Is There a Precautionary Principle?

Christopher D. Stone

The author is the J. Thomas McCarthy Trustee Professor of Law, University of Southern California, Los Angeles. He may be reached at www.cstone@law.usc.edu. This Article isbased on remarks presented at the Expert's Meeting on "Genetically Modified Foods, Biosafety, the Codex and the Regulation of International Trade" at the Institute for Advanced Studies of the United Nations University in Tokyo (Mar. 18, 2000); and at the UNU Seminar Series; Inter-Linkages and Biosafety: "The Cartagena Biosafety Protocol, the WTO, and the FAO/WHO Codex Alimentarius Commission-Seeking Coherence Between Multilateral Regimes," at the New York United Nations Headquarters (Sept. 22, 2000).

[31 ELR 10790]

Introduction

Progress has always brought, along with rosy prospects, shadowy perils. The risky byproducts of technology, combined with an enhanced appreciation of hazards, is making us edgier than ever. The "precautionary principle" is being widely proposed as a response. The term has come to be routinely included in multilateral environmental agreements and declarations,¹ and is also appearing in local laws² and scattered judicial opinions.³ Notoriously, however, "the" precautionary principle's meaning—or "meanings," for it has been put forth in so many versions, often with cognate phrasing,⁴ as to belie the pretensions of the definite article—remains obscure.

The United Nations' General Assembly Resolutionon the World Charter for Nature (1982), in addressing "activities which are likely to pose a significant risk to nature," declares that "where potential adverse effects are not fully understood, the activities shall not proceed."⁵ But of course the adverse effects of activities are never "fully understood"; taken literally, the directive would be: "Don't do anything."

Other versions, such as the 1992 Rio Declaration on Environment and Development, address would-be regulators with a triple negative: not having "full scientific certainty" is not a reason not to postpone "cost-effective measures to prevent environmental degradation."⁶ But "full scientific certainty" is a bit of a red herring. In the debates over climate change, deforestation, genetically modified organisms, and so on, practically no one is urging that nothing be done pending "full scientific certainty." The controversies are characteristically over how much, if anything, should be done in light of what we do and do not know both about the science and the costs.⁷ On that score, the triple negative offers us nothing helpful. Even more puzzling is the acceptance of the "cost-effective" qualification, a constraint one would have imagined most backers of the precautionary principle to oppose.⁸

One reason for the muddle is that we are dealing, in the main, with the language of international diplomacy. The negotiators of multilateral agreements, particularly at the framework stage, may be unable to reach agreement on much more than the words themselves.⁹ In negotiating at [31 ELR 10791] that level, some degree of vagueness is less a vice than an art form, a prerequisite of garnering consent to move forward. Moreover, a term of uncertain meaning can constitute—and some version of the precautionary principle might qualify as—a seminal moral commitment. As Dan Bodansky has observed, international environmental norms are largely about "how states speak to one another."¹⁰

But whatever the role of indeterminate terms at "framework" levels, the qualities that make for an acceptable declaration about how states should speak to one another (or a good rallying cry at grass-roots levels) do not satisfy the needs of carrying out legal and administrative regimes. When a term is embodied in law, a whole slew of bureaucracies has to know what to do with it—particularly when meaning is contested in third-party dispute resolution. And the more open the meaning, the larger the room for contentious and self-serving interpretations. In the trade area, most prominently, there is concern that, as long as the precautionary principle remains nebulous, trading nations will mask as "'precautionary' health protection measures" border controls actually designed to shield domestic producers from foreign competition.¹¹

Thus, as the precautionary principle advances into law, it is increasingly frustrating that there is no convergence either as to what is means, or as to what regions of action (environment, public health) it is supposed to apply. One would hope that the necessary clarification and detail would come, in the first instance, from the principle's advocates, primarily in the environmental nongovernmental organizations (ENGO) community. But the advocacy literature is not consistently illuminating.

In one representative explication, we are informed that the precautionary principle commands an activity to be terminated "where there is sufficient evidence that [it] is likely to cause unacceptable harm to the environment."¹² Unspecified is what are to be deemed sufficient evidence and unacceptable harm and by reference to what likelihood and why harm to the environment should be privileged over effects on other goods. Does it apply to an activity that comes with a 1% probability of $ 1 million environmental damage even if it has an expected net welfare benefit, all risks included, of $ 10 million? In other words, is "unacceptable" meant to suggest an environment-risk threshold at which a proposal is disqualified from further consideration, irrespective of a much more probable improvement in welfare overall?¹³ And why/when terminate a risky activity, rather than temper it with threat of legal liability, should harm occur?¹⁴

In other embodiments, the precautionary principle is pitched in terms of burden of proof. For example, we are told that it "revers[es] the traditional burden of proof in environmental pollution cases, so that the burden is placed upon the body proposing a possibly harmful activity to show that no harm will be caused."¹⁵ But, of course, to demand prior proof that an action will cause no harm is on its face extreme.¹⁶ Imagine what the world would be like if so frugal a lid had been placed on innovation for the past century or so; not only would most technology have been thwarted (think of electricity), but perhaps also much environmental regulation, as well.¹⁷

Even if rephrased more moderately, an environment-slanted burden of proof fails to address the many proposed environment-threatening activities that are themselves aimed at protecting the environment. Consider decisions whether to "bioremediate" nuclear wastes and oil spills with microbial agents. The conflict is a tough choice between the risks of living with the status quo, the untreated hazard, as against the risks of the remediation agent getting loose and going awry. Collisions that take this form are widespread, e.g., the risks of nuclear plants versus those of fossil fuels, and little clarified by a rule that enjoins us to tilt the burden of proof to work "for, rather than against, the environment."¹⁸

Part of the lingering unclarity stems from the failure of the precautionary principle literature, by and large, to make much connection with the impressive bodies of work on decisions under uncertainty, cost-benefit analysis, and risk management. Thus, there are surprisingly few references to risk aversion, expected utility, option value, and such related [31 ELR 10792] terms that, even in only moderately technical usage, could help clarify what is being proposed.

Even more surprising than the principle's detachment from the pertinent social science literature is its disconnect from mainstream legal literature that could be quite helpful. The literature mythologizes the concept's originality by tracing the idea to the "Vorsorgeprinzip"—literally, the "principle of prior care and worry"—that is said to have arisen out of German democratic socialism of the 1930s.¹⁹ But surely the law was cultivating prior care and worry long before the term "precautionary" had been coined, or "Vorsorgeprinzip" translated. Indeed, cultivating prior care and worry is one of the things the law does. More pertinently, there has taken hold, in the United States and elsewhere, a large body of regulatory law that not uncommonly forces regulated bodies to reduce their activities to a level lower than what the mathematically expected damage would warrant—which is presumably a crucial part of what the advocates of the precautionary principle intend.²⁰ A fuller engagement with this body of legal scholarship would also help clarify how the concerns that motivate the principle in its various versions might be carried through at implementation levels.

This isolation is regrettable. The precautionary principle's patrons are serious people calling attention to serious concerns. But there appears to be so little content, and so much confusion, that it may be helpful to return to some elementary principles of risk, uncertainty, and law. The overarching aim is not to reconstruct the precautionary principle. Indeed, I shall maintain the unlikelihood that anyone can enunciate a principle that is at once both universally applicable across all risks, and more useful than "take care." The point is to lay a foundation for clarifying what properties of the unfolding challenges might warrant what sorts of cautious responses.

