale disaster or catastrophe is regarded as sufficient

to prohibit the application of new technologies that offer considerable benefit

to humans. In this sense, the precautionary principle is a variation of a prin-

ciple of risk-aversiveness, so that one takes maximal regard of possible large-

scale or catastrophic disasters, however remote and despite the benefits that

might accrue from the technology. This understanding of the precautionary

principle would constrain one to accept the likelihood of a number of deaths in

order to avoid the remote possibility of even greater catastrophes.

For example, the precautionary principle has been invoked to prohibit the

introduction of genetically modified organisms, until one can with a very high

degree of certainty rule out the possibility of catastrophic outcomes. Those

who embrace the precautionary principle would accept the starvation of

millions in third-world countries who could be fed by genetically modified

grains, rather than assume a very remote and very unlikely interruption of the

ecological balance as a result of unexpected or unforeseen genetic effects as a

result of genetically modified organisms.

Although the availability of genetically modified foodstuffs might aid the

starving, this very important good would be seen to be outweighed by a

possible, albeit unlikely, ecological catastrophe. New technologies are thus

held to be guilty until proven innocent (Saunders, 2000). Marc Moreno who

reports that according to a panel of food policy experts, the ban – on the basis

of the precautionary principle – of genetically modified food (GM) has no

scientific ground and causes starvation in the developing world (Moreno,

2002). In this vein, Goklany acknowledges that in order to meet food demands

additional deforestation of millions of hectares will be required. In 1997,

according to the Food and Agricultural Organization, it was estimated that

already 1,510 million hectares were devoted to cropland and by 2050 an

additional 1,600 million hectares of habitat land would be lost (FAO, 2000).

He also notes that an annual increase of productivity of 2% (through the use of

generically modified crops) would be translated into at least 422 million

THE PRECAUTIONARY PRINCIPLE: A DIALECTICAL RECONSIDERATION 303

hectares of cropland currently under plow that could be returned to nature or

made available for habitat or other human uses, thus increasing environmental

benefits (Goklany, 2001, pp. 30–32). He therefore concludes that, considering

that the benefits related to the use of genetically modified crops outweigh the

risks, biotechnology in agriculture can provide a means for addressing the

problem of malnutrition around the globe without necessarily neglecting

the environment. As he points out ‘‘the rewards of GM crops greatly outweigh

their risks. Although it would be a mistake to go full steam ahead on GM

crops, it would be a bigger mistake to stop them in their tracks. The wisest

policy would be to go as fast as possible while keeping a sharp lookout, and

staying on the track to improvements in human and environmental well-

being’’ (Goklany, 2001, p. 56).4 To proceed in this fashion would require

reconsidering the implications of the precautionary principle as usually

interpreted in order to take into account the potential damaging consequences

of not promoting scientific and technological development.

Because of the difficulty of proving that new technologies will not involve

unanticipated catastrophic outcomes, the precautionary principle if interpreted

strictly, as shown from the example of policy responses regarding food from

genetically modified crops, would seem to place an unjustifiable burden on all

technological progress. It would not only appear to forbid anything but the most

gradual introduction of most new technologies, but also give equal grounds for

the suspension of technological interventions for which there has not been ample

time to assess unforeseen risks.5 For instance, one might hypothesize that a wide

range of current pharmaceutical agents may carry with them unforeseen

consequences for the development of senile dementia, etc. With a sufficiently

active imagination one could bring much of contemporary biotechnology under

suspicion without a ready ability to lift the cloud of uncertainty.

The concern to give proper weight to possible catastrophic outcomes is

further augmented by discounting particular benefits, especially possible eco-

nomic benefits. Among many of the proponents of the precautionary principles,

there is a view either that it is improper to give any weight to economic benefits

or that the importance of such benefits has been improperly inflated. Nancy

Myers, for instance, claims that the World Trade Organization and the North

American Free Trade Agreement ‘‘institutionalized . . . the ascendancy of

commerce over environmental and public health concerns’’ and hence cost-

benefit assessments, it is argued, dictate that products or technological

innovations outweigh the costs of possible environmental harms (Myers,

2002, p. 214). If all consideration of economic benefits were removed from

304 H. TRISTRAM ENGELHARDT, JR., & FABRICE JOTTERAND

cost-benefit calculations, a considerable burden would have been placed on the

development of promising new technologies. This criticism of the weight to be

given to economic benefits opens the larger issue of how to compare different

genre of benefits and harms.

