Regulating Risks after "ATA"

Whitman v. American Trucking Association was one of the most eagerly awaited regulatory decisions in many years. But the Court's understated, steady, lawyerly opinion was a bit of an anticlimax, representing a return to normalcy and leaving many open questions. The Court was correct to say that the relevant provision of the Clean Air Act forbids consideration of cost; it was also correct to refuse to invoke the nondelegation doctrine. Importantly, the Court left in place a set of important lower court decisions, allowing agencies to consider costs unless Congress expressly concludes otherwise. The Court also raised some new questions about the constitutional status of the Occupational Safety and Health Act. Perhaps most important, the Court has not ruled out nonconstitutional challenges to the particulates and ozone standards, or to ambient air quality standards generally. In suggesting that some of those challenges should succeed, especially against the new ozone regulation, this essay urges that Justice Breyer's concurring opinion in the case may well be the most influential in the future, because it fits most comfortably with other trends in the law of risk regulation, signaling the emergence of a kind of "cost-benefit state."

create a series of cost-benefit default rules, allowing agencies to disregard trivial risks, requiring agencies to show significant benefits from regulation, allowing agencies to consider the substitute risks introduced by regulation, and authorizing agencies to take costs into account when statutes are not explicit on the point.7 Within the judiciary, the recent conflicts have pressed two questions in particular: When must agencies, including the Environmental Protection Agency (EPA), consider costs?8 When, if ever, will Congress be required to legislate with particularity?9 The latter question is not often thought to be especially interesting, because courts have not invoked the nondelegation doctrine to invalidate a federal statute since 1935.10 But a number of commentators have suggested that it is past time to revive the doctrine, partly in order to ensure legislative attention to trade-offs rather than pleasant homilies." And in the last decade or so, some courts, concerned about the grant of open-ended discretion to the executive, have shown an unmistakable interest in doing exactly that.12 Because it promised to help answer both of these questions, Whitman v American Trucking Associations, Inc."3 was the most eagerly anticipated case in administrative law in many years. In some circles, a dramatic decision was anticipated, raising the possibility of significant new directions in the law of environmental protection and risk regulation. On the central issues, however, the Court's unanimous, steady, lawyerly, even formalistic opinion amounted to a return to the status quo-and to an understated but unmistakable rebuke to the court of appeals for the District 7See Cass R. Sunstein, Cost-Benefit Default Principles, 99 Mich L Rev 1651, 1654(2001. 8See, for example, National Resources Defense Council, Inc. v EPA, 824 F2d 1146, 1154-66 (DC Cir 1987; George Warren Corp. v EPA, 159 F3d 616, 623-24 (DC Cir 1998). of Columbia Circuit. In its principal ruling, the Supreme Court reaffirmed long-settled law to the effect that in setting national ambient air quality standards, the EPA is not permitted to consider costs.14 The Court also rejected a constitutional attack on the Clean Air Act, reestablishing long-settled law allowing Congress to delegate broad discretionary authority to regulatory agencies.15 And through this return to the status quo, the current Court treated the court of appeals for the District of Columbia Circuit in the same way that the 1970s Supreme Court treated that court of appeals-attacking new judicial innovations and calling for at least a form of "hands off."'6 The Supreme Court ruled quite broadly on both of the key issues in ATA, but it also reasoned in an extremely unambitious way, invoking statutory language and precedent while saying little about the meaning of the key statutory provisions and offering astonishingly little in the way of theoretical ground for its reluctance to invoke the nondelegation doctrine. If we consider the importance of the case, the opinion seems a bit of an anticlimax, potentially even a judicial return to 1970s environmentalism. We know that the relevant provisions of the Clean Air Act do not allow costs to be considered, but we know little about what those provisions actually mean. We know that the Court is not much interested in reviving the nondelegation doctrine, but we do not know why it lacks that interest.
In evaluating the Court's approach, I will argue that the Court reached an entirely sound result, and in the process disapproved of a court of appeals initiative that fully deserved disapproval. But at the same time, the Court's formalistic opinion was exceedingly unhelpful on the central questions, in a way that leaves open many issues for the future. In the end, it will not be possible to avoid some of the conflicts between 1970s environmentalism and the cost-benefit state, and I will suggest that the conflicts should generally be resolved in favor of the latter. In making that suggestion, 14 For the initial part of the long settlement, see Lead Industries, 647 F2d at 1148-56.
"5 See, for example, Touby v United States, 500 US 160, 164-67 (1991). 16 The leading case here is Vermont Yankee Nuclear Power Corp. v National Resources Defense Council, Inc., 435 US 519, 558 (1978). See Antonin Scalia, Vermont Yankee: The APA, the D.C. Circuit, and the Supreme Court, 1978 Supreme Court Review 345, 359-75. Professor Scalia's general (not unqualified) approval of the Supreme Court's rejection of policy initiatives by the D.C. Circuit finds a kind of echo in his majority opinion in ATA. I will argue on behalf of a large role, even after A TA, for the costbenefit default principles, suggesting that they deserve a continuing place in the lower courts. I will also urge that serious nonconstitutional challenges to ambient air quality standards remain viable after A TA. The two standards in the case involve ozone and particulates; I will urge that the ozone standard is more vulnerable than the particulates standard, simply because, for low levels of ozone, the evidence of adverse public health effects is so much weaker.
Two particularly interesting sets of issues lie beneath the dry surface of the ATA opinion. The first involves the state of the nondelegation doctrine. To be sure, the Court has signaled its lack of enthusiasm for the doctrine-indeed, it has given its clearest signal yet on that point. At the same time, the Court's rejection of the approach of the court of appeals produces new questions about the nondelegation doctrine in some important domains of risk regulation, above all involving the Occupational Safety and Health Act (OSHA)."7 As we shall see, ATA raises anew a serious constitutional question about a key provision of OSHA.
The second and more pressing set of issues stems from the Court's remarkably thin and unhelpful discussion of the meaning of the Clean Air Act (CAA). The court of appeals' decision was prompted by evident uncertainty about what the key provisions actually asked the EPA to do, and the Supreme Court did precious little to resolve that uncertainty. As we shall see, the same concerns that led the court of appeals to invoke the nondelegation doctrine might well reemerge on remand in ATA-and in many other contexts involving risk regulation. Above all, the court of appeals sought to require the EPA to attempt to quantify the risks that it was seeking to control, so as to ensure that the agency was attacking large problems rather than small ones."8 After ATA, this goal might be achieved through other, more modest routes, not involving the Constitution at all. While the ATA Court spoke largely in the terms of 1970s environmentalism, it did so in the context of an Article I challenge and a highly adventurous effort to inject cost considerations into an apparently cost-blind statutory provision. As I will urge, the decision should not be seen as an attack on more modest judicial innovations, nonconstitutional in nature, that attempt to increase the sense and rationality of risk regulation.
In a period in which the Court is often criticized for allowing its own political convictions to overwhelm its duty of fidelity to the law, the quiet, lawyer-like analysis in ATA deserves respect, however many questions remain unresolved. In the process of exploring those questions, I investigate one of the most interesting features of ATA: the sharply conflicting approaches of the Court's two specialists in administrative law. Justice Scalia's opinion for the Court is lawyerly, formalistic, textualist, and apparently indifferent to the matter of consequences. Justice Breyer's concurring opinion is highly pragmatic, going well beyond the legal materials to try to make sense of the regulatory regime that Congress created. With respect to the law of risk regulation, I believe that Justice Breyer's opinion, an unambiguous rejection of 1970s environmentalism, is the harbinger of the future, and potentially the most important opinion in the case. Invoking Justice Breyer's concerns, I suggest that a number of challenges remain available to national ambient air quality standards-and that the most plausible of these challenges, connected directly with Justice Breyer's opinion, promise to increase the sense and rationality of national environmental policy.
This article comes in four parts. Part I describes the background. Part II discusses the Court's opinion and offers an evaluation. Part III briefly explores the three concurring opinions in the case, with particular reference to Justice Breyer's emphasis on recurring issues in risk regulation. Part IV discusses the future. I pay special attention here to the problems not resolved in ATA, including the place of cost-benefit default principles, the constitutional status of OSHA, and the possibility of nonconstitutional challenges to the EPA's particulates and ozone regulations, and indeed to ambient air quality regulations in general.