Paradigm Risks

A tour of basics does well to begin with the oft-repeated, foundational claim of the precautionary principle movement, that "it costs more to repair environmental damage than to prevent it."²¹ The homily is often true; but often, not. After damage has occurred we are apt to look back with regret at cases where, had we had foreknowledge, a $ 1 investment would have spared us $ 10 in damages. But while it is worth $ 1 to prevent a loss of $ 10, few would deem it worth $ 100 to avoid. In terms of the maxim, a pound of cure isn't worth a ton of prevention.

How to balance out prevention and cure, and react to it, is the province of risk assessment and risk management.²² These are things humankind has been at from the beginning, for as long as there have been fires, food, animals, natural disasters, and ill-willed or just careless neighbors. Even with millennia of experience behind us, however, no one has figured out a single principle. On the contrary. The proliferation of risks (and the capacity to detect risks) has led to an increase and specialization of categories and technique, rather than convergence.²³ Some risks are to some degree compensated in private contract, such as heightened wages for employees who accept the heightened risks of dangerous employment. Other risks not amenable to direct payoffs, such as those of driving, are tolerated within limits because living with them (realizing the net benefits) is mutually beneficial across the risk-sharing community. Against the risk of structures collapsing, civil engineers focus on the "worst case" and then apply safety margins.²⁴ Food ingredients and workplace chemicals are commonly banned in excess of certain levels (using such factors as acceptable daily intake (ADI)), based on animal tests, without further inquiry into costs and benefits.²⁵ But some environmental health fears trigger correlations of human illnesses to exposures, leaving the question of regulation subject to cost-benefit analysis, rather than hard and fast cutoffs. The risks that a device such as a nuclear plant or rocket will fail are typically addressed by analyzing the likelihood of a chain of failures of critical components. Often risk is addressed through comparative risk analysis (CRA): a level of activity or output is accepted, and the question turns toward the least risky way of achieving the accepted level. And so on.²⁶ Even more frustrating for seekers of a single principle, people notoriously underwrite widely varying expenditures to avoid the "same" risk, say, of death, depending upon whether the threat is from falling off a ladder or crashing in an airplane.²⁷

The wide range of categories both of risks and of responses poses immediate difficulties for anyone proposing a single version of "precaution" that will carry into all implementation sectors. But, to illustrate the basics, let us proceed by taking one important class of risks, those, like natural disasters and casualties, with enough data points that some form of a posteriori analysis is possible.

Imagine a person contemplating building a house (of a certain cost and quality) on a hillside that is periodically exposed to fires. Using experience tables for the given area [31 ELR 10793] (capturing the proximity to brush, distance from fire stations, and so on), and the vulnerability to loss of the proposed material, the risks can be calculated by applying probabilities constructed from experience. Suppose that the house has a total value of $ 200,000. The expected value of fire loss in any year may be $ 200. But the owner's aversion to the risk may induce her to expend more, say, $ 250, as a certainty equivalent (CE) to eliminate the risk entirely. The CE can be regarded as establishing a budget within which the risk manager seeks to locate the lowest cost combination of reducing the risk (buying fire retardant materials and smoke detectors); transferring part or all of the uneliminated risk to others, ordinarily through an insurance company (whose charge ideally reflects the insured's efforts at risk reduction); and the retention of some remainder—that is, leaving a part of the risk uncovered (ordinarily a deductible).²⁸ That is not to say what the rational solution for any person will be. The optimal depends upon one's benefit of living in the hills (the cost of not living in the hills), one's attitude towards risk—perhaps to the particular risk of fire—and the disutility of losing dollars at various levels of wealth. But in principle, at least, the methodology is not very complicated. We live and deal with risks of this sort every day.

The "New" Risks

The precautionary principle is motivated by concern for a new generation of risks, in large measure connected with the novelty of emerging technologies. Return to our hypothetical home builder to provide the basis for a contrast. The builder can foresee the set of perilous outcomes (fire, flood, burglary), and knows (or can estimate with high level of confidence) the probability that can be assigned to each outcome over a period, and even the value of various risk management responses.²⁹

But when we shift our attention to new technologies, such as living genetically modified organisms (LGMOs), we find that they come to us with no (or, only a roughly comparable) history. We are therefore doubtful we can anticipate all outcomes: Will it blind-side us with some peril unthought of even in science fiction? Nor can we with confidence assign a probability to those outcomes we can foresee. Nor can we well predict the effectiveness or costs of suggested risk-reduction and harm mitigation measures. Moreover, in the case of insuring a house, market priced replacement costs provide an ascertainable measure of and cap on losses. But to "price" some of the novel perils, we have to depend on the opinions of experts, and the experts often cannot reach consensus on anything—on data, theory, outcomes, probabilities, or the "worst" and "best" cases.

Not all novel challenges stem from novel technologies. There are such risks as acquired immunodeficiency syndrome (AIDS) and asteroids, and also novel circumstances, such as unprecedented cumulative risks. We have not faced, in recorded history, so rapid a rise in atmospheric carbon congestion, or so widespread a fragmentation of habitats. Neither situation can be traced to a single innovative technology. But what is challenging, again, is a deficit in experience.

Worse, with some technologies we cannot even produce a history. To illustrate with a contrast, a new building material can be subjected to a series of trials that enable us to defer the decision to release to market until we have produced information. But our ability to calculate some risks, principally to human health, is constrainedby public policy. For example, to determine how dangerous some things really are, the statistically favored method would be to use human control groups. But at a point where controlled human experimentation becomes politically unacceptable, we make a societal decision not to find out how dangerous the substance really is, and choose course on the basis of incomplete data.³⁰ With other hazards, the data is out of reach at any price. For example, there is no test to discover with high confidence what will happen if the congestion of greenhouse gases in the atmosphere rises to two or four times the pre-industrial level. Any analysis that relies upon frequency is out of court.³¹

One worries, too, about hazards that may unfold in paths that are nonlinear—"surprising" not only in direction but in velocity. For example, rising sea levels are worrisome in any climate change scenario. However, if the rise is smooth and gradual, at least we can learn and adjust, replacing dwellings inland as old housing stock depreciates, building sea walls, and so on, in whatever combination is most effective. But if rise in sea level should accelerate precipitously (as from unanticipated positive feedback mechanisms), learning to cope may be compressed, and costs compounded by haste's errors.

These, then, are some of the justifiable concerns that motivate the movement for caution. How are we to transform the valid concerns into institutional mechanisms?

Normal Regulatory Precaution

The fact that these motivating risks are characteristically novel does not mean that they lie outside the reach of traditional mechanisms. The question for a precautionary movement is to identify more exactly, what are the weaknesses of current strategies that need to be rethought, remedied, or supplemented?

Let us take as representative of governmental risk management the abatement of pollutants.³² Figure 1 represents [31 ELR 10794] the basics of the ordinary cost-benefit approach. The figure adopts the standard assumption that the social benefits of each unit of emission eliminated declines, the more that has already been eliminated. This is represented in the downward sloping benefits curve. The costs assumption is the opposite: that marginal costs of abatement rise as it becomes increasingly expensive to eliminate each remaining trace. The conventional optimal level of abatement is Q, at which marginal benefits and marginal costs of further reductions equate.

