Environmental health policy decisions: the role of uncertainty in economic analysis.

Introduction

Environmental health professionals will be asked increasingly to participate in accounting for how dollars are spent in the areas of human health and environmental protection. What are the net benefits, for example, of sanitation? Are case management methods cost-effective strategies for helping lead-burdened children? This kind of accounting will be both inevitable and vital in regulatory response. As a result, environmental health professionals may find that, increasingly, they need to secure a place for the human health aspects of environmental economics and management. For example, reducing the risk of human exposure to benzene was clearly articulated as an intended benefit of controlling evaporative refueling emissions, but it was sometimes hard to keep this goal in sight during the ensuing debate about the relative economic advantages of controlling emissions by modifying the gasoline pump and controlling them through the design and manufacture of the automobile. This trend will continue as debates over economic development intensify and the definition of social welfare broadens.

These observations are informed on the one hand by the link between economic and quantitative risk analysis and, on the other hand, by the critical role to be played by the National Environmental Health Association in ensuring a position for public health concerns in economic decisions about development and environmental policy. As part of a larger discussion of critical issues for the 21st century, the importance of this link emerged in a thoughtful series of articles by Walker and Davis et al. on the future of environmental health (1-4). This article follows that train of thought while noting that regulatory reform will increasingly call for more economic analysis in decisions about environmental health policy. For deeply felt policies, the difficulties of economic analysis will no doubt be a lightning rod for much heated controversy. For this reason alone, informed participation of environmental health professionals is critical to the debate.

This article addresses the role of irreversibility and uncertainty in environmental health policy decisions. The U.S. Environmental Protection Agency's (U.S. EPA's) [PM.sub.2.5] rule for regulating emissions of small pollutant particles provides one example of an environmental policy decision that involves irreversibility and uncertainty. The presence of irreversibility and uncertainty gives value to flexible policy responses, including optimal timing of policy and incremental strategies, the success of which ultimately depends on the contributions of all environmental health professionals. So too for policies emphasizing information and learning, carrying still further lessons for the statistical design and analysis of continuing research initiatives and quantitative analyses of human health risks. Finally, a proper benefit-cost analysis of health and environmental policy reserves a place for the value of lives saved, reduced morbidity, and other benefits of policy. There will nevertheless be no substitute for the ready .voices and informed advocacy of public health officials.

Economic Analysis in Environmental Health Regulation

Ideally, benefit-cost analysis provides an analytical framework for weighing in economic terms the trade-offs involved in policies that affect the environment and human health. While this kind of analysis is not an omnibus tool, it can be used appropriately to set priorities, rank and select alternatives, and evaluate performance. Amidst growing national concerns about the costs, reach, and effectiveness of environmental, health, and safety regulations, a group of prominent economists met recently to discuss the issue under the auspices of the American Enterprise Institute, the Annapolis Center, and Resources for the Future. In light of reforms favoring greater reliance on economic analysis in policy decisions, there emerged from these discussions a set of principles for guiding and improving quality in the use of benefit-cost analysis in environmental, health, and safety regulations (5,6).

In summarizing their statement of principles, the authors proclaim that decision makers should be encouraged to use economic analysis to reach decisions and set regulatory priorities rather than be precluded from doing so. Such analyses should be required for all major regulatory decisions, should be subject to peer review, and should be conducted according to an established set of guidelines. In cases in which costs and benefits cannot be quantified, sound decisions can still be reached with the help of qualitative descriptions of the advantages and disadvantages of policy alternatives. In all of these recommendations and throughout the discussion, one important theme recurred: the ubiquity of uncertainty [TABULAR DATA FOR TABLE 1 OMITTED] and thus, implicitly, the role of statistical inference.

Benefit-cost analysis in environmental health regulation involves a complex aggregate of economic estimates of costs and benefits, scientific estimates of environmental impact, and biomedical estimates of risks to human health. The quality of a benefit-cost analysis and, ultimately, the design of environmental health policy depend on a host of factors, including flexibility, uncertainty, and economic assumptions about the value of reducing morbidity risk and the value of other health-related improvements. Theoretically, the analysis of these factors falls within the purview of the field of environmental and resource valuation, as discussed by Freeman (7). From a practical perspective, however, contributions from environmental health professionals, including biostatisticians, chemists, physicians, economists, and risk and public policy analysts, are critical.