I. THE SETTING
To understand ATA, it is necessary to know something about what happened in the EPA and the court of appeals. For many years, the EPA had refused to issue a new national standard for particulates, notwithstanding scientific evidence apparently indicating grounds for action. Eventually the EPA was spurred to act by a decision of a district court, with an exceedingly rapid time table for action.19 So spurred, the EPA issued a new standard for small particulates in 1997;20 at the same time, the EPA issued a standard for ozone.21 In both cases, the scientific issues are extremely complex, and reasonable people might well argue about whether the evidence adequately supported the EPA's decisions. But a general reading of the evidence suggests an important distinction, one that is relevant both to ATA and to the future of risk regulation. Under the EPA's own data, the particulates standard promises significant health gains, both in terms of mortalities averted and in terms of morbidity.22 Thus, for example, the EPA estimates that the particulates standard would save over 300 lives annually and prevent 6,800 cases of chronic bronchitis. But EPA data suggest a much more mixed picture for ground-level ozone. (Some data on both proposed regulations are set forth in the appendix to this article.) The evidence is more mixed for two reasons. First, the mortality and health gains from the new standard appear much more modest; it is possible that the ozone standard would save no additional lives per year. Indeed, tighter regulation of particulates, going well beyond the EPA's rule, would appear to do a great deal more to protect health than would the new regulation of ozone. Second, the evidence for ozone is more mixed because ground-level ozone provides protection against cataracts and skin cancer.23 To be sure, cataracts and skin cancer are not the most serious of health problems, in light of the very high cure rate for skin cancer. But if the protective effects are taken into account, it is not entirely clear that the new regulation of ozone will actually produce a big improvement in terms of health.
The relevant provision of the Clean Air Act requires the EPA to set primary standards "the attainment and maintenance of which . . . are requisite to protect the public health," with "an adequate margin of safety."24 The American Trucking Association and others launched a variety of challenges to the particulates and ozone standards. Of these challenges, among the most interesting was based on the evidence just described-a claim that the EPA was obliged to take into account the health benefits, and not merely the health risks, of ground-level ozone. The court of appeals accepted this objection, concluding that the EPA had to consider all "identifiable effects" of ozone, both positive and negative.25 The government did not appeal this important aspect of the decision, whose implications for future EPA action remain unclear, in part because of the EPA's unenthusiastic response to the ruling.26 But the two major rulings of the court of appeals involved still larger issues. In the court's view, the statute simply did not allow the agency to consider costs, and hence a cost-blind approach to statutory standard was not merely permissible but required.27 In this way the court rejected the petitioners' imaginative effort to produce a departure from the long-standing understanding that national ambient air quality standards would be set without regard to costs. But construed as a "benefits only" provision, the court thought that the relevant provision of the CAA raised extremely serious constitutional problems under the nondelegation doctrine, which grows, of course, out of the constitutional vesting of "all legislative power" in "a Congress of the United States."28 In the court's view, the reason for the constitutional difficulty was simple: The relevant provision, as interpreted by EPA, seemed to allow the agency to choose whatever standard it liked best. "[T]he agency rightly recognizes that the question is one of degree, but offers no intelligible principle by which to identify a stopping 24  point."29 In the court's view, "EPA's formulation of its policy judgment leaves it free to pick any point between zero and a hair below the concentrations yielding London's Killer Fog."30 The Constitution could not be satisfied without some clearer principle distinguishing permitted from prohibited regulation. Rather than invalidating the statute, however, the court allowed the EPA to cure the difficulty. In the court's view, it remained possible that the EPA could construe the statute in a way that would remove the constitutional doubts." Such a narrowing construction would be both necessary and sufficient for validity. But what, specifically, did the court have in mind? The question is important both for understanding the Supreme Court's decision in ATA and for understanding the future of national ambient air quality standards after ATA. Clearly the court of appeals sought a high degree of quantification-indeed, a high enough degree of quantification to allow the EPA to explain, with some precision, why one standard would be "requisite" to protect the public health, while other standards would not be. In this way, the court signaled its interest in moving well beyond 1970s environmentalism, toward an approach that would rely on numbers rather than general concern and overall impressions. Above all, the court sought a specification of the kinds of numbers-in terms of likely adverse health effects under the existing standard-that would call for increased regulation, and the kinds of numbers that the EPA would consider tolerable, and hence would not call for increased regulation. In the key passage, the court wrote: "On the issue of likelihood, for some purposes it might be appropriate to use standards drawn from other areas of the law, such as the familiar 'more probable than not' criterion. Of course a one-size-fits-all criterion of probability would make little sense. . . . Nonetheless, an agency wielding the power over American life possessed by EPA should be capable of developing the rough equivalent of a generic unit of harm that takes into account population affected, severity, and probability.""32 What the court apparently wanted was some sort of general measure of adverse effects, one that would give an explanation of why, in any given case, the agency regulated to one "point" rather than another. The agency might, for example, value a statistical death at "100," treat a case of chronic bronchitis as some fraction of that, and do the same for every other adverse health effect. The universe of harms from exposure might well be aggregated in this way-so that the gains from the chosen level of regulation would be X, whereas the gains from less stringent regulation would be 80 percent of X, and the gains from more stringent regulation would be 120 percent of X. With quantification of this sort, it would be possible to know why the agency chose one point rather than another for particulates and ozone, and also to ensure that the agency's particular choices squared, more or less well, with its particular choices for other pollutants. The result would be to ensure cross-pollutant coherence, and also to provide some coherent explanation, in specific cases, of why industry is wrong to urge less stringent controls and why environmentalists are wrong to request more stringent controls. Perhaps the court suggested this "generic unit of harm" approach partly because of the odd fact that the particulates and ozone regulations were before it in the same case-and because it seemed hard to understand why the EPA did not regulate particulates more stringently, and ozone less so. Of course there are many complexities in doing what the court of appeals sought. To give just one illustration, it is not clear if the agency should focus on the probability of harm faced by each individual, or instead on some statistical measure of aggregate harms, faced by the population as a whole.33" But the EPA does seem to have sufficient information to produce some sort of overall measure of harm, in a way that would satisfy the court's concerns.34 Within the world of administrative law, a key question is whether the Constitution, or some other source of law, requires the EPA to be so quantitative, as a way of explaining why it chooses one degree of regulation rather than another, either more or less stringent. The court of appeals held that the Constitution does indeed imposes this requirement on the EPA, as a way of cabining its own discretion.
The government promptly appealed, objecting above all to the nondelegation ruling. In a bit of a surprise, the respondents crosspetitioned, arguing that the Act should be construed to allow the EPA to consider costs in setting ambient standards. The crosspetition was analytically prior, and this was the question that the Supreme Court addressed first.35