Figure 1

[SEE ILLUSTRATION IN ORIGINAL]

If the risk generator is not internalizing all risks, in theory the regulator can induce the socially efficient level of care by tampering either with price or with quantity. When price is the lever, the polluter is charged a "tax" (efficient tax) calibrated to bring home the full social costs of the pollutant.³³ Confronted with the efficient tax, the polluter will invest in the most efficient abatement alternatives, up to the level where that is cheaper than paying the tax.

We have not yet introduced risk. Just as homeowners would pay a premium to insure against fire (that is, pay more in dollars to avoid the loss than the mathematically expected loss in dollars) so, too, a society might do the same to avoid or mitigate certain anticipated environment-connected damage.³⁴ The standard approach is to raise the charge we impose on the risk-generating activity (or throttle emissions), shifting our target from an efficient (in the sense of wealth-maximizing) to a risk-adjusted level. The choice expends CDE (see Figure 1), the difference between the expected benefits and expected costs. But if CDE is set at the level of the CE (what we are prepared to pay to eliminate the risk), no more and no less, there is no clear basis for objection. CDE is a social risk premium, one that can be expended in many ways, ranging from foregone consumption to an increased (precautionary, if you will) investment in monitoring and even in literal insurance.

That, at any rate, is a simplified, condensed version of the standard, risk-adjusted cost-benefit model for addressing a representative set of risk-generating activities. Starting with this model as a foundation, it is possible to give a sharper focus to the inquiry: What is there about novel risks, such as LGMOs and climate change, that advocates of the precautionary principle might claim requires special departures?

(Moderate) Lack of Knowledge

The first claim might be that we lack the knowledge—the data and understanding—that the model requires. We commonly do not know with certainty either the costs (for example, of prospective abatement measures) or the benefits (for example, of associated health and safety gains). To appraise the impact of ignorance on our institutional options, it may serve to distinguish "moderate" ignorance of normal hazards from "deep" ignorance of potentially severe, even cataclysmic damage, which I will address below.

The basic response to moderate ignorance of normal hazards involves locating a balance of reliance between ex post and ex ante measures. In ex post measures, exemplified by torts and civil and criminal penalties, the law stays its hand until an injury has occurred. The ex ante strategies, exemplified by regulatory standards, such as building and safety codes, are designed to eliminate undesired levels of risk before harm happens.³⁵

The ex post strategies are not disregardful of risks; confronting risks generators³⁶ with the prospect of future sanction is calculated to motivate present caution. The difference is that the ex ante mechanisms are government-prescribed mandates designed to avert harm; the ex post mechanisms leave to the managers' expertise how to devise the most cost-beneficial avoidance of legal judgments, just as laissez-faire presumes them most capable of avoiding losses inflicted by the market. The ex ante devices are thus more "cautionary" in the sense of cutting off risks of excessive damage in advance, but less cautionary in the sense of risking the costs of excessive regulation.³⁷

Presumably, an important part of what the precautionary principle advocates favor can be expressed as seeking an expansion of ex ante measures—of opting for the risks of excessive regulation over those of excessive damage. In general, reliance on ex ante regulation grows in response to situational defects in the deployment of ex post strategies. The most favorable conditions for ex post mechanisms are when the actor responsible for the harm can be (1) identified, (2) brought to court; (3) proven liable; and (4) forced to "make good" on the judgment. It is easy to see that with regard to many modern perils, various of these elements are in doubt. As hazardous agents become more durable and wide-ranging, or prone to interaction with other chemicals in the environment, causal chains from harm back to source are obscured. The prospects of proving who was at fault for [31 ELR 10795] what decline. Jurisdiction is particularly problematical in the instance of transboundary disputes. If the damage is to a commons area, such as the high seas or seabed, the absence of an injured party—a holder of property rights—complicates redress further. And, even if a judgment should be rendered, there in no guarantee the wrongdoer can be made to pay up. Bars to judgment, including sovereign immunity, limited liability and bankruptcy, dilute the law's threats.

These are among the reasons why, depending on circumstances, we force companies engaging in risky activities to adopt mandatory workplace conditions, to abjure certain materials, to restrain output to certain levels, to install the best available technology (BAT), and so forth. But the crucial work requires turning from general principles to inescapable specifics: under what circumstances are which ex ante measures warranted, and subject to what constraints? No version of the precautionary principle ventures an answer.

Large Uncertainties

What do we do when unsureness is larger? One temptation is to accept the pattern of civil engineering and build for the worst case, or even worst case plus safety margin.³⁸

The question is important for policy purposes, because a decision to emphasize the worst case warrants a high level of "premium" against harm, and perhaps even special "worst-case" instruments. For example, Robert Costanza and Charles Perrings have made a provocative case for environmental bonds, which the risk-generator would post against the "worst case."³⁹ Several factors, however, restrain the range of cases in which minimax (minimizing the chances of the worst outcome) and bonding have appeal.⁴⁰ The basis for preferring "worst case" in the construction context includes the facts that, in building a bridge, the risks are relatively well-defined, and there is an engineering solution for each peril (load, wind, etc.), the costs of which are discernable and acceptable. But any proposal to give heavy weighting to the "worst case" (in disregard of the "best case") should at least justify why it is rejecting the best credible information. This explains why "worst-case" analysis, originally a mainstay under the National Environmental Policy Act (NEPA), came to be rejected by the Council on Environmental Quality (CEQ).⁴¹

"Worst-case" considerations come up in more complicated ways, however, and the inclination to give special weight—somehow—to the downside risks is not easily disposed of. One basis for special weighting arises out of perils that may unfold along any of several distinct paths; we may be able to supply well-defined probabilities for outcomes along each path, but be highly uncertain as to the probabilities among the n paths themselves. Such a second order uncertainty is likely to characterize "open system" perils that involve uncertain and ill-understood interactions and feedback. We can say with fair confidence: "If X occurs, then such and such; if Y, then such and such"; but we have no basis for allocating probabilities between X and Y.

That is the situation illustrated in Figure 2. The payoffs are expressed in dollars from a business as usual, no-intervention policy. The curve on the left, the worst case, can be thought of as constructed on the assumption of a positive feedback mechanism, for example, that global warming will trigger conditions that accelerate warming (melting glaciers reducing planetary albedo). The best-case curve assumes a negative feedback, such as that warming will trigger favorably cooling cloud conditions, or a dazzling, unsubsidized mitigation technology.

Figure 2 worst-best scenarios

[SEE ILLUSTRATION IN ORIGINAL]

Now, if we had confidence in the relative probabilities of the two curves, we could integrate them into a single risk distribution curve. But the more that our unsureness between the outcomes approaches true ignorance, basing policy on an assumption of, say, equiprobability, is as arbitrary as assuming the worst case. But, so, too, are both the strong pessimistic assumption, the left edge, A, and the weaker pessimistic assumption, the worst-case curve's expected value, EW.

Whether or not there are hypothetical circumstances and defensible axioms under which to justify minimax in the sense of selecting either EW or A as the "working" level of harm, there are certainly real dilemmas that come at us in clouds of unsureness. In these clouded cases, we are unable to identify the real level of risk with much confidence, much less to establish the "premium" warranted to reduce or eliminate it. Worse, the same circumstances in which we are in the dark as to the benefits of control are typically marked by large uncertainty as to the costs of control.