Finally, one must note that the precautionary principle is often interpreted

so as to give equal if not greater weight to concerns with the environment in

and of itself, not simply as harms to the environment may have indirect

impacts on human welfare. Here the question is not simply of comparing

benefits and harms, but the question of whose harms and benefits should be

compared and in what way. That is, how is one to compare the possibility of

harm to animals, ecosystems, and the environment with the possibility of

harms and benefits to humans? This weighting of the environment, especially

ecosystems in and of themselves, is noted by Alston Chase who asserts that the

precautionary principle reflects concerns regarding benefits and harms that

are biocentric6 rather than humanistic or human-centered. As he puts it,

‘‘biocentrism is the fundamental value conveyed in most treaties or protocols

promoting the Precautionary Principle’’ (Chase, 1997, p. 5). In these terms

one can justify the starvation of millions of people for the sake of the well-

being of the ecosystem.

The assessment of risks to the environment requires an account of how to

compare harms and benefits to humans as well as to other living organisms and

the environment generally. Such comparisons would require a complex

account differentiating diverse benefits and harms as these have impact on

humans, animals, ecosystems, and the environment in itself. Such rankings of

goods and harms fall beyond a factual description of the consequences of

particular technological interventions. It requires choosing one among a

number of competing moral visions. This circumstance is stressed by Joe

Thornton in his assessment of policies pertaining to environmental and health

issues. He notes the obvious: the assessment of risk presupposes endorsing a

particular vision that ranks harms and benefits with respect to interests in

human versus environmental flourishing. In his volume, Pandora’s Poison:

Chlorine, Health, and a New Environment Strategy, he notes that ‘‘I do not

claim balance or objectivity, because these are neither appropriate nor possible

in this kind of effort’’ (Thornton, 2000, p. ix). Thornton recognizes the

complex constitution of controversies involving a heavy political and moral

overlay. To sort out such controversies, one needs to look with care at the

geography of the different influences, as well as endorse a particular approach

to weighting harms and benefits (Engelhardt & Caplan, 1987).

THE PRECAUTIONARY PRINCIPLE: A DIALECTICAL RECONSIDERATION 305

However one sorts out the proper assessment of harms and benefits, the

cultural force of the precautionary principle would seem to place the burden of

suspicion on technological innovation and progress, in that all innovative

technological interventions carry with them an unassessable prospect of an

unanticipated, large-scale, catastrophic side effect.7 This conclusion would

seem to follow, given intuitions that give a greater weight to possible significant

catastrophic outcomes over equal but less catastrophic costs in human lives and

suffering. This conclusion is further fortified by discounting economic benefits

and adding a biocentric accent to the calculation of benefits and harms. All of

this seems to lead to regarding the precautionary principle as hostile to

biotechnological progress. This conclusion will now be brought into question.

III. THE NEED FOR RAPID RE-ADAPTATION OF HUMANS

TO AN EVER-CHANGING AND OFTEN

THREATENING ENVIRONMENT

Without addressing the issue of how to compare harms and benefits, one can

bring into question the putative conclusion that the precautionary principle

will under all circumstances place a burden against technological progress.

The arguments developed below show that the precautionary principle, if

properly understood, should support at least certain areas of biotechnological

innovation, rather than constitute an impediment. In what follows, the focus is

given to human welfare. With a few changes, the focus could be brought to

bear on ecosystems as well. As developed, the argument does take into

account concerns with the ecosystems insofar as they would constitute a threat

to the long-range survival of the human species.