A. COSTS? NO ELEPHANTS IN MOUSEHOLES
The first question was whether national ambient air quality standards should be based on an assessment of benefits alone, or instead on some kind of balancing of benefits and costs. Recall that the relevant provision of the CAA requires the EPA to set primary standards "the attainment and maintenance of which ... are requisite to protect the public health," with "an adequate margin of safety."''36 At first glance, the statute appears to be indifferent to cost. All that matters is that pollution be controlled so as "to protect the public health." This view of the provision is strengthened by the fact that the statute requires national standards to be based on a "criteria" document, which is required, in turn, to discuss all "identifiable effects" of the pollutant on public health and welfare.37 The criteria document is not required to discuss the economic effects of regulation itself. Thus the analysis seemed simple: If the ambient standard is required "to protect the public health" and to be "based" on the criteria document, and if that document is not supposed to discuss costs, then it would seem to follow that standard setting may not be done with reference to costs.
In essence, the Court accepted this analysis. Proceeding in a textualist manner, with an emphasis on statutory language and struc-35 I do not discuss the third ruling, involving the relationship between the new ozone rule and statutory provisions governing ozone. This third ruling has considerable practical importance, but it does not raise recurring issues. 36 42 USC ? 7409(b)(1) (1994). ture (but no reference to the legislative history), it held that standard setting must indeed be cost-blind.38 The Court said that the key statutory term, "public health," requires an analysis of the effects of the pollutant, and that costs are irrelevant. At first glance, this does indeed seem correct: If a provision asked an agency to do what is "requisite" to protect the "public health," the more natural (though not inevitable) reading is that the agency is to concentrate on benefits, not on balancing benefits against costs. But the American Trucking Associations and their allies pressed some good questions from common sense: Can the EPA possibly issue national standards on the basis of an inquiry into public health only? If the costs of compliance are extremely high, won't the EPA inevitably impose more lenient regulations than it would if the costs are extremely low? Isn't it obviously worthwhile to achieve some small improvement in air quality, if the costs of compliance are (say) $500,000, and obviously less worthwhile to do so if the costs of compliance are (say) $10 billion? These questions seem especially powerful in light of the fact, acknowledged by the EPA, that there is no "safe threshold" for many pollutants, including those involved in ATA itself. If there is no safe threshold, a costblind analysis would seem to require the EPA to eliminate pollutants from the ambient air-a result that would ban automobiles, coal-fired power plants, and much more. Consideration of costs would appear necessary to avert this ludicrous conclusion. And the argument seemed to draw further strength from the apparent fact, urged by credible observers, that the EPA had in fact considered costs, although tacitly and without public supervision.39 In fact, it makes most sense to interpret the "requisite to protect the public health" language to require the EPA to make a showing of significant risk, a point that goes some (not all) of the way toward answering these objections.4 If the remaining risks are trivial, additional regulation should not be seen as "requisite to protect the public health." But the Supreme Court's major answer to these points was simple: Tell it to Congress.41  real question was whether "public health" could be understood to be a reference not only to environmental effects, but also to the adverse effects introduced by regulation. Invoking the dictionary as its principal authority,42 the Court suggested (not so helpfully) that the ordinary meaning of "public health" is "the health of the public," that this is the "most natural of readings," and that the natural reading is inconsistent with the claim that costs are relevant to the EPA's decision. 43 To be sure, the cross-petitioners did not rely on policy arguments alone. They also urged (and the Court acknowledged, also with the dictionary's aid) that "public health" is a not an unambiguous phrase. In their view, "many more factors than air pollution affect public health," and "a very stringent standard might produce health losses sufficient to offset the health gains achieved in cleaning the air-for example, by closing down whole industries and thereby impoverishing the workers and consumers dependent upon those industries."44 There is evidence to support this claim. Expensive regulation does appear to produce health risks.45 The Court did not reject the empirical claim, but it held that arguments of this kind could not defeat the more "natural" interpretation of the Act. Indeed, the Court analyzed the statutory structure to give further support to that interpretation. Unlike the apparently cost-blind provision governing national ambient air quality standards, other provisions of the Clean Air Act explicitly permitted cost to be taken into account. For example, a key provision of the Act asks the EPA to consider costs in selecting the "best system" for emissions reductions from new pollution sources.46 The fact that some statutes expressly referred to cost seemed to support the view that the provision at issue in ATA, most naturally read as cost-blind, did in fact have this meaning. To this the Court added that Congress was "unquestionably aware" of the potentially adverse effects of expensive regulation, and hence allowed the EPA to waive the compliance deadline for stationary sources upon a showing, inter alia, that "'the continued operation of such sources is essential . . . to the public health or welfare.' "47 All this was enough to suggest that the basic provision of the Act banned consideration of cost. Because that basic provision is "the engine that drives nearly all"48 of the subsequent provisions of the Act, Congress would not lightly be taken to have required cost consideration in such subsequent provisions.49 Congress "does not alter the fundamental details of a regulatory scheme in vague terms or ancillary provisions-it does not, one might say, hide elephants in mouseholes."s0 Hence the statute's terms "adequate margin" and "requisite" should not be taken to allow consideration of cost, because it is "implausible that Congress would give to the EPA through these modest words the power to determinate whether implementation costs should moderate national air quality standards."" To be sure, a number of provisions of the Act expressly require the EPA to generate information about compliance costS.52 But in the Court's view, these provisions are designed to allow the EPA to assist the states in finding low-cost strategies for attainment."s3 They do not suggest that the EPA is allowed to consider costs on its own. All of these considerations converged on a single conclusion: Ambient standards must be based on considerations of health, and cost simply does not matter.

B. DELEGATION?
If the EPA's task is to assess public health alone, is the statute an unconstitutional delegation of power? Common sense might suggest that there is a serious problem here. The question of what level is "requisite to protect the public health" seems unaccompa-14 THE SUPREME COURT REVIEW [2001 nied by statutory standards. If there are no safe thresholds, perhaps the EPA could (or must) require some pollutants to be eliminated entirely from the ambient air. But even in the absence of safe thresholds, perhaps the EPA could decide that certain risks are to be treated as residual or ordinary-the stuff of everyday life. If the statute does not tell the agency what it must do, perhaps its range of discretion is unacceptable. Note in this regard that many people do not believe that certain imaginable steps are "requisite" to protect their own health: They walk at night in dangerous neighborhoods; they eat peanut butter; they fail to exercise; they gain weight; they drive cars, sometimes (believe it or not) above the speed limit; they use cell phones, sometimes while driving; they own large dogs. In all of these cases, the risks may well be nontrivial as a statistical matter, and perhaps the EPA could build on such practices in deciding what level of regulation is "requisite." But without statutory guidance, perhaps the Act gives the EPA excessive discretion-the discretion essentially to choose the level of risk regulation that it wants, with essentially no legislative guidance.
A statute requiring cost-benefit balancing might, on this line of reasoning, raise similar nondelegation issues, at least in the absence of legislative guidance on how to assess both costs and benefits. Without such guidance, is an agency permitted to say that a statistical life is worth $1 million? $10 million? $50 million? Can an agency discount lives saved in the future? By 1 percent? By 7 percent? If the statute allows agencies to value statistical lives as they choose, there might seem to be an unacceptably high level of discretion. Thus a constitutional attack on the "requisite to protect the public health" language would raise doubts about many other provisions, including those that give costs a role to play in regulatory choices.54 Troubled by the wide range of discretion apparently enjoyed by the EPA, the court of appeals held that the Constitution required the agency to come up with clear standards by which to explain why it would exercise its discretion to regulate to one "point" rather than another. But as a constitutional matter, the Supreme Court found the lower court's approach quite implausible. "In a 54 For a discussion of costs in regulatory choices, see Sunstein, 98 Mich L Rev at 324-30 (cited in note 34); Sunstein, The Arithmetic of Arsenic (cited in note 34). delegation challenge, the constitutional question is whether the statute has delegated legislative power to the agency."ss Because this is the question, an agency cannot "cure" an otherwise unconstitutional delegation of power merely by adopting a narrow construction. "The idea that an agency can cure an unconstitutionally standardless delegation of power by declining to exercise some of that power seems to us internally contradictory."s6 Thus the agency's "voluntary self-denial"" has no bearing on the real constitutional question, which is whether Congress has cabined agency discretion with the requisite "intelligible principle." Turning to that question, the Court agreed with the Solicitor General that the statute required the EPA to use the latest scientific knowledge to establish standards "at a level that is requisite to protect the public health from the adverse effects of the pollutant in the ambient air."'58 The statutory idea of requisite-to-protect meant "sufficient, but not more that necessary," to provide that protection.59 In the Court's view, this standard was constitutionally acceptable, because it was "strikingly similar" to limitations found acceptable in other cases.60 The Court had, for example, upheld a grant of authority to the Attorney General to designate a drug as a controlled substance if this step was "necessary to avoid an imminent hazard to the public health."''61 The Occupational Safety and Health Act contained a quite similar provision, which was also found constitutionally valid.62 The Court acknowledged that "the degree of agency discretion that is acceptable varies according to the scope of the power congressionally conferred."63 Hence Congress "must provide substantial guidance on setting air standards that affect the entire national economy."64 But it was nonetheless unnecessary for Congress to provide a "determinate criterion" to establish how much regulation is too much.65 To be sure, the EPA would be making "judgments of degree."66 But the decision about what is "requisite" would not run afoul of the nondelegation principle.