In these circumstances, precaution warrants a more active role than tampering with taxes and output levels. The role may range from direct management of some activities, such as geo-engineering, afforestation, and other public works projects, to financial leadership, such as subsidies to [31 ELR 10796] make up "gaps" in the coverage offered in private insurance markets.⁴²

Finally, the fact that these dilemmas are characterized by great unsureness suggests an orientation toward curing deficits in theory and information. Accordingly, the best initial response will often be to reject "locking in" to any particular long-term solution, simply on the grounds that our present best guess will almost surely prove far off the mark. Instead, one major option should be filling a basket of options that can be drawn from in mid-course, as learning improves.⁴³ Indeed, the acceleration of learning is itself a powerful cautionary strategy. The government can, for example, underwrite research and development, sponsor basic research and the gathering of data, test, and extend impact study requirements. These options should all be a major part of the precautionary conversation.

Risks to the Environment Demand a Separate Precautionary Accounting

Several versions of the precautionary principle would have us terminate an activity "where there is sufficient evidence that an activity is likely to cause unacceptable harm to the environment."⁴⁴ These versions implicitly ask us to disaggregate the expected return from an activity into two separate flows: a distribution of the probable damage to the environment and a distribution of all other costs and benefits. If the environmental costs, viewed in isolation, are unacceptable, the activity is to be stopped regardless of expected net benefits.

Versions in this form raise several problems. First, there is the question whether we can intelligibly measure and distinguish damage "to the environment" from damage to other things, such as public health and property values, that vary with environmental quality, and are indeed often the principal motivators of our caution.⁴⁵ However, to the extent that harm to the environment can be detached, some separate, initial focus makes sense. The attention of those proposing an activity frequently rests on private costs and revenues. Negative externalities, such as damage to the marine eco-system and anything else that does not translate into foreseeable and prompt commercial losses, are apt to go unaccounted for, absent special institutional requirements that they be assembled. That is the virtue of the many information-forcing mechanisms such as environmental impact statements (EIS).

More controversial is what to do with the disaggregated data once costs to the environment have been forecast. Typical versions of the precautionary principle decline to integrate the environmental costs with other cost-benefit-relevant data, so that if there is some nontrivial likelihood of environmental harm (exceeding some level), all further calculations are curtailed.

Leaving aside the problem of the still-undefined terms, the normative basis for generally privileging the environment over other values is not evident. To illustrate, suppose that the issue under consideration is the field testing of some genetically modified crop. Assume that the benefits are all food and health related and that the risks all fall on the environment.⁴⁶ Assume, also, that we have generously accounted for the nonmarket measured value of the environmental losses through appropriate contingent valuation methods. Why should we filter social decisions through a principle that says, "don't cause serious harm to the environment" (or, "the burden is on those who would cause serious harm to the environment") rather than "don't impair expansion of the food supply" (or "the burden is on those who would reduce the food supply")? The proponents of genetically modified "golden" rice want to play off against the probability of environmental damage the probability that the availability of such rice will eliminate 500,000 cases of blindness each year.⁴⁷

An argument for privileging at least some select environmental assets might appeal to incommensurability, the notion that there are some conflicts that we are reluctant to submit to comparison of costs and benefits. (It is said that most of us would reject, rather than calculate an answer to, the question, "what price would you require to sell your child into slavery?") On such grounds, anyone proposing an action that would put at risk certain designated environmental assets (as well as certain cultural ones, such as the Sphinx) might be forced at least to demonstrate both an exceptionally high and clear welfare benefit and a very low risk to the asset.

[31 ELR 10797]

In summary, a precautionary approach is certainly right, insofar as it underscores that, absent some special cautionary provisions, the environment is apt to receive inadequate protection.⁴⁸ Threats to nonmarket-measured benefits should be generously priced by contingent valuation. What is problematic—what requires fuller argument certainly as a general principle—is only why, when the environment has been so inventoried and "priced," risks to it should be privileged over other risks.

Irreversible Risks to the Environment Demand a Separate Precautionary Accounting

Some of the precautionary principle literature adopts a more moderate posture: It is not risk of every environmental harm that triggers a disregard or discount of benefits, but only those of irreversible harm. Yet, it is not clear why irreversibility should be elevated to an independent critical factor.

To begin with, irreversibility is an elusive concept.⁴⁹ If a tropical forest is converted for grazing land, it will never grow back (in the same way), and unique native species may be irretrievably lost. This is a good reason to give any such "development" a sober preview. On the other hand, the consumption of anything—from a tomato to woodland—is inherently irreversible. In life we put at risk lots of goods whose loss cannot be reversed, from limbs to heirlooms, if the benefits, adjusted for risk, are worth it. We insure (or look to the law) to restore us, if not to the same arm or heirloom, at least to an equivalent pre-loss welfare. In the same vein, if the expected benefit of the grazing land exceeds the expected lost value of the forest and species (the value of the species times the probability of its eradication), more than irreversibility is required to show that the grazing land should not be regarded as a compensating equivalent.

"More than irreversibility" might be sought in quasi-option value. It is always possible that the forest to be cleared will turn out to be the exclusive habitat of some plant species with as yet undiscovered medicinal potential. It can be said of a decision to transform the land use (1) that it is (practically) irreversible and (2) that we are likely to acquire better information about the costs and benefits of the decision with the passage of time. Under these conditions, there is undoubtedly some value in postponing decisions that will reduce biodiversity richness, while uncertainties in future supply and demand, and in properties of the environment, are clarified.⁵⁰ The point is valid. But, again, to say there is a positive option value (the sum an individual would be willing to pay to preserve the option while further information is garnered) is not to say that the value exceeds the opportunity costs (the benefits of "development"). In a sense, quasi-option value merely reminds us to add to the other aims of environmental stewardship a circumspect management of informational opportunities.⁵¹

There is another notion related to irreversibility that merits drawing out. It has no special term, but can be illustrated thus. If an auto manufacturer, after release of a new model, discovers that the accelerator sticks, the product can be recalled and corrected, so as to put a ceiling on damage. In that way, while the harm caused during the interim cannot be undone, exposure to further risk can be avoided or contained. By contrast, there is fear that, for example, an LGMO, once released into the environment, may persist to have wide-ranging, cascading outcomes. We can halt further releases. But as regards the genies that are out of the bottle, there is no simple "product recall" to put them back. At the worst, the ill effects may be unstoppable, or, if they can be reversed or contained, the costs could prove extreme. (Here are situations where the old adage about "you can't put the toothpaste back in the tube" rings true.) My sense is that our cautionary intuitions in such circumstances can be expressed as well in terms of the potential magnitude (seriousness) of the damage as in terms of irreversibility (read: costly mitigation) or irreplaceableness. But whatever the vocabulary, such scenarios are troublesome and call for appropriate responses. In the case of products, one would expect caution to take the form of, for example, stringent pre-release testing; for natural resources, such as habitats, it would take the form of demanding scrupulous EIS, sensitive to the value of deferring consumption while examining options.

The Failure to Account for Distributional Considerations

Distributional inequities provide another grounds on which to argue for reducing the level of an activity below that warranted by net, risk-adjusted benefits. To illustrate, imagine an activity that is unambiguously defensible in traditional cost-benefit terms: even the least probable, worst-case scenario of environmental harm, should it eventuate, would not offset the least optimistic benefits. But suppose that the distribution of risks is as depicted in Figure 3. All outcomes are positive for the risk generator, A, and negative for the risk bearer, B. (For example, Nation A conducts nuclear weapons tests in the Pacific, upwind of nation B.) The figure is drawn to accord with the expected gains to the winner being ample enough that the winner could pay off the loser's losses and still come out ahead.⁵² The issue is, what if the losses are not compensated, as is particularly likely given the inadequacies of international law.