Irreversibility and Uncertainty in Economic Analysis

Recent examples of this kind of cross-disciplinary effort can be found in the pages of this journal in the work of El-Gazzar and Marth, Robbins and McSwane, and Ballas et al. (8-10). Their work raises the following questions: How do such analyses fit into the basic theory of modern benefit-cost analysis in environmental policy? And as demand for such analyses grows and the community of environmental health professionals becomes increasingly involved, what are the lessons and themes from environmental economics for the design, implementation, and evaluation of environmental health regulation?

Policy debates that rely on traditional benefit-cost analysis usually revolve around the expected flow of costs and benefits, as well as the social value of time or discount rates. The grounds for decision making are the net benefits of policy as a now-or-never proposition. Environmental problems, however, are characterized by what economists call irreversibilities, uncertainties, and flexibilities: three themes of environmental economics that significantly shift the grounds for decision making and the ultimate design of policy. In the face of the irreversible effects of developing pristine lands, for example, a basic idea emerged in the mid-1960s from seminal discussions about the value of preserving national parks. Over the last quarter century, that idea has gradually developed into the theory of optimal decision making in environmental economics and management. The result today is that, at least in theory, environmental policy decisions must reconcile certain opportunity costs and benefits, the value of flexible timing, the scope of the regulation, and the prospects that the future will in some way bring missing or incomplete information to light. Table 1 summarizes the basic features of environmental health policy decisions in terms of the principal elements and their implications for policy design.

Economic analysis of environmental health policy decisions treats the social costs of regulation and the social benefits of reduced risks in terms of their monetary value. Statistical models of economic changes, environmental effects, and human health risks are integral to the approach. Implementation requires some monitoring of key environmental, economic, and health variables, particularly for applications involving learning strategies, optimal timing, and sequential policies.

The phrases "sunk costs" and "sunk benefits" express a central concept in the analysis of irreversible decisions. A sunk cost is the irrecoverable cost of adopting an environmental policy, such as the cost of installing new scrubbers on factories, scrapping old machinery for new fuel-efficient models, or paying higher prices for better-grade fuels. A sunk benefit, in contrast, is a negative opportunity cost or preventive benefit of adopting an environmental policy, such as the benefit of avoiding irreversible environmental damage, of preserving fragile ecosystems, of saving human lives, or of reducing morbidity. Understanding the trade-offs imposed by such costs and benefits is key to understanding modern benefit-cost analysis.

The basic lesson is that when environmental health decisions are being made under conditions of uncertainty, it is in the interest of society to proceed flexibly with respect to the timing, design, and scope of regulation. Irreversible decisions in environmental health regulation entail certain sunk costs. The treatment of those sunk costs in traditional benefit-cost analysis biases now-or-never decisions toward policy adoption at full levels. Irreversible decisions also, however, entail certain sunk benefits. The treatment of those sunk benefits in traditional benefit-cost analysis biases now-or-never decisions against policy adoption. A recognition of these effects leads the analysis toward the optimal timing of policy, optimal level of response, and sometimes toward go-slow or gradual policy adoption. The traditional treatment precludes the possibility that new scientific, economic, or biomedical information will come to light and significantly affect the design, timing, or scope of a proposed regulation. Add the prospect of learning about the future to the effect of irreversibility, and the value of flexibility takes on a significant and very practical meaning. The example provided in the sidebar at right illustrates the basic themes.

U.S. EPA estimates that the new [PM.sub.2.5] rule, which regulates small pollutant particles, will save some 20,000 lives a year and reduce the number of asthma cases by about 250,000. These results would constitute a clear public health benefit, but the estimate is based on uncertain biology and controvertible medical science. It turns out that [PM.sub.2.5] particles congest the air during calm weather when there is no wind to blow them away The epidemiological evidence linking [PM.sub.2.5] pollutants and, for example rates of asthma rests largely on correlations, which leaves open the possibility that other agents are involved in the incidence of asthma during calm weather conditions. The circumstances are ideal, it seems, for indoor air to become stagnant, stuffy, and laden with unhealthy concentrations of ordinary household chemicals, spores of molds, dust mite feces, and so on. These factors compound the uncertainty about the likely public health benefit of the [PM.sub.2.5] rule. At the same time, the new rule will also involve considerable, but no less uncertain, social costs - like the sunk costs of new emissions controls, taxes, and investments in new technologies.