C. EVALUATION: TWO AND A HALF CHEERS
On both of the key issues, the Court's unanimity should not be surprising. Indeed, the Court's fidelity to preexisting law-even its steady, unimaginative, lawyerly opinion-gives reason for some confidence in a Court that has been much criticized for allowing its own political convictions to play an excessive role in its interpretation of the law. It is far from unreasonable to speculate that some of the justices lacked much enthusiasm, on political grounds, for a cost-blind Clean Air Act, and that some of them would like Congress to give narrower delegations of authority to regulatory agencies. Nonetheless, the Court took the existing legal materials, in the form of statutory language and judicial precedents, extremely seriously. The Court was right to say that those materials supported both results. The best interpretation of the CAA does not allow the EPA to consider costs when setting national standards. Moreover, there is no good argument for a large-scale revival of the nondelegation doctrine; and if the doctrine is to be used in the most extreme cases, ATA was not such a case.
1. Cost. To say this is not to deny that the case had its complexities, or to urge that it would be impossible to imagine an opinion coming out the other way on either issue. At a minimum, the Court might plausibly have said that in deciding what margin of safety is "adequate," the EPA is permitted to take account of costs, not merely benefits. In addition, the phrase "requisite to protect the public health" is not altogether clear. The idea of "public health" is ambiguous, and perhaps cost is relevant to whether a certain level of stringency is "requisite" to protect it. Especially in view of the fact that expensive regulations can have harmful effects on public health-a fact of which Congress was aware-it could be urged that whether a particular regulation qualifies, under 6s Id, quoting American Trucking, 175 F3d 1027, 1034 (DC Cir 1999) (per curiam). 66 Id at 914. that phrase, as "requisite" is a function of cost, not merely benefit. A reasonable court might well be drawn to this interpretation, at least if it were seen as the only way of making sense, rather than nonsense, of the statute as a whole. Such an opinion would not have been beyond the pale.
Nor would this approach entirely lack precedential support. In the Benzene Case,67 the Court interpreted the Occupational Safety and Health Act in a way that would ensure against regulatory absolutism-not by requiring cost-benefit balancing, but by saying, without textual support, that the agency could not regulate a risk not shown to be "significant." In dissent, Justice Marshall, sounding very much like Justice Scalia in ATA, contended that conventional legal tools could not justify imposing a "significant risk" requirement, which, in Justice Marshall's view, was a judicial fabrication.68 In defending its conclusion, and its somewhat irreverent approach to the statutory text, the Benzene plurality emphasized that the government's approach "would give OSHA power to impose enormous costs that might produce, little, if any, discernible benefit.""69 It would not have been a gigantic stretch from this line of reasoning to a holding, in ATA, that the EPA should be permitted to consider whether an enormous outlay of expenditures can be justified by the benefits received. In any case, a number of lower court decisions have not stayed close to the statutory text where common sense seemed to require a measure of regulatory flexibility.70 Nonetheless, the path that the Court chose was unquestionably the easier one, because it fit so much more naturally with the statutory language. The Court was right to say that "requisite to protect the public health" is more naturally read to focus on benefits alone. To this I might add three further points. First, the pressure on the Court was greatly relieved by the fact that in the implementation process, costs could indeed be considered, and at multiple stages. This point sharply distinguished the CAA from OSHA, which lacks any comparable implementation stage.71 If Justice Scalia's majority opinion in ATA is close to Justice Marshall's dissenting one in the Benzene Case, it might be partly because the consequences of a cost-blind approach to EPA standard setting are not quite what they seem-as Justice Scalia was undoubtedly aware. Second, the "requisite to protect the public health" language might well be regarded as imposing a significant-risk requirement. The Court said little on this issue, a gap to which I will turn in due course; but even on the Court's view, the CAA does not require the EPA to remove all risks, regardless of their magnitude. If a regulation leaves small risks, it can be argued, reasonably enough, that more stringent controls would not be "requisite." For this reason the EPA's position in ATA was far more cautious and modest than OSHA's more extreme position in the Benzene Case. Third, textualism might be regarded as a kind of "penalty default," imposing a burden on Congress, and on relevant interest groups, to provide a corrective for any problems introduced by a cost-blind approach. There is room for reasonable debate about whether Congress is sufficiently responsive to penalty defaults of this kind. But some comfort can come from the fact that a legislative corrective is certainly possible, and has in fact been provided in a somewhat analogous context.72 2. Delegation. On the nondelegation issue, the Court's opinion is also entirely plausible, indeed correct, but even less satisfying. At least at first glance, the Court appears to have delivered a death blow to those who have sought to revive the nondelegation doctrine. It appears to have delivered such a blow less by its holding than by its evident lack of interest in preserving much future space for the doctrine. Why, exactly, did the Court do this? The Court's opinion is quite wooden on the point. The Court said little about 7 This is because the Clean Air Act requires states to produce implementation plans to bring about compliance with federal standards, whereas OSHA regulations are directly binding on the private sector. 72 See 21 USC ? 376(b)(5)(B) (1994) ("Delaney Clause"), as discussed in Public Citizen v Young, 831 F2d 1108, 1111-18 (DC Cir 1987. In 1996, Congress amended the Delaney Clause in various ways, softening its absolutist character by replacing the flat ban with a requirement of a "reasonable certainty of no harm," defined in the legislative history to reflect a policy of reducing cancer risks in the exposed population to no more than one in 1 million additional lifetime risk. 21 USC ? 346a(b)(2)(A)(ii) (1994( & Supp 1999 the nondelegation doctrine in principle-about what values it serves, about why courts should or should not be reluctant to invoke it. The Court's analysis consisted mostly of a lawyer-like (law clerk-like?) effort to show that the delegation in the CAA was no different from those delegations found acceptable in many other cases. This was an example of the dullest and the most unambitious form of analogical reasoning-the sort of thing recently associated with Chief Justice Rehnquist, or even former Chief Justice Burger, and very different from what usually emerges from Justice Scalia's pen.
Most strikingly, the Court made no effort to explain exactly why it was so willing to allow Congress to grant considerable discretionary authority to the EPA. Nor did the Court explain why it was uninterested in reviving the nondelegation doctrine in general. With respect to analogies, the Court basically skimmed the surface, arguing, plausibly if not entirely convincingly, that the delegation here was not evidently greater than the delegation in other cases. The Court offered exceedingly little in the way of detail. To be sure, the "requisite to protect the public health" language does not seem, in the abstract, to be as open-ended as other statutory language, such as the "public interest" standard in the Federal Communications Act. But recall here the central concern of the lower court in ATA: What limits does the Clean Air Act impose on the EPA's exercise of discretion? If this question turns out to be hard to answer, it would not be terribly difficult to distinguish most of the prior cases. The EPA, for example, is authorized to cover a much larger portion of the American economy than was involved in the precedents, and it would be at least plausible to say that its discretion was broader as well. To know whether this latter argument is convincing, it would be necessary to say much more about the meaning of the relevant provision of the Clean Air Act. But here the Court's efforts were thin to the point of being comical. The repetition of statutory terms-"requisite to protect the public health"-was not much aided by the Court's (repeated) explanation that these terms meant "not lower or higher" than necessary to protect the health of the public. 73 We might say that with respect to the nondelegation doctrine, the Court's opinion is remarkably shallow, in the sense that it lacks any theoretical ambition, but also remarkably wide, in the sense that it appears to dispose of a wide range of imaginable nondelegation challenges.74 The Court has made it clear that the nondelegation doctrine will be used, if at all, only in the most extreme cases. But it has given little indication of why it is taking this approach. Is it possible to say what lies beneath the surface here? Some clues come from Justice Scalia's separate writing on the topic.75 Justice Scalia has objected to the nondelegation doctrine on the ground, not that it lacks constitutional roots, but that it is not subject to principled judicial enforcement. The line between permissible and prohibited delegations is inevitably one of degree: How much discretion is too much discretion? Judicial efforts to answer that question are not susceptible to rule-bound judgments, and hence particular rulings would inevitably have the appearance, and perhaps the reality, of judicial partisanship. In these circumstances a judicially enforced nondelegation doctrine might even be said to violate that very doctrine's aspiration to rule-bound law. A judge who aspires to reduce judicial discretion will inevitably be skeptical about the nondelegation doctrine.
I believe that these concerns help explain the Court's surfaceskimming in ATA-and also that the concerns are well founded, in general and in the particular case. These institutional problems are best combined with a point about the nature of the regulatory state. There is no good evidence that a large-scale revival of the nondelegation doctrine would make constitutional government work better, from any point of view. In this light the case for reviving the doctrine seems weak indeed.76 In fact, the purposes underlying the doctrine can better be achieved through other routes, including substantive reform, the hard look doctrine, and narrowly targeted "nondelegation canons."77 In ATA itself, it is reasonable to think that the relevant provisions were not so open-ended as to run afoul of a sensible nondelegation doctrine. Indeed, it would 74 For a discussion of shallowness/depth and narrowness/width, see Cass R. Sunstein, One Case at a Time: Judicial Minimalism on the Supreme Court 10-23, 244-58 (Harvard, 1999 be possible to say that those provisions are valid even if the doctrine should be used in the extreme cases.
But all this leaves a nagging question: What does the Act mean? Here the Court's answer is embarrassingly thin. The respondents urged that the statutory standard-"requisite to protect the public health"-could mean anything at all. To respond to their complaint, it would be necessary to explain how the standard creates ceilings and floors on agency action. I will return to the question in Part IV.