[31 ELR 10798]

Figure 3 distributional objections

[SEE ILLUSTRATION IN ORIGINAL]

One can conjure circumstances in which uncompensated risk-shifting could be defended.⁵³ Nonetheless, the presumptions against it find support on grounds both of economic efficiency and intuitive fairness. The impact of uncompensated harms can and of course should be reduced by fortifying the law's compensation mechanisms. But as long as the prospect of settling up ex post are limited, there is all the more warrant to foster precautionary mechanisms ex ante. How far, is hard to say. Mechanisms that stifle externalities find favor, as do those that build on information-forcing and consent. For example, when the risk accompanies a commodity in transit, boundary crossing can be conditioned on the risk receiver's consent, as under the prior informed consent provisions of the Basel Convention and Cartagena Protocol.⁵⁴

Distributional issues arise from activities whose impact crosses generational boundaries as well. Ways of storing radioactive and highly toxic wastes that are most cost effective to the living externalize risks on the remote unborn. But there are several disanalogies when we turn to the unborn. One difference is that special compensation for victims, which is costly but not impossible in dealing with strangers in space, is practically out of the question in dealing with strangers in time. Suppose we were to set up an escrow fund for the benefit of those who may be injured in the 22d century by corrupted nuclear storage canisters. What possible assurance could we have that the fund would not be "raided"? Hence, where future damage is concerned, compensation is not an option.

One alternative to employ special ex ante safety measures on behalf of temporally remote strangers, such as exceptionally sturdy nuclear waste disposal areas, and more durable warning signs. A more general approach is to identify and avert threats to crucial life-support systems. As an example, imagine if we were to identify an asteroid heading toward collision with earth in 2300, and which could be deflected only if we were to begin costly technical measures immediately—in other words, if we agree to shoulder the costs for their benefit. (I incline to vote that we bail them out, even though they will not be able to return the favor.) Other precautionary conundrums involve discount rate. Inasmuch as the events we are concerned with are temporally remote, the adoption of almost any positive discount rate shrinks to invisibility the present value of practically every calamity. Thus, the appropriate level of discount is critical, but also, alas, a minefield of theoretical paradoxes and ethical controversies that a thorough-going precautionary movement may not be able to avoid.⁵⁵

My own view is that we are so likely to be off the mark in forecasting our progeny's most urgent challenges that the optimal way to express our care will be to endow general (as distinct from calamity-focused) benefits, such as financial health and adaptable political and economic infrastructures. In all events, the "precaution" we ought to take on behalf of spatially and temporally remote persons raise genuine issues that bring us beyond discussions of causes and effects into deepest moral philosophy.

Failure to Account for Strong Differences in Evaluations of, and Attitudes Toward, Risks

Finally, sometimes the intuition that we ought to be cautious reflects an awareness that different people (and in the international context, nations), react to risks differently. Sometimes the risks are objectively different. For example, the risks of skin cancer induced by ozone shield thinning fall heaviest on caucasians living at high latitudes, and lowest on darker-skinned populations at mid-latitudes. In other circumstances, variations in risk assessment may largely reflect differences in the degree to which the available data and recommendations of experts (ordinarily Western) are trusted. Variance in expected outcomes is always a motivator of caution—something one pays to reduce. This suggests that some of the "precautionary" motivation can be satisfied by investment in generating information, such as byproduct labeling, transparency of institutions, and "right to know" and "duty to warn" statutes.⁵⁶

There remains, however, a set of cases in which more risk information, alone, will not reduce and reconcile conflicting evaluations of the uneliminated risks. In the tumult over LGMOs, for example, the opposition arises not merely from diverging interpretations of the limited empirical data. Conflicting national and cultural values are involved, including control over what we eat, and attitudes toward science, sovereignty, and capitalism. And different evaluations—different risk targets, with differing willingness to "take the risk"—can arise out of differences in wealth. As one commentator has said of the Bergen Declaration: "One person's 'unacceptable consequence' is another's 'regrettable necessity.'"⁵⁷

When evaluations or attitudes toward risk thus clash, there is a case for caution even if, on the best objective data, expected benefits are on net positive and the distributive issues are put to the side. Here, the caution is against tearing the fabric, the obligations and rewards, of neighborliness. [31 ELR 10799] "Precaution" in these circumstances being rooted in respect for others, and in the benefits of cooperation, the response includes improving the legitimacy and consent-building capacities of global institutions. Super majority and "consensus" voting requirements may be indicated when deeply felt values collide. The World Trade Organization's (WTO's) Agreement on Sanitary and Phytosanitary Measures suggests a compromising tactic: a member invoking distinctive national risk concerns as a basis for refusal to import a potentially hazardous agent is allowed leeway in the level of border protection it selects, as long as it applies nondiscriminatorily comparable levels of risk protection across the board, including domestically.⁵⁸

Conclusion

Humankind has been dealing with perils from the beginning. Indeed, much of the activity that characterizes civilization can be conceived as responses to uncertainties: the family, irrigation projects, city walls, agriculture, savings and insurance. Risk management is nothing new. But some of the new threats, mostly technology-driven, are potentially more far-reaching in space and in time, more serious in magnitude, and the means of containing them less well understood. Caution should be high on everyone's agenda. The proponents of the precautionary principle deserve credit for their effective and insistent advocacy.

Yet, the precautionary principle—both the law and the literature taken together—is in disarray. To begin with, there is no "the" precautionary principle there. There are droves of differing versions, none of which is particularly helpful. Some sound downright wrong. It is not settled what areas the proponents aim to cover—the environment, exclusively, or health and safety as well. Are its ambitions restricted to informing when inaction is not justified, such as in its widely held but largely sterile triple negative: not having absolute causal proof of harm is not a justification for not acting? Or is the aim to stipulate when risk-reducing action is justified?⁵⁹ The "principle" is so shapeless it has even been offered as a basis for penalizing those who were not sufficiently cautious or caring in the past ("precaution," we are informed, "put into reverse").⁶⁰

Without going further, we can lay to rest the claim that the precautionary principle has entrenched itself in customary international law. The answer is "No"; or, at least, not in any significant sense. Agreed, there is a growing state practice to invoke the term. But the formulations "precautionary approach," "precaution based-approach," etc. vary, while definitions and interpretations are rampant, vague and contentious.⁶¹ Thus, while there may be a plausible argument that use of some such term is becoming part of customary law through state practice,⁶² there is no customary expectation of what conduct it implies, a prerequisite to any obligatory performance.

Certainly more clarity is possible. But there are several reasons to doubt that any single principle can be fashioned more trenchant than "take care." To begin with, the concerns that motivate caution are too varied to reduce to a unifying formula, or even guide. Often, we are cautious that the costs of an activity will exceed the benefits. At other times, the caution is to avoid morally disproportionate risks on nonconsenters across space and time. There are too many different conditions of knowledge and of ignorance. Sometimes uncertainty as to costs and benefits is only modest. At other times, our unsureness is great, as to outcomes, distributions, and even as to the disutility to be assigned to heretofore unexperienced, catastrophic states. Strategic objectives vary among eliminating, reducing, mitigating, and the spreading of losses. The mechanisms include (to name just a few) bonding, EIS and prior justification procedures, labeling, a broad basket of regulatory (ex ante) standards, pre- and post-market monitoring, outright prohibition of certain activities, and public works projects. Presumably, a higher certainty of harm should be required before imposing an absolute ban on an activity than before imposition of labeling. Temporally remote events make discount rate critical. And food poisoning is simply scarier than melting ice caps or fading species.