The [PM.sub.2.5] rule clearly will have irreversible consequences (some partially irreversible and some fully so) for the economy and for public health. The case illustrates why Arrow et al. gave uncertainty so much prominence in their statement of guiding principles for benefit-cost analysis (5). There are economic uncertainties about the cost of pollution abatement and the benefits of improved public health. There are scientific uncertainties about the ecological fate and transport of air pollutants and the exact effect of small particles on human respiratory disease. These considerations argue for flexible policy response to the problem, ongoing monitoring programs, and funding of new or continuing research on the effects of air quality on human health. Finally, the case illustrates why regulations emerging from environmental policy generate so much heated controversy, particularly over methods of risk assessment and ways of figuring uncertainty into the decision-making process.

The Future of Environmental Health

Irreversibility and uncertainty involved in environmental health policy decisions such as U.S. EPA's [PM.sub.2.5] rule confer value on information and on flexibility in the design and implementation of policy. The realization of that value and the full consideration of public health priorities in economic analysis depends in part on the contributions of all environmental health professionals, especially those involved in some area of quantitative risk assessment.

Quantitative risk assessment is a chain of essential links - hazard identification, exposure assessment, dose-response assessment, risk characterization, and risk management - leading to the policy responses that determine environmental health regulation. Environmental health professionals can participate in this process in a number of ways. Walker, discussing the future of quantitative risk assessment, focused on the regulatory impact of scientific and methodological problems in the hazard identification process (2). His insights are relevant to every aspect of quantitative risk assessment, as are comments made by Johnson in his discussion of exposure assessment (11). As the lessons of environmental economics show, scientific uncertainties and methodological limitations in quantitative risk assessment strike at the heart of decision making and risk management strategies.

Flexible strategies typically need to call upon environmental health professionals to help monitor key environmental, economic, and health variables whose conditions evolve in an uncertain world. When uncertainties tell us to go slow or to do more research, the goal is to learn about, reduce, or even resolve the uncertainty. More often than not, descriptions of uncertainty depend critically on statistical inference, which puts a premium on the use of the most efficient statistical designs and analyses in establishing, for example, dose-response relationships. The same efficiencies should apply in monitoring programs, whether those programs are being used to optimize policy timing or as part of retrospective studies and policy evaluations. Increasingly, modern statistical techniques, including new likelihood methods, the Monte Carlo method, and bootstrap techniques, are designed to give more accurate estimates of the margins of error. It may be best to model uncertainty directly with dispersion models, volatility models, or other appropriate models of variability. Finally, sensitivity analysis can be employed as standard practice in case control studies. Such analyses can provide, for example, an assessment of just how robustly a hazard exposure affects the risk of disease.

Conclusion

Walker and Davis et al. forecasted several thought-provoking challenges that now confront the field of environmental health (2-4). Among these challenges are issues of economic analysis, including pressing ones of cost containment, financing, and the value of prevention dollars. How will the money be spent and on which issues? Will health priorities be heard above the din of economic debate? Many observers would argue that some but not all of the environmental regulation of the past two decades would pass a traditional benefit-cost analysis. Less can be said of a modern analysis since, with the exception of global warming, little has been done to evaluate the practical implications of the basic theory used to weight the opportunity costs (and benefits) of particular environmental problems and policies. Therefore, another challenge for the community of environmental health professionals and the National Environmental Health Association is to ensure that the lessons and themes of environmental economics are applied and communicated properly in future work.

Acknowledgements

The author thanks Hu Ying for careful comments on the original manuscript, Peter Guttorp for encouragement and support, and an anonymous reviewer for helping improve the exposition. Many thanks to all at the National Research Center for Statistics and the Environment, University of Washington, for much hospitality.

Although the research described in this article has been funded in part by U.S. EPA through agreement CR25173-01-0 to the University of Washington, it has not been subjected to the agency's required peer and policy review and therefore does not necessarily reflect the views of U.S. EPA. No official endorsement should be inferred.

REFERENCES

1. Walker, B. (1992), "Perspectives on Quantitative Risk Assessment," Journal of Environmental Health, 55(1):15-19.

2. Walker, B. (1992), "Environmental Health in the 21st Century: A Role for the National Environmental Health Association," Journal of Environmental Health, 55(3):37-40.

3. Davis, T., L. Gordon, R. Powitz, R. Roberts, B. Stern, C. Treser, and C. Wiant (1993), "Future of Environmental Health: Part One," Journal of Environmental Health, 55 (4):28.

4. Davis, T., L. Gordon, R. Powitz, R. Roberts, B. Stern, C. Treser, and C. Wiant (1993), "Future of Environmental Health: Part Two," Journal of Environmental Health, 55(5):42.