III. FUNDAMENTALISM, SEMANTICS, AND PRAGMATISM
ATA produced three separate opinions of considerable interest. Justice Thomas's concurring opinion is noteworthy because of his brief remarks on the nondelegation doctrine. Justice Stevens's far longer concurrence is noteworthy because of the caveat he seeks to add to the majority's view. Justice Breyer's opinion is the most important of all, because of its effort to make sense of the regulatory scheme, and because of its attempt to provide some guidance for the future of risk regulation.
A. FUNDAMENTALISM Justice Thomas agreed that the CAA provided "no less an 'intelligible principle' than a host of other directives that we have approved.""78 But he objected that the parties had provided "barely a nod to the text of the Constitution," which does not speak of "'intelligible principles.' "79 Instead, he urged, "it speaks in much simpler terms: 'All legislative powers herein granted shall be vested in a Congress.' ""80 In Justice Thomas's view, some statutes might contain "intelligible principles" but still involve a delegated decision whose significance "is simply too great . . . to be called anything other than 'legislative."'81 For this reason, Justice Thomas indicated his willingness "to address the question whether our delegation jurisprudence has strayed too far from our Founders' understanding of separation of powers."82 At first glance, Justice Thomas's reasoning seems mysterious, perhaps even ridiculous. The requirement of an "intelligible principle" is many decades old.83 It would be truly astonishing if the Court were to abandon the requirement as an understanding of the nondelegation principle. In any case, the idea that "all legislative power" is vested in a Congress is hardly self-interpreting; the question is what this provision means. It would even be possible to read the constitutional background to suggest that Congress can delegate such legislative power as it wishes. If this conclusion is overdrawn-as I believe that it is84-then the question becomes when some amount of executive discretion becomes "legislative" in nature. The requirement of an intelligible principle is simply a means of answering that question.
But underneath the implausible formalist veneer, Justice Thomas might well have a point. His key sentence is this: "I believe that there are cases in which the principle is intelligible and yet the significance of the delegated decision is simply too great for the decision to be called anything other than 'legislative.' " Perhaps Justice Thomas is suggesting that an otherwise acceptable delegation would be objectionable if the area or context-the "significance"-suggested that an intelligible principle is not enough. So understood, Justice Thomas's view is fully consistent with the majority's suggestion "that the degree of agency discretion that is acceptable varies according to the scope of the power congressionally conferred."86 On this view, the nondelegation doctrine is not entirely without life. Where an agency is given the power to regulate much of the American economy, the statutory principle must be more intelligible than it must otherwise be. All statutory principles must be intelligible, but some must be more intelligible than others. This is a perfectly reasonable suggestion.

B. SEMANTICS
In an opinion joined by Justice Souter, Justice Stevens urged that the Court should admit that agency rulemaking authority counts as "legislative power," but that it is not unconstitutional for that reason.87 If national ambient air quality standards had been issued by Congress, "everyone would agree that those rules would be the product of an exercise of 'legislative power.' "88 In Justice Stevens's view, an agency that makes rules, under an acceptable delegation, is also acting "legislatively." For Justice Stevens, the proper way to analyze the case is to acknowledge that the agency acted legislatively, but to hold that the exercise of legislative power is legitimate because it was constrained by a sufficiently intelligible principle.
This seems to be a purely semantic debate, and nothing at all appears to turn on it. Where the majority would say that legislative power had not been delegated, because an intelligible principle could be found, Justice Stevens would urge that the delegation of legislative power was acceptable, because an intelligible principle could be found. Justice Stevens was apparently writing for the sake of conceptual clarity. But it is unclear whether his approach counts as an improvement on that score. The Constitution vests legislative power in Congress, and on the standard view, legislative power cannot be delegated. In any event, the executive has long had the power to make rules, and there is nothing odd in saying that rulemaking, if confined by some principle, is an executive function if the executive is making rules. True, the same exercise would be legislative if Congress were carrying it out, a point that suggests that whether an activity counts as "legislative" or "executive" depends, in large part, on whether the Congress or the executive is acting. But where is the oddity in that conclusion? The Court's approach seems as sensible as that offered by Justice Stevens.
A good way to make progress would be to say that words, not excluding the word "legislative," have meanings connected with, and not divorced from, the purpose for which they are used. There is nothing odd in saying that rulemaking is "executive," when sufficiently cabined by statutory standards, if undertaken by the executive branch-but that the same sorts of decisions would be "legislative" if done by Congress. It seems to me that Justice Stevens has been led astray by the ordinary language use of "legislative." Even if agency rulemaking is legislative in one, perfectly intelligi-87 Id at 920-21 (Stevens concurring in part and concurring in judgment). 88 Id at 921.

THE SUPREME COURT REVIEW
[2001 ble sense, it is not legislative, in the sense of Article I, section 1, if an executive agency has engaged in it.