The implication is that there is not and cannot be a single precautionary principle elastic enough to wrap around all alternative institutional needs. Indeed, there is not a single vocabulary. Ordinary cases of risk focus attention on expected outcomes, in the language of risk aversion and utility; cases of externalized costs raise distributional discussions of "fairness." And, particularly when we seek global guidance, we are confronted with national and cultural variance not only in rankings of harm, but also in tolerance and tastes for risks. "One person's 'unacceptable consequence' is another's 'regrettable necessity.'"

Again, these comments are in no way meant to deflate the successes of the precautionary principle campaign in bolstering concern for risk in national and international arenas. But as public attention is increasingly focused, those who want to sensitize institutions to risks and uncertainties need to spend more time developing cautious norms and cautious procedures appropriate to the particular character of distinct domains. I do not see this challenge as a matter of "how to apply" the precautionary principle in this area and in that. I am questioning the claim that there is a precautionary principle there.

1. See Peter H. Sand, The Precautionary Principle: A European Perspective, 6 HUMAN & ECOLOGICAL RISK ASSESSMENT 445, 445-46 (2000) (counting 14 "hard" multilateral agreements). David VanderZwaag, The Precautionary Principle in Environmental Law and Policy: Elusive Rhetoric and First Embraces, 8 J. ENV'T L. & PRAC. 355 (1999), identifies 14 different formulations in treaty and non-treaty declarations. See also http://www.ec.gc.ca/cepa/ip18/e18_01.html#J13 (last visited May 7, 2001).

2. See Sand, supra note 1, at 448-49.

3. See Vellore Citizens Welfare Forum v. Union of India, A.I.R. 1996 S.C. 2715. Sand, supra note 1, observes few cases in which imposition of a precautionary principle operated unambiguously of its own force as a judicially imposed constraint on administrative action, even in states that have had a strong version on their books for some time. VanderZwaag, supra note 1, provides a useful review of judicial allusions, which predominantly stop short of endorsement.

4. VanderZwaag, supra note 1. Sometimes it appears as "the precaution-based approach," as in Commission of the European Communities, Communication From the Commission on the Precautionary Principle, FAO/WHO Codex Alimentarius Commission, Codex Committee on General Principles, Brussels, Codex Document CX/GP 00/3-Add.3, Feb. 2, 2000 [hereinafter EC Communication]. See also the response of the United States at Additional Comments of the United States, Risk Analysis: (1) Working Principles for Risk Analysis FAO/WHO Codex Alimentarius Commission, Joint FAO/WHO Food Standards Programme, Codex Committee on General Principles, Codex Alimentarius Commission, Fifteenth Session, Paris, France, 10-14 Apr. 2000, at 3 ("Does the 'precaution-based approach' mean the same as the proposed 'precautionary principle'?"). See http://www.fsis.usda.gov/OA/codex/confpaper.htm (last visited Aug. 7, 2000).

5. G.A. Res. 37/7, U.N. GAOR, 37th Sess., Supp. No. 51, at 17, U.N. Doc. A/37/51 (1982), reprinted in 22 I.L.M. 455 (1983).

6. "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." Rio Declaration on Environment and Development, June 14, 1992, princ. 15, U.N. Doc. A/CONF.151/5 (1992), reprinted in 31 I.L.M. 874, 879 (1992). Much the same standard is adopted by the Biodiversity Convention, Convention on Biological Diversity, June 5, 1992, pmbl., S. TREATY DOC. No. 103-20, 31 I.L.M. 818, 822-23 (1992) (entered into force Dec. 29, 1993), and the Bergen Declaration, Bergen Ministerial Declaration on Sustainable Development in the ECE Region, G.A. Preparatory Committee for the United Nations Conference on Environment and Development, 44th Sess., Annex I at 19, U.N. Doc. A/CONF.151/PC/10 (1990), 1990 YEARBOOK INTERNATIONAL ENVIRONMENTAL LAW 429, para. 7 (also reprinted in 20 ENVTL. POL'Y & L. 100 (1990)).

7. See generally The Role of Risk Analysis in the 1992 Framework Convention on Climate Change, 15 MICH. J. INT'L L. 215 (1993).

8. Peter Sand may have part of the explanation: the qualifying "cost" in "cost effective" was apparently added as an "oral correction" by the United States. Sand, supra, note 1, at 447 & n.1. "Cost effective" is sometimes distinguished from "cost beneficial" as requiring, not wealth maximization, but the least costly way to achieve an independently fixed regulatory goal, usually in terms of physical outcomes (whether or not that goal can be cost justified). Because the term was injected without elaboration, whether cost effective is to be understood as synonymous with cost beneficial or in a restricted means-ends sense, is another matter that is simply not clear.

9. The Treaty of European Union (1992), as amended, adopted the precautionary principle at Article 130 REC without any definition. Treaty Establishing the European Community, Feb. 7, 1992, O.J. (C 224) 1, 61, (1992), [1992] 1 C.M.L.R. 573 (1992), art. 130. Margot Wallstrom, the European Environment Commissioner, rejected the U.S. complaint, see supra note 4, that the European Union never defined the principle in its communication to the Codex Alimentarius Commission. See supra note 1. Wallstrom "speaking to journalists … said that 'there is a definition … in the communication.' Asked what it was [she] responded, 'it depends on the case it is being applied to …. We have stated that when it is applied it should not be discriminatory or disproportionate.'" See Wallstrom Rejects U.S. Charge That Paper Fails to Define EU Precautionary Principle, Int'l Trade Rep. (BNA), Apr. 6, 2000, at 570.

10. Daniel Bodansky, Customary (And Not So Customary) International Environmental Law, 3 IND. J. GLOBAL LEGAL STUD. 105, 119 (1995).

11. EC Communication, supra note 4, at 3.

12. James Cameron & Juli Abouchar, The Precautionary Principle: A Fundamental Principle of Law and Policy for the Protection of the Environment, 14 B.C. INT'L & COMP. L. REV. 1, 3 (1991).

13. James Cameron, who has probably produced the largest body of explicating texts has come around to "put best, 'the precautionary principle stipulates that where the environmental risks being run by regulatory inaction are in some way uncertain but non-negligible, regulatory inaction is unjustified.'" James Cameron, The Precautionary Principle, in TRADE, ENVIRONMENT, AND THE MILLENNIUM 242 (Gary P. Sampson & W. Bradnee Chambers eds., 1999). Among other problems, adopting the contorted negative form (when not acting is not justified) is at best an oblique way of dealing with the common case where regulatory action is a condition of running environmental risks, such as when a dam cannot be built, a product licensed, or a tanker permitted without regulatory approval.

14. See discussion infra.

15. David Freestone & Ellen Hey, Implementing the Precautionary Principle: Challenges and Opportunities, in THE PRECAUTIONARY PRINCIPLE AND INTERNATIONAL LAW: THE CHALLENGE OF IMPLEMENTATION 265 (David Freestone & Ellen Hey eds., 1996).