5. Arrow, K.J., M.L. Cropper, G.C. Eads, R.W. Hahn, L.B. Lave, R.G. Noll, PR. Portney, M. Russell, R. Schmalensee, V.K. Smith, and R.N. Stavins (1996), Benefit-Cost Analysis in Environmental, Health and Safety Regulation: A Statement of Principles, Washington, D.C.: AEI Press.

6. Arrow, K.J., M.L. Cropper, G.C. Eads, R.W. Hahn, L.B. Lave, R.G. Noll, PR. Portney, M. Russell, R. Schmalensee, V.K. Smith, and R.N. Stavins (1996), "Is There a Role for Benefit-Cost Analysis in Environmental, Health and Safety Regulation? Science, 272(5259):221-222.

7. Freeman, A.M. (1993), The Measurement of Environmental and Resource Values: Theory and Methods, Washington, D.C.: Resources for the Future.

8. El-Gazzar, F.E., and E.H. Marth (1993), "Foodborne Disease: Investigative Procedures and Economic Assessment,"Journal of Environmental Health,. 55(2):24-28.

9. Robbins, M., and D. McSwane (1994), "Sanitation Doesn't Cost, It Pays: Is It True and Can We Prove It?" Journal of Environmental Health, 57(5): 14-20.

10. Ballas, C., R. DiPentima, and F.A. Rusczek (1995), "Cost Benefit Of Providing Comprehensive Case Management for Lead-Burdened Children, Journal of Environmental Health, 57(9):15-18.

11. Johnson, B.L. (1992), "A Precis on Exposure Assessment," Journal of Environmental Health, 55(1):6-9.

RELATED ARTICLE: The Value of Flexible Timing - A Numerical Example

The value of flexible timing and the basic effect of uncertainty can be simply illustrated. Consider a policy whose purpose is to reduce the stock of some pollutant thought to have ill health effects on some human population at risk. Suppose that each affected person has a .004 chance of dying during the next year. Suppose further that the pollution control policy is designed to reduce that chance, but that the amount of reduction is unknown because of an uncertain link between the environmental pollutant and the health risk. The uncertainty lasts exactly one period and is forever resolved by the toss of a coin. In particular, suppose there is a zero reduction with probability .3 or a reduction of .001 with probability .7. The chance of dying therefore either remains at .004 or decreases to .003.

Next, suppose that value of a statistical life is $1,000 per individual for a policy that would reduce each individual's chance of dying by one in a thousand. The discount rate is 10 percent per period of benefit, and the cost of regulation is $5,000 per individual. If society acts today based on future expectations, then the value V of the policy is as follows:

V = -5,000 + [summation of] 698[(1.1).sup.-t] where t = 0 to [infinity] = -5,000 + 7,677 = $2,677

The net present value of policy is positive, which suggests that society should act today. So far, however, this conclusion ignores the opportunity cost of acting rather than waiting for tomorrow's resolution of the uncertain health link. Suppose instead that society proceeds with the policy only in the event that the link is positive. Since the chance that the link is positive is seven tenths, the value V of the policy is then given by

V = (0.7)[(-5,000/1.1) + [summation of] 997[(1.1).sup.-1] where t = 1 to [infinity]] = 4177/1.1 = $3,797

In this way society acts on the policy tomorrow only if new information proves that the policy is bound to reduce the risk of dying by one statistical life in a thousand. Waiting for that information increases the value of the policy from $2,677 to $3,797, so it is better in this case for society to wait.

Irreversibility is the principal driver of the decision to wait. Otherwise society would simply recover the sunk cost of $5,000 per individual if the policy fails. Nevertheless, society will only wait provided it has the option to act tomorrow. The value of this flexibility is $1,120 ($3,797 - $2,678), which may be thought of as the amount society would be willing to pay for a policy that is flexible about timing over one that requires a now-or-never decision. Considered from another angle, an option value of $1,120 means that society would be willing to pay a sunk cost as high as $6,760 tomorrow (in place of the $5,000 today) to replace an inflexible now-or-never decision with a flexible decision.

How does the degree of uncertainty affect option value? In the calculation above. the odds are seven to three that the health risk declines by one in a thousand. These odds entail some degree of certainty about the policy outcome. Increase the odds to a more certain nine to one, and the option value of waiting drops by two orders of magnitude to a mere $11, in contrast, say, to even odds, which effectively double the value of waiting. In other words, scientific certainty about a positive link between policy and health is an incentive to act. Or, interpreting the lesson in a broader sense, information has value, as, perhaps, do policies that learn incrementally along the way.

Corresponding Author: Michael J. Phelan, Associate Professor of Statistics, Chapman University, 333 N. Glassell St., Orange, CA 92866.