C. PRAGMATISM
In its own way, Justice Breyer's opinion was the most remarkable one in the case. It offers an intriguing comparison to Justice Scalia's majority opinion, indeed an opinion that might easily be imagined as a majority opinion of a very different sort. Justice Breyer's separate opinion seemed to have three goals. The first was to show that the legislative history of the Act, and not merely its text, strongly supported the majority's conclusion.89 This part of Justice Breyer's opinion is part of a long-standing disagreement between Justice Scalia and Justice Breyer,90 and raises no novel problems.
Justice Breyer's second goal was to endorse a principle of interpretation, one that would allow agencies to consider costs in the face of statutory silence or ambiguity. "In order better to achieve regulatory goals-for example, to allocate resources so that they save more lives or produce a cleaner environment-regulators must often take account of all of a proposed regulation's adverse effects, at least where those adverse effects clearly threaten serious and disproportionate public harm. Hence, I believe that, other things being equal, we should read silences or ambiguities in the language of regulatory statutes as permitting, not forbidding, this type of rational regulation."91 The reason that the EPA could not consider costs here was that other things were not "equal"; the statutory structure and history showed "a congressional decision not to delegate to the agency the legal authority to consider economic costs of compliance."92 This is a centrally important cautionary note, one to which I will return. Justice Breyer's third goal was to establish the rationality of the legislative scheme as he understood it. He wanted to show, as Justice Scalia did not, that the "natural" construction made sense. His overall suggestion was that as construed, the statute would, at pertinent phases, allow consideration of costs after all, and that the EPA would be permitted to construe the statute reasonably, by making it less draconian than it appeared. To this end he offered four points, each of considerable interest: 1. As part of its cost-blindness, the statute was designed not to take pollution control technology as a given, but to force technological innovation. This was, and is, an entirely realistic hope.93 In fact, the catalytic converter was developed as a result of a seemingly draconian statutory mandate. And because the statute was expressly designed to force technological innovation, regulatory efforts to calculate the costs of compliance were "both less important and more difficult"94-because the relevant calculations would be based on speculation about the cost of unknown future technologies. These calculations "can breed time-consuming and potentially unresolvable arguments about the accuracy and significance of cost estimates."9s In these circumstances, cost-benefit analysis might itself fail cost-benefit analysis.
2. Even as interpreted by the Court, the Act allows cost and feasibility to be considered.96 These factors are relevant, for example, to states deciding on the mix of control devices used to achieve compliance; and those facing economic hardship can seek an exemption from state requirements. The EPA is also permitted to consider costs, not in setting standards, but in setting deadlines for attainment. Congress is also available to extend deadlines if necessary.97 In fact, noncompliance with national standards has been a persistent pattern under the Act, in part because compliance can be so costly. The relevant provision of the CAA might be costblind; but this is not at all true for the statute as a whole.
3. The EPA is not required to "eliminate every health risk, however slight, at any economic cost, however great.""98 Standards "'requisite to protect the public health"' need not produce "a world that is free of all risk-an impossible and undesirable objec- tive."99 In fact, the terms "requisite" and "public health" should be understood in their context. It is relevant for the EPA to consider "the public's ordinary tolerance of the particular health risk in the particular context at issue."100 Hence the EPA is allowed to produce a kind of common law of "acceptable" risks, rather than eliminating all risks as such. In deciding what is "requisite to protect the public health," the EPA is allowed "to consider the severity of a pollutant's adverse health effects, the number of those likely to be affected, the distribution of the adverse effects, and the uncertainties surrounding each estimate."'1o On this count, Justice Breyer is saying something more cautious than, but something not terribly far from, what was said by the court of appeals. In his own way, Justice Breyer was urging the EPA to develop standards for separating acceptable from unacceptable risks.
4. The EPA is allowed to "consider whether a proposed rule promotes safety overall."102 If a rule causes "more harm to health than it prevents," it is unlawful.3"' In this way Justice Breyer endorsed the lower court's conclusion that the EPA is required to consider the health benefits of ground-level ozone, not just the health risks.
What is most important about Justice Breyer's opinion is the effort not merely to read the statutory terms, but also to make sense out of them-to show that the statutory framework is not as silly, or absurd, as it might seem to be in the abstract. There is a noteworthy contrast here between Justice Scalia's approach, for the majority, and the approach favored by Justice Breyer.'04 And because it pays such attention to the pragmatic issues, Justice Breyer's opinion might well prove to be the more important for the future. Each of the four points just mentioned has significant 102 Id. 103 Id. 104 I will not discuss here the more general jurisprudential issues raised by formalism and pragmatism in statutory interpretation. A large issue is empirical-whether one or another approach can be defended on pragmatic grounds. If judges are undistinguished, formalism starts to look a lot better; hence the best defenses of formalism are pragmatic in character. implications for EPA decisions and indeed for regulation more generally, as we shall shortly see.

IV. IMPLICATIONS -AND THE FUTURE
ATA leaves many issues open. For the future, the three crucial questions involve (a) the place of the cost-benefit default principles, which have played a large role in the last two decades of federal administrative law; (b) the status of the nondelegation doctrine; and (c) the legal standards governing ambient air quality standards, including the very standards involved in ATA.

A. COST-BENEFIT DEFAULT PRINCIPLES
The Court was well aware that a number of lower court decisions have established a new interpretive principle: Where statutes are ambiguous, agencies are allowed to consider costs.1?5 What is the status of this principle after ATA? While things are not entirely clear, the best answer is that the principle is unaffected. In fact, Justice Breyer went out of his way to endorse the basic idea.o06 Justice Breyer was careful to say that courts "should read silences or ambiguities in the language of regulatory statutes" to permit consideration of "all of a proposed regulation's adverse effects," at least where those effects would "clearly threaten serious and disproportionate public harm."'17 For its part, the majority specifically referred to the cases establishing the principle and worked to distinguish them from the case at hand: "None of the sections of the CAA in which the District of Columbia Circuit has found authority for the EPA to consider costs shares [this provision's] prominence in the overall statutory scheme."'18 Thus the Court was at pains to cite, with apparent approval, the key cases creating the basic principle, and appeared to be saying that the EPA might well be permitted to consider costs if the statute did not expressly forbid it from doing so.
On the other hand, Justice Breyer was clearly concerned that the Court's approach would override the cost-benefit default principle.109 Remember his words: "[i]n order better to achieve regulatory goals ... regulators must often take account of all of a proposed regulation's adverse effects, at least where those adverse effects clearly threaten serious and disproportionate public harm. Hence, I believe that, other things being equal, we should read silences or ambiguities in the language of regulatory statutes as permitting, not forbidding, this type of rational regulation."110 This point was meant as a rejoinder to the majority, which Justice Breyer took to be saying that to allow costs to be considered, Congress was required to be "clear." But at first glance, Justice Breyer's concern seems baseless. The Court was saying only that in view of the clarity of the main provision of the Clean Air Act, judges would be reluctant to find permission to consider costs elsewhere, since Congress "does not alter the fundamental details of a regulatory scheme in vague terms or ancillary provisions-it does not, one might say, hide elephants in mouseholes."111 This is a standard approach to statutory interpretation. It does not suggest that where a statute's "fundamental details" are vague, they will be interpreted to forbid consideration of cost. But it would not be impossible to read the Court's opinion a bit more broadly. Recall that in concluding that the EPA need not consider costs in issuing national standards, the Court emphasized that some provisions of the CAA explicitly refer to costs, and explicitly require them to be taken into account. Here the Court was using the canon of construction "expressio unius est exclusio alterius": the expression of one thing is the exclusion of another. In the particular context of environmental statutes, the "expressio unius" canon could have explosive implications. When Congress does not explicitly refer to costs, agencies may not consider them, and for one simple reason: Congress often does explicitly refer to costs. If the canon is to govern the future, the cost-benefit default principles are in some trouble.
There is a further point. The Court seems to suggest that a statute should not be taken to confer broad discretionary authority on agencies: "[W]e find it implausible that Congress would give to the EPA through these modest words the power to determine whether implementation costs should moderate national air quality standards."ll2 To support the view that ATA is best taken to disallow agencies to interpret ambiguous statutes to allow consideration of costs, it would be necessary to make a simple, two-step argument. First: Statutes should be construed so as to give agencies less rather than more in the way of discretion. Second: A construction of a statute that would allow agencies to decide whether to consider costs significantly increases agency discretion. Now the claim here is not that a statute requiring cost-benefit analysis is itself disfavored on delegation grounds. The claim is instead that an interpretation should be disfavored if the consequence of the interpretation would be to authorize the agency to decide whether to engage in cost-benefit balancing. If this claim is accepted, then the default rule in favor of allowing agencies to consider costs stands as repudiated.
But it is most unlikely that the Court would accept these lines of argument. The "expressio unius" canon can be a useful guide to statutory construction, and the more natural, cost-blind reading of "public health" is certainly supported by the fact that some provisions of the CAA make explicit reference to costs. But here as elsewhere, the "expressio unius" idea should be taken with many grains of salt. If Congress has not, in some ambiguous statutory term, referred to costs, it will often be because Congress, as an institution, has not self-consciously resolved the question whether costs should be considered. The fact that Congress explicitly refers to costs under other provisions is not a good indication that, under an ambiguous text, costs are statutorily irrelevant. This would be an extravagant and therefore implausible inference. The use of the "expressio unius" approach in ATA is best taken as a sensible way of fortifying the most natural interpretation, and not at all as a way of urging that explicit references to cost, in some provisions, mean that costs may not be considered under ambiguous provisions. And if Congress has not resolved the issue either way, the agency is entitled, under standard principles of administrative law, to consider costs if it chooses.13 What about concerns about agency discretion? Agencies are typ-112 Id.
ically allowed to interpret statutory ambiguities,114 and in countless cases in which that principle is invoked, the agency exercises a great deal of discretion over basic issues of policy and principle.115 To allow an agency to decide to consider costs is not to allow it to exercise more discretion than it does in numerous cases. Where the statute is unclear, agencies should be authorized to seek rational regulation; and nothing in ATA suggests otherwise. This is especially so in light of the fact, emphasized by both the Court"116 and Justice Breyer,"117 that the Clean Air Act allows the EPA to consider costs at numerous stages in the implementation process. I conclude that ATA is best taken not to question the cost-benefit default principle, and indeed that the most reasonable reading of the opinion is that the Court has explicitly embraced that principle.