16. A comparable "no harm" position has been rejected by the World Trade Organization (WTO) in three cases under the Agreement on the Application of Sanitary and Phytosanitary Measures, in which refusal to accept imports was defended on the grounds there could be some risk. In all three, the WTO Panels and Appellate Body rejected that argument because science can never prove that an action will result in absolutely zero risk, and thus sound policy requires the assessment of risks. See Steve Charnovitz, The Supervision of Health Regulation by World Trade Rules, 13 TUL. L. REV. 271 (2000). However, "no harm" may emerge as a defensible standard situationally, especially where countervailing benefits are questionable. The Food Quality Protection Act of 1996 requires levels of pesticide chemical residues to be "safe," defined to mean "a reasonable certainty that no harm will result from aggregate exposure to the … residue, including all anticipated dietary [and other] exposures." 21 U.S.C. 346a(2)(A) (1999). Australia and New Zealand have implemented "white list" procedures under which the importation of all plant species is barred, unless they have been determined to be safe, an approach rejected by the United States in the 1970s. See Biological Invaders Sweep In, 285 SCIENCE 1839 (1999).

17. See Frank P. Cross, Paradoxical Perils of the Precautionary Principle, 53 WASH. & LEE L. REV. 851, 861 (1996) (observing that negative impacts of many regulations could not have been ruled out adequately to meet some formulations of the precautionary principle).

18. Freestone & Hey, supra note 15.

19. See Timothy O'Riordan & James Cameron, The History and Contemporary Significance of the Precautionary Principle, in INTERPRETING THE PRECAUTIONARY PRINCIPLE 14 (Timothy O'Riordan & James Cameron eds., 1994). Sonja Boehmer-Christiansen provides interesting background in The Precautionary Principle in Germany: Enabling Government, in id. at 38. "Sorge [the root, connotes] fretting or worrying about … or providing for. A squirrel storing nuts for winter, a factory taking out an accidents insurance policy, a family saving for a rainy day would all be acting with Vorsorge." Id.

20. I use the terms "caution" and "precaution" interchangeably in this Article. Precaution may have a slightly distinct nuance (suggesting measures before an activity has begun, rather than during), but not one that merits distinction in the context of the precautionary principle.

21. See the epigraph from Gro Harlem Brundtland in Cameron & Abouchar, supra note 12, at 1.

22. Insofar as the tasks can be detached, assessment involves quantifying the likelihood of an outcome; management involves such matters as deciding levels of acceptable risk, which risks to control, and in what ways.

23. See Vlasta Molak, Introduction and Overview, in FUNDAMENTALS OF RISK ANALYSIS AND RISK MANAGEMENT (Vlasta Molak ed., 1997).

24. See discussion infra.

25. An excellent treatment of U.S. food, drug, and cosmetic laws, which characteristically eliminate all risk-bearing agents beyond a certain threshold, is Peter Barton Hutte, Law and Risk Assessment in the United States, in FUNDAMENTALS OF RISK ANALYSIS AND RISK MANAGEMENT, supra note 23.

26. As precaution against the most devastating threat technology has raised, nuclear war, the contending risk-reduction principles include, along with nonproliferation of nuclear weaponry, mutually assured destruction (MAD).

27. See, e.g., Richard Zeckhauser, The Economics of Catastrophes, 12 J. RISK & UNCERTAINTY 113, 116 (1996) (maintaining that "large concentrated losses ['catastrophes']" get substantially overweighted).

28. Strictly speaking, to the extent there is a deductible, risk is not removed "entirely."

29. Frank Knight's division into outcomes and probabilities is not uncontroversial: they can be conflated. So, too, can ignorance and uncertainty, which I separate for purposes of this exposition.

30. See A. Dan Tarlock, Genetic Susceptibility and Environmental Risk Assessment: An Emerging Link, 30 ELR 10277, 10279 (Apr. 2000) (societal decision "not to know" certain risks). Of course, an aversion to human testing can also lead either to the nonapproval of a drug that might have been beneficial, or to the release to market of drugs that (it turns out) were under-sampled, and therefore less safe. See Denise Grady, Calculating Safety in a Risky World of Drugs, N.Y. TIMES, Mar. 6, 2001, at F1. The 1997 amendments to the food and drug laws actually encourage firms to undertake pediatric clinical drug trials by extending patent protection for qualified tests, see 21 U.S.C. § 355a (Supp. 2000).

31. Even this needs qualification. We are not "clueless." First, there are, if not controllable experiments, geologic data; second, as computer simulations improve, it is conceivable that, for many phenomena, the confidence level of computer simulations will exceed that of inferences from "real" experiments.

32. While we focus here on pollution, the same model applies to many other risk-generating activities, such as deforestation and fishing (the impact of a marginal tree or fish). But no single model can illustrate how all risks are accommodated; for example, it is generally accepted practice to eliminate substances of highly toxic and carcinogenic potential when they pass beyond a physical threshold. See Hutte, supra note 25. And for illustration of "worst-case" and safety factors in civil engineering, see the discussion infra.

33. Of course, we do not normally deploy a literal "tax" in the first instance. We more often adjust the toll the generator will pay through intensified legal liability (discussed more fully below); but any "gap" between the actor's expected legal liability and the efficient level can be made up by a tax.

34. Notice, however, the disanalogy between the "premium" of foregone consumption "to ensure" against, say, climate change, and a real insurance premium. In the case of actual insurance, barring insolvency of the insurer, the insured will be paid. Reducing carbon emissions, for example, does not really "ensure" us against losses, should they occur.

35. Note that ex post and ex ante strategies are not mutually exclusive, and in fact are often deployed jointly against a particular risk.

36. As we know from the Coase Theorem, the idea of there being a risk generator and a victim is an oversimplification: damage is the result of conflicting activities.

37. The magnification of ex post sanctions, such as punitive damages and fines, raises the specter of excessive caution as well.

38. It would be useful to include comparison of how risks are managed in the area of food additives, for example; concepts such as no observed effects level (NOEL) and minimum residue level (MRL) incline toward a "worst-case" posture.

39. Robert Costanza & Charles Perrings, A Flexible Assurance Bonding System for Improved Environmental Management, 2 ECOLOGICAL ECON. 57 (1990) (advocating that undertaker of perilous activities bond up to the current best estimate of maximum damage).

40. Bonds familiar to law probably work best in situations where there is a small number of actors, and the conditions of forfeit are well identified, such as where an excavator posts bond against the collapse of a neighboring property's support. My impression is that bonding is less suitable in circumstances of multiple actors and joint inputs, and where both the forfeiting event and the damage are apt to be problematical.

41. Charles F. Weiss, Federal Agency Treatment of Uncertainty in Environmental Impact Statements Under the CEQ's Amended NEPA Regulation 1502.22: Worst Case Analysis or Risk Threshold?, 86 MICH. L. REV. 777 (1988) (tracing the judicial and administrative treatment of uncertainty under NEPA and supporting the CEQ's replacement of worst-case analysis with a qualitative probability threshold).