The long-range survival of humans depends on the capacity of humans to

withstand threats from an environment often significantly hostile to the

survival of humans, indeed, to the long-range survival of any species of

organisms. Among those threatening elements are new viruses, new variations

of old viruses, and bacteria that have become altered so as to be drug-resistant

and/or toxic to humans in new ways. Similar threats to human survival can be

envisaged in terms of viruses and other life forms that might threaten the

human food supply and the environment. Given the network of rapid global

travel, quarantine over any significant period of time is likely to be ineffective

without a near total paralysis of international trade (see for instance the

analysis of Bailey (2002) in relation to plant biotechnology). From Ebola and

306 H. TRISTRAM ENGELHARDT, JR., & FABRICE JOTTERAND

AIDS to new forms of influenza and SARS, recent history has provided

numerous possibilities for environmental confrontations that could lead to

large-scale, indeed catastrophic, loss of human life.

The protean possibilities for future threats of a large-scale, indeed

catastrophic magnitude, given a reasonable interpretation of the precautionary

principle, would require the vigorous development of a biotechnology suf-

ficient to produce not simply new antimicrobials and new vaccines, but able

genetically to modify humans as well as the organisms that serve as foodstuffs

for humans. Given the prospect of a catastrophic development of a hyper-

virulent microbe threatening either humans or their foodstuffs, the human

ability both to kill such threatening microbes as well as rapidly to readapt

humans and their foodstuffs to resist such threats would be obligatory under

the precautionary principle. The precautionary principle should require

vigorously supporting technological and scientific progress.8

This result is an important acknowledgement. There are two sides or

dimensions of the precautionary principle. On the one hand, the precautionary

principle requires considering the untoward consequences of new tech-

nological innovation. On the other hand, the precautionary principle requires

considering the untoward consequences of not supporting technological

innovation. In short, one must not only fear catastrophes that will flow from a

technology, but also the catastrophes that will flow from its absence.

The question then is how to compare the two sides or dimensions of the

precautionary principle. Which set of unforeseen, large-scale, and cata-

strophic consequences should be given greater weight and why? Possible

catastrophes frame technology or frame its absence. To begin with, there are

factual considerations. Given the recorded history of disastrous epidemics

when communication among humans was less global than today, one might

very well have grounds to tilt the balance in favor of giving greater weight to

the unforeseen consequences likely to flow from the failure to accelerate

biotechnological progress and encourage biotechnological innovation. Should

such reflections on the history of the hostility of environments to organisms in

general and to humans in particular be credible, then there would be a strong

moral argument grounded in the precautionary principle in favor of sustaining

a bias in favor of biotechnological progress and innovation. In this

circumstance, the precautionary principle would need to be reinterpreted in

order to be understood as substantively technology-friendly.

The greater the plausibility of bioenvironmental threats, the greater the

obligation will be to encourage the development of an appreciation of

THE PRECAUTIONARY PRINCIPLE: A DIALECTICAL RECONSIDERATION 307

biomedicine and the biomedical technologies as core to the human enterprise.

When the precautionary principle is combined with any moral vision that

gives weight to obligations to future generations, then the biomedical

technologies will be core to the human endeavor of ensuring the survival of the

human species. In short, a more balanced appreciation of the precautionary

principle should transform the principle from being central to an anti-

technological ethos to a principle that when rightly understood is a cardinal

foundation of an ethos supportive of biotechnological innovation. In addition,

insofar as such innovation turns out as a fact of the matter to be enhanced by

larger-than-usual profit-margins in the pharmaceutical and medical device

industries, then one will wish to avoid forms of cost containment, tort liability,

and tax policies that encumber profitability in this industry. In short, a more

balanced consideration of the principle may shed important light on a broader

range of risks associated with biotechnology, namely, those connected with a

failure wholeheartedly to support it.