B. NONDELEGATION AGAIN? NONDELEGATION EVER?
At first glance, the Court's nondelegation ruling seems to be a kind of return to the status quo-as an effort to place the doctrine where it has been since 1935: in the constitutional coffin. This is a reasonable reading of the opinion, with the proviso that the doctrine remains available, now as before, for the extreme cases. What makes a case extreme? Apparently an extreme case would be one in which the agency has far more discretion than does the EPA under the "requisite to protect the public health" language. Cases of that sort are, by the logic of the Court's opinion, very few and far between. Schechter Poultry"1 remains good law, and the Court was careful to say that when the area of agency authority is very broad, the statutory principle will certainly have to be intelligible. But the Court's basic message was that its own precedents suggest that almost all nondelegation challenges will be unavailing-and indeed that a mere repetition of any statutory standard will be a sufficient response. 114 See id. "' See, for example, id at 855-56; Babbitt v Sweet Home Chapter of Communities for Greater Oregon, 515 US 687, 708 (1995);Young v Community Nutrition Institute, 476 US 974, 981 (1986). 116 Whitman, 121 S Ct at 910-11.
1. An unusual line of cases. In one sense, however, the Court's nondelegation ruling was no mere return to the status quo, and for a simple reason: It places some long-standing doctrine in disarray. In a series of cases, of which ATA was the culmination, the court of appeals for the District of Columbia circuit has held that a narrowing construction by an agency will be a sufficient and necessary condition for saving an otherwise objectionable delegation of authority.119 The doctrine originated in a challenge to the statute granting the President the authority to fix prices and wages.120 The relevant provisions appeared not to limit the President's discretion-to allow the President to set prices and wages however he chose. In upholding the statute against constitutional attack, the court said not only that Congress had set out an intelligible principle, but also that the executive was obliged to come up with "subsidiary" principles to cabin its own discretion.121 This part of the Court's opinion owed its origins to some imaginative writing from Kenneth Culp Davis.122 The requirement of "subsidiary" principles seemed important to the decision, but it need not be seen as indispensable to it. It would be possible to read the relevant statute, in its context, as sharply limiting executive discretion, and thus to uphold it without relying at all on the need for a narrowing construction by the agency.123 But the basic idea was significantly extended in an important case involving the Occupational Safety and Health Act.124 The relevant provision authorized the agency to set standards "reasonably necessary or appropriate" to provide safe and healthful places of employment 725-30 (1969). 123 In brief: The wage and price freeze was a response to the perception that the nation was in the midst of "cost-push" inflation. That was the statutory background. If that is what the statute was about, then the President could not play favorites, or reward his friends and punish his enemies. I say a bit more on this issue below. ( delegation of authority? Would it be sufficient to say that the statute told the agency to do no more, and no less, than was "reasonably necessary or appropriate"? The court of appeals did not think that that would be sufficient. It was concerned that the statutory terms could mean any number of things. They could mean, for example, that OSHA should engage in cost-benefit analysis; perhaps a standard is not "reasonably necessary or appropriate" unless the benefits justified the costs. Or they could mean that OSHA was supposed to regulate all "significant risks" to a maximally protective point, subject perhaps to a constraint that the regulation be "feasible" for industry.126 Or they could mean something else. Because of the statutory terms' apparent plasticity, the court was obviously tempted to strike down the statute on nondelegation grounds. But instead of doing so, the court remanded the case to OSHA, concluding that the agency could save the statute with a narrowing construction. On remand, the agency generated what the court found-barely-to be a sufficient response.127 According to the agency, the statute required it to regulate only "significant risks," and only to the point of "feasib[ility]," and within those constraints the agency was required to select the standard that would be most protective of workers.128 The court said that this was enough to satisfy the constitutional concern.
Notwithstanding the Supreme Court's ruling in ATA, the approach of the court of appeals is not impossible to understand. If we are concerned about an absence of accountability, and also about values associated with the rule of law, a narrowing construction at the agency level can be important. Such a construction can expose the agency's standard to public oversight and review; it increases transparency and to that extent accountability. And by ensuring that agency action will be undertaken pursuant to a clear standard, a narrow construction can go a long way toward alleviating the concern of arbitrary, unpredictable agency action, treating the similarly situated differently. But we should be careful not to say that the purposes of the nondelegation doctrine are the doctrine itself, or to dissociate the doctrine from Article I, its legal source. At most, a narrowing construction can be helpful when the 1]