42. See CHRISTOPHER D. STONE, THE GNAT IS OLDER THAN MAN 192-96 (Princeton Univ. Press 1993).

43. See Robert J. Lempert, Robust Strategies for Climate Change, 45 CLIMATE CHANGE 387 (2000); ROBERT J. LEMPERT & MICHAEL E. SCHLESINGER, ADAPTIVE STRATEGIES FOR CLIMATE CHANGE, INNOVATIVE ENERGY STRATEGIES FOR CO₂ STABILIZATION (Cambridge Univ. Press, forthcoming). Charles Perrings has argued for some sequencing of commitment in the face of some imaginable bad cases; we should be initially cautious (to "over-invest" in safety) in time period (1) to allow the data to enrich through learning at least until period (2). See Charles Perrings, Reserved Rationality and the Precautionary Principle: Technological Change, Time and Uncertainty in Environmental Decision Making, in ECOLOGICAL ECONOMICS: THE SCIENCE AND MANAGEMENT OF SUSTAINABILITY 153 (Robert Costanza ed., 1991).

44. See supra note 11. Michael Kloepfer has defined the essence of the Vorsorgeprinzip as that "environmental dangers [harm to the environment] and damages shall be avoided as far as possible." MICHAEL KLOEPFER, UMWELTRECHT 74 (1989).

45. Often objections to disruptions of the environment are based upon the costs exacted on human health and other indicators of human welfare. Among the costs of cutting down a tropical forest are those connected with a reduction in biodiversity; that loss, in turn, diminishes the inventory of pharmaceutical "leads," which can surely be regarded as a health cost. The cleared trees can no longer sequester carbon dioxide, which is connected, through climate change, to property damage. It is not clear how such costs are to be allocated across sectors, if disaggregation is called for. Perhaps "environmental costs" should include only those costs not assignable to other categories.

46. In fact, there would be impacts to consider on human health, costs of labor and land.

47. See David Barboza, AstraZeneca to Sell a Genetically Engineered Strain of Rice, N.Y. TIMES, May 16, 2000, at C8. In two thoughtful monographs, Indur M. Goklany maintains that the precautionary principle's advocates incline to apply it with indefensible selectivity among danger areas, ignoring scenarios to which our "caution" may fate us. For example, a precautionary banning of LGMOs ignores perils of undernourishment, malnutrition, and disease that could be alleviated by plants genetically engineered to deliver vitamins, resist salinity and pests, and so on. Goklany suggests a number of qualitative criteria for risk preference in the face of conflicting risks, including that human health should dominate environmental health, and that near-term harm should dominate distant harm (on several grounds, including enhanced opportunity for technological rejoinders). See INDUR M. GOKLANY, APPLYING THE PRECAUTIONARY PRINCIPLE TO GENETICALLY MODIFIED CROPS (Center for the Study of American Business, Policy Study No. 157, 2000) and INDUR M. GOKLANY, APPLYING THE PRECAUTIONARY PRINCIPLE TO GLOBAL WARMING (Center for the Study of American Business, Policy Study No. 158, 2000).

48. There is of course the question whether in any area the cost of the mechanisms, such as the EIS procedure, is within the "budget." It may be that insofar as special treatment is defensible, it is equally defensible in the case of every public good, such as those arising from the environment, public health, or tranquility.

49. As distinct from irreversible choice. See C.D. Kolstad, Fundamental Irreversibility in Stock Externalities, 60 J. PUB. ECON. 221, 232 (1996) ("When today's actions only affect tomorrow's costs or utility, but do not restrict tomorrow's choices, there is no irreversibility effect."). Kolstad also points out the "irreversibility" of measures designed to protect the environment from irreversible damage—in many cases precautionary environmental capital expenditures. Imagine an endangered species of fish whose habitat would be eradicated by drought. To respond, we invest $ 1 million in a dam (which offers no other benefit). Later, with change in climate, we learn that the investment was uncalled for. Could one not characterize the unrecoverable capital (and foregone benefits) as "irreversible"? Id. at 229.

50. See Kenneth J. Arrow & Anthony C. Fisher, Environmental Preservation, Uncertainty, and Irreversibility, 88 Q. J. ECON. 312 (1974); Anthony C. Fisher & W. Michael Hanemann, Option Value and the Extinction of Species, 4 ADVANCES IN APPLIED MICRO-ECON. 169 (1986).

51. Myrick Freeman suggests that option and quasi-option values cannot in all cases be considered positive. Indeed, gathering the desired information may require some level of destruction. MYRICK FREEMAN III, THE MEASUREMENT OF ENVIRONMENTAL AND RESOURCE VALUES 261-65 (1993). Freeman casts doubt on whether such allowances can really survive as a conceptually separate component of value in calculating policies under uncertainty. He would "expunge" it. Id. at 264.

52. That is, the activity meets the Kaldor-Hicks criterion for efficiency.

53. For example, A, the risk generator, might be poor, and B, the bearer, rich, so that the action would reduce disparities in wealth and could even be depicted (assuming declining marginal utility of wealth) as advancing utility. Or, even if A is richer, the gains to A might be immense relative to the losses to B.

54. See Cartagena Protocol on Biosafety to the Convention on Biological Diversity, Feb. 23, 2000, art. 6, at http://www.biodiv.org/biosafe/BIOSAFETY-PROTOCOL.htm (last visited May 3, 2001).

55. Do we adopt a thorough-going utilitarianism that seeks, as a goal, to equalize total, or average, utility across space and time? And what if some generations benefit from a policy, while others lose?

56. Sand, supra note 1, at 451-52.

57. David Fleming, The Economics of Taking Care: An Evaluation of the Precautionary Principle, in THE PRECAUTIONARY PRINCIPLE AND INTERNATIONAL LAW: THE CHALLENGE OF IMPLEMENTATION, supra note 15, at 147. Fleming makes solid arguments as to why damage to the environment is likely to be systematically under-estimated, requiring special cautionary measures as a corrective.

58. Agreement on the Application of Sanitary and Phytosanitary Measures, Final Act Embodying the Result of the Uruguay Round of Multilateral Trade Negotiations, MTN/FA II-A1A-4, Apr. 14, 1994, WTO Agreement, Annex 1A, available in WTO Goods, Agriculture-Sanitary and Phytosanitary Measures, art. 5.5, at http://www.wto.org/wto/goods/spsagr.htm (last visited May 3, 2001). The rules are much more complex than "national treatment" under the General Agreement on Tariffs and Trade, however. See Charnovitz, supra note 16, at 272-73.

59. "The precautionary principle says that in some cases—particularly where the costs of action are low and the risks of inaction are high—preventive action should be taken, even without full scientific certainty about the problem being addressed." AARON COSBEY & STAS BURGIEL, INTERNATIONAL INSTITUTE FOR SUSTAINABLE DEVELOPMENT, THE CARTAGENA PROTOCOL ON BIOSAFETY: AN ANALYSIS OF RESULTS 5 (2000).

60. "Precaution is essentially forward looking, but … those who have created a large ecological burden already should be more precautious. In a sense this is precaution put into reverse." O'Riordan & Cameron, supra note 19, at 18.

61. "How can the Commission claim that its 'precautionary principle' is 'enshrined' in various international conventions [citing] when the Commission's term ['precaution-based approach'] and concept … never appears in the text of these conventions?" U.S. Response, supra note 4, at 4.

62. "May," because even this modest position would have to account for apparent dissent by the United States and Canada. And surely, regrettably, the practice of states, especially absent conventions, involves a lot of incautious activity. But see 1 PHILIPPE SANDS, PRINCIPLES OF INTERNATIONAL ENVIRONMENTAL LAW 212 (1995), who believes the argument is at least credible.