IV. THE ARGUMENT FROM IGNORANCE

GOES BOTH WAYS

At the very least, this dialectical exploration of the precautionary principle

shows its other side and excluded dimension, thus indicating one of the major

difficulties involved with arguments from ignorance. Evenly applied, the

precautionary principle invites us to give at least as much weight to the

catastrophes we may face from not developing a certain technology as from

developing the technology. Were one of the opinion that the historical record

of devastating epidemics and other environmental changes was not sufficient

to tip the balance vigorously in favor of technological innovation on the basis

of the precautionary principle, then both appeals to ignorance would simply

cancel each other out. In that case, the precautionary principle would be

devoid of force.

It must be acknowledged that this analysis of the precautionary principle

focuses on its application only in areas where it would bear on technologies

whose unavailability could foreseeably lead to catastrophic human harms.

Thus, there may be some (surely not these authors) who might be of the view

that the precautionary principle should preclude the use of cell phones until

the magnetic waves involved had been tested on primates for a sufficiently

long period so as to assess the possibility of long-range adverse outcomes.

308 H. TRISTRAM ENGELHARDT, JR., & FABRICE JOTTERAND

Were the precautionary principle to be employed to block the further use of

cell phones, this concern might not be as easily outweighed by the health risks

from the unavailability of cell phones.

This brief reflection leaves us with two conclusions, at least for some. First,

a balanced appreciation of the precautionary principle leads to any

unanticipated result: rather than setting cautionary blocks to biotechnological

development, the principle should, given a number of plausible empirical

assumptions, encourage biotechnological development. Second, if the factual

assumptions necessary to tip the balance in favor of the precautionary

principle as supporting biotechnological innovation are brought into question,

then the default position will be to deprive the principle of any credible force,

at least in a significant range of biotechnologies. Either the precautionary

principle means something that most have not anticipated (i.e., it is

technology-friendly), or, at least in many areas, it is rendered void by the

possibility of contrary catastrophic possibilities.9

NOTES

1. For a critical appraisal of risk assessment with regard to nuclear power, see MacLean (1987).

2. The precautionary principle has roots in the German Vorsorgeprinzip (‘foresight-planning’), which constitutes, according to Julian Morris, ‘‘a founding principle of German environ- mental policy in the mid-1970s’’ (Morris, 2000, p. 1). Morris, however, points out that in the United States the precautionary principle has implicitly been used since the 1950s, especially by political conservatives groups that opposed the fluoridation of water. The argument was two-fold: first it was argued that fluoride was used as rat poison and second this involuntary mass medication was ‘‘a step on the road to socialism.’’ (Morris, 2000, p. 2). In the 1960’s, the same precautionary principle was implicitly used by left-wing activists to oppose nuclear power. Finally, in the 1970s social scientists referred to the principle in a more general framework. For an overview of the definition and origin of principle see Morris (2000, pp. 1–21).

3. The precautionary principle can be found under many forms in different treaties such as the Montreal Protocol, the Convention on Biological Diversity, the Helsinki Convention on Marine Protection in the Baltic, the Treaty on The Precautionary Principle by the European Union, the Biosafety Protocol and the Treaty on Persistent Organic Pollutants. One of its most influential statements is found in the Rio Declaration on Environment and Development (1992). Principle 15 requires that ‘‘in order to protect the environment, the precautionary approach shall be widely applied by States according to their capabilities. Where there are threats of serious or irreversible damage, lack of full scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental degradation.’’ Avail- able [On-line]: http://sedac.ciesin.org/pidb/texts/rio.declaration.1992.html

THE PRECAUTIONARY PRINCIPLE: A DIALECTICAL RECONSIDERATION 309

Another influential statement, The Wingspread Statement, which followed a gathering at Wingspread in 1998 (headquarters of the Johnson Foundation in Racine, Wisconsin), likewise presents a strong endorsement of the precautionary principle in public health and environment decision-making. It states ‘‘while we realize that human activities may involve hazards, people must proceed more carefully than has been the case in recent history. Corporations, government entities, organizations, communities, scientists and other individ- uals must adopt a precautionary approach to all human endeavors. Therefore, it is necessary to implement the Precautionary Principle: When an activity raises threats of harm to human health or the environment, precautionary measures should be taken even if some cause and effect relationships are not fully established scientifically. In this context the proponent of an activity, rather than the public, should bear the burden of proof. The process of applying the Precautionary Principle must be open, informed and democratic and must include poten- tially affected parties. It must also involve an examination of the full range of alternatives, including no action.’’ A full version of the Wingspread Statement is available on-line: http:// www.sehn.org/state.html

4. For further details, see Goklany’s chapter on ‘‘The Risks and Rewards of Genetically Modified Crops’’ (2001, pp. 29–56).

5. Risk assessment remains at the core of the discussions surrounding the precautionary principle. Some proponents of the principle, Peter Saunders and Mae-Wan Ho, for instance, argue that it is based on good science and therefore ‘‘a compelling case for the application of the precautionary principle’’ can be made that the precautionary principle can be used to protect the environment and human health (Saunders & Ho, 2003). Interestingly, however, in the first edition of the handbook on the precautionary principle written for the Science and Environmental Health Network, the authors acknowledge that ‘‘risk assessment and other ‘sound science’ approaches to decision-making are highly reliant on policy and scientific assumptions, which are frequently unscientific or subjective’’ (Tickner, Raffensperger, & Myers, p. 14). Chase likewise notes that the precautionary principle is unreliable as ‘‘a means of making quantitative assessments of alternative courses of action’’ and ‘‘does not provide . . . a calculus by which to weigh and compare economic costs against ecologic benefits, or vice versa’’ (Chase, 1997, p. 5).

6. Biocentrism involves ‘‘The belief that . . . the Biosphere or ecosystem takes precedence over the well being of humanity’’ (Chase, 1997, p. 5).

7. A good example of some of the values underlying support for the precautionary principle is found in a document issued in September 2001 by the Canadian government (Government of Canada, 2001). Although the document does not constitute an official position of the govern- ment (it is only a discussion paper), it reflects cultural assumptions concerning technological innovations and progress widespread in Canada. The document raises the question of the need for regulations at the national and international level, and it emphasizes risk management, especially in relation to new technologies (such as biotechnology, for instance). These two concerns characterize public suspicion of technological development while, at the same time, paradoxically, desiring the benefits of new technologies: ‘‘Public opinion surveys show that Canadians want to reap the benefits of change (e.g., biotechnology), but they also want their governments to protect them from the risks. As a result, governments are often called upon to balance new or emerging risks and potential opportunities [technological innovations and progress], and to manage issues where there is significant scientific uncertainty. The decisions they make can have profound effects on societies, trade and economies’’ (Government of Canada, 2001, p. 1). For a general assessment of the document see Lee and Barrett (2002).

310 H. TRISTRAM ENGELHARDT, JR., & FABRICE JOTTERAND

8. Further reflections are needed with respect to the role of risk-taking in scientific progress. This issue goes beyond the scope of this short note. Space exploration illustrates the difficulty in evaluating the exact nature of proper risk assessment. Space exploration may at first blush appear to offer only ‘‘limited’’ benefits in biotechnology at the price of potential dangers for the people involved in it and the rest of the population (the risk of the crash of a space shuttle in inhabited areas is real, as the breaking apart of Columbia as shown). An application of the precautionary principle as usually understood, considering the calculus of risks and benefits, would simply stop any such undertakings.

9. For a critical assessment of the precautionary principle see Morris’ Rethinking Risk and the Precautionary Principle (2000). The contributors to this volume address a variety of issues in relation to precautionary thinking and conclude that the principle has become an arbitrary imposition of regulations that are mostly counterproductive. In other words, the precau- tionary principle appears to be a rhetorical device unable to provide concrete guidance but with untoward consequences. As Chauncey Starr remarks ‘‘the precautionary principle exists only as a rhetorical statement; it provides no useful input to decision making. Expert opinions should be sought, but be recognized as conservatively biased. The search for science-based guidance is commendable, but is rarely achievable’’ (2003, p. 3).

REFERENCES

Bailey, R. (2002). The looming trade war over the plant biotechnology. Center for Trade Policy Studies, 18, 1–17 [On-line]. Available: http://www.freetrade.org/pubs/pas/tpa-018.pdf

Chase, A. (1997). Some cautionary remarks about the precautionary principle. ‘Countdown to Kyoto’: The consequences of the mandatory global carbon dioxide emissions reductions, Australian APEC Study Center, Canberra, 19–21 August 1997.

Engelhardt, H.T., Jr., & Caplan, A.L. (1987). Scientific controversies. New York: Cambridge University Press.

Engelhardt, H.T., Jr. (2000). The Foundations of Christian Bioethics. Lisse: Swets & Zeitlinger.

Food and Agricultural Organization. (2000). FAOSTAT Database. Goklany, I.M. (2001). The precautionary principle: A critical appraisal of environmental risk

assessment. Washington, DC: Cato Institute. Government of Canada. (2001). A Canadian perspective on the precautionary approach/

principle: Proposed guiding principles. September 2001. [On-line]. Available: http:// www.pco-bcp.gc.ca/raoics-srdc/docs/Precaution/Booklet/booklet-e-allfonts.pdf

Lee, S., & Barrett, K. (2002). Comments on: A Canadian perspective on the precautionary approach/principle discussion document. University of Victoria & Environmental Health Network, March 28, 2002. [On-line]. Available: http://www.sehn.org/canpre.html

MacLean, D. (1987). Understanding the nuclear power controversy. In: H.T. Engelhardt, Jr. and A.L. Caplan (Eds.), Scientific controversies (pp. 567–582). New York: Cambridge University Press.

Moreno, M. (2002). Ban on genetically modified crops causing starvation, food experts say. CNSNews.com, September 26, 2002.

Morris, J. (2000). Defining the precautionary principle. In: Julian Morris (Ed.), Rethinking risk and the precautionary principle. Oxford: Butterworth-Heinemann.

THE PRECAUTIONARY PRINCIPLE: A DIALECTICAL RECONSIDERATION 311

Myers, N. (2002). The precautionary principle puts values first. Bulletin of Science, Technology and Society, 22, 210–219.

Saunders, P. (2000). Use and abuse of the precautionary principle. ISIS submission to US Advisory Committee on International Economic Policy (ACIEP) Biotech. Working Group, 13 July 2000.

Saunders, P., & Ho, M.-W. (2003). The precautionary principle is science-based [Online]. ISIS Report, 4. Available: http://www.i-sis.org.uk/sapp.php

Starr, C. (2003). The precautionary principle versus risk analysis. Risk Analysis, 23, 1–3. Thornton, J. (2000). Pandora’s poison: Chlorine, health, and a new environment strategy.

Cambridge, MA: MIT Press. Tickner, J., Raffensperger, C., & Myers, N. (2000). The precautionary principle in action,

(1st ed.). Windsor: Science and Environmental Health Network.

312 H. TRISTRAM ENGELHARDT, JR., & FABRICE JOTTERAND

Engineering Ethics

advancing+risk+assessment+brief.pdf

__MACOSX/._advancing+risk+assessment+brief.pdf

advancing+risk+assessment+full+report.pdf

__MACOSX/._advancing+risk+assessment+full+report.pdf

choosing+pleasures+risk.doc

__MACOSX/._choosing+pleasures+risk.doc

env+risk+yucca+mt.pdf

__MACOSX/._env+risk+yucca+mt.pdf

fecundity+as+basis+for+risk+assessment+in+mulloscs.pdf

__MACOSX/._fecundity+as+basis+for+risk+assessment+in+mulloscs.pdf

phil+scrutiny+of+evidence+of+risk.pdf

__MACOSX/._phil+scrutiny+of+evidence+of+risk.pdf

risk+201107_OMBdraft-OzoneRIA.pdf

__MACOSX/._risk+201107_OMBdraft-OzoneRIA.pdf

risk+fear+blame+public+safety.pdf

__MACOSX/._risk+fear+blame+public+safety.pdf

risk+precautionary+princ.pdf

__MACOSX/._risk+precautionary+princ.pdf