REGULATING RISKS AFTER ATA 33
Court is otherwise in equipoise. The nondelegation concern is not eliminated by such a construction. This is basically the Court's response in A TA.129 And it is evident that after ATA, the court of appeals' approach is entirely unacceptable. The Supreme Court has made clear that a narrowing interpretation by the agency cannot save an otherwise objectionable delegation. And in the end the Court's reasoning does seem to make sense. If the problem is that Congress has failed to lay down standards for agencies to follow, how can the agency's own standards resolve that problem? The question seems all the harder to answer in light of the fact that the source of the nondelegation doctrine is Article I, section 1, which vests legislative power in a "Congress of the United States." The purpose of the nondelegation doctrine, it would seem, is to require Congress to legislate.
With respect to that requirement, agency narrowing is neither here nor there. 2. The line revisited-and OSHA's fate. How, then, should the now-rejected court of appeals decisions be understood? Might the Court have (inadvertently?) given a small boost to the nondelegation doctrine by suggesting that a narrowing construction cannot be helpful? These questions are not simple to answer.
The case of freezing wages and prices is the easier to handle.
To be sure, Congress did not give clear standards in the text of the statute. But statutory terms receive meaning from context, and the context behind the Act suggested a desire to protect the nation from a certain kind of inflationary pressure, captured in the notion of "cost-push inflation," in accordance of which unions and employers create a kind of psychological spiral, one that needs to be broken through law.130 To be sure, the statute itself did not refer to this theory, and perhaps a committed textualist would want to strike down the statute because of its open-ended terms. But it is clear, from context, that Congress did not seek to give the President the authority to freeze prices and wages in a way that involved political favoritism. If the statute should be interpreted so as to avoid the constitutional difficulty, it would not be an intolerable stretch to say that any executive action should be reviewed with the particular context in mind-and hence that statutory purpose sharply limited presidential discretion. Perhaps it will be responded that Congress should be required to say all this clearly in the statutory text itself. But would much be gained by requiring this step? This is far from obvious. Thus the lower court's decision in the case of wage-price freezes probably remains good law, on the grounds that: (1) the statute should be construed so as to be constitutional rather than the opposite, and (2) the key provisions, read in their context, do not allow the President to play favorites, but on the contrary require him to act in accordance with the statutory purpose as understood in the context that motivated it. Note that the point here is not that courts are supplying, on their own, an intelligible principle to discipline agency discretion. Instead I am urging that the statute, taken in its context, supplies that principle.
The contested provision of OSHA is much harder. The phrase "reasonably necessary or appropriate" seems, on its face, to leave everything open. To be sure, the phrase might also appear similar to the standard that the Court found sufficient in ATA. But there is a big difference. In ATA, the Court held, immediately before upholding the statute against nondelegation attack, that the statute required a "health only" determination, and that it did not allow consideration of costs."' The ATA Court held the statute constitutional in part because (in its view) Congress was clear on this point; Congress itself decided that costs would not count, and hence the agency was not permitted to create the legal standard out of thin air. It seems to follow that if a statute itself requires consideration of costs, and hence a form of cost-benefit balancing, it would also be constitutional, notwithstanding the high levels of residual discretion that would remain. Indeed, no one seems to argue that there is a nondelegation problem with a statute that calls for costbenefit balancing, even though significant discretion is left with the agency.
What makes the relevant provision of OSHA much harder is that it seems to give the agency discretion to decide whether the statute does or does not allow consideration of costs-and thus to decide what the statutory standard is. This was Chief Justice Rehnquist's objection to another provision of OSHA132-an objection that was rejected on the ground that that provision could be authoritatively construed sharply to discipline the agency's discretion.133 Nothing in ATA, in short, resolves the question whether a court should uphold a statute that leaves the agency the authority to construct its most fundamental meaning, by deciding, for example, whether the statute requires the agency to show that benefits justify costs, that a significant risk has been shown, or something else. It would therefore be possible to conclude that the "reasonably necessary or appropriate" language is too open-ended, because it allows the agency to decide, without any real limits, on the substance of the statute that it is implementing.
It is important to be careful with this point. As we have seen, lower courts have created an interpretive principle authorizing agencies to consider costs if they see fit, and ATA seems to approve of this principle. It would be implausible to suggest that a statute is unconstitutional if it allows an agency to decide whether or not to consider costs; this kind of decision has been found acceptable in many contexts, and ATA cannot be taken to say that the underlying statutes are now unconstitutional. After A TA, the appropriate answer to the constitutional question is this: Statutes that allow agencies to decide whether to consider costs admittedly confer considerable discretion. But after ATA, this is the furthest thing from fatal. The key provisions of the relevant statutes typically contain sufficient limits on agency discretion, and typically these provisions govern a small domain of the economy,134 a relevant factor under A TA. Even if agencies are allowed to decide whether or not to consider costs, they do not, under those statutes, have anything like a blank check.
But OSHA's "reasonably necessary or appropriate" language seems quite different on these counts. OSHA's authority involves no small domain, and in the abstract, the "reasonably necessary or appropriate" language allows the agency to choose the statutory 132 See Industrial Union Department, AFL-CIO v American Petroleum Institute, 448 US 607, 671 (1980) (Rehnquist concurring in judgment) (objecting on nondelegation grounds to 29 USC ? 655(b)(5)).
'33 See id at 639-46 (Stevens) (plurality) (interpreting provision in a way that implicitly rejects the nondelegation challenge); see also Whitman, 121 S Ct at 912-13 (interpreting the Benzene Case as rejecting nondelegation challenge). standard. The significance of ATA is that if this conclusion is to be avoided, it cannot be for the reason invoked by the court of appeals. It must be because the statute is best construed to cabin agency discretion to some degree. We know, for example, that the statute bars measures that are unreasonable, or inappropriate, as means of achieving safe workplaces. In this way the statute can be taken to require cost effectiveness, and also to require the agency to pursue the end of worker safety. But is this enough? The end of worker safety can be pursued in multiple ways. To be sure, the constitutional doubt would be removed, under ATA, if the statute were construed to require cost-benefit balancing.135 We can easily imagine a judicial opinion that would so construe the statute; perhaps a step is "reasonably necessary or appropriate" if and only if it survives balancing, all things considered. But the court of appeals was probably right to say that this interpretation, while possible, is not ordained by the statutory text. And it is because the interpretation is not ordained that the court remanded the case to the EPA for an authoritative construction. The problem, after ATA, is that such a construction is, with respect to the nondelegation issue, neither here nor there.
In these circumstances, future courts have only three options. The first would be a version of the route taken in ATA: to point to other statutes giving agencies broad discretion (such as the "public interest, convenience, and necessity" standard of the Federal Communications Commission) and to urge that the disputed provision is not much different. The difficulty with this approach is that it would be somewhat irresponsible to invoke the analogies, without making some effort to show how the agency is not permitted to do whatever it chooses. The second option would be to generate an interpretation of the provision that adequately cabins agency discretion. This would certainly be possible, but the strategy in ATA-merely repeating the statutory language, with the added words "no less and no more"-would not be helpful. Certainly an interpretation that requires cost-benefit balancing would be adequate. The third option would be to strike down the statute on nondelegation grounds. It is ironic but true that this route may have been made more rather than less likely as a result of the Supreme Court's rejection of the approach of the court of appeals.
Should the court take this step? In general, I think that the ATA Court was entirely correct to suggest that the nondelegation doctrine deserves little place in modern constitutional law. For this reason, some combination of the first and second routes would probably be best: an effort to construe the statute to impose some limitations, perhaps by calling for cost-benefit balancing, alongside a recognition that a great deal of discretion is constitutionally legitimate. But if the nondelegation doctrine deserves any place at all, the "reasonably necessary or appropriate" language, in context, would not be the worst imaginable place for judicial invalidation.

REFERENCE TO PARTICULATES AND OZONE)
After ATA, a major question is how a plaintiff might be able to challenge a national standard, if costs cannot be considered and if the constitutional route is unavailable. With its unhelpful "not lower or higher" language, ATA offers only a little guidance here."36 If the regulation is less stringent than is "requisite to protect the public health," with an "adequate margin of safety," it will be unlawful. If the regulation is more stringent than is "requisite," it will also be unlawful. In the easy cases, at least, the lessons are clear. A regulation will be subject to challenge, as insufficiently stringent, if it allows significant adverse health effects. If we accept the EPA's evidence (see Appendix, Table Al), it would therefore be reasonable to argue that the EPA was required, and not merely permitted, to produce a new and perhaps even more stringent regulation for particulates. A regulation will also be subject to challenge if significant adverse effects cannot be expected at levels that the EPA forbids. On a reasonable reading of the evidence governing ozone (see Appendix, Table A2), the new EPA regulation might be unlawful for that reason.
In the end I will urge that the particulates rule should probably be upheld, but that the ozone standard should probably be remanded, so that the EPA can give a better, more quantitative ex-planation of why it chose the particular regulatory "points" that it selected. This judgment is tentative and highly dependent on the details. What is more important than the conclusion is an appreciation of the three possible ways in which any national ambient air quality standard, including the particulates and ozone standards, might be analyzed in court.
1. A soft look. The court might say that there is substantial scientific evidence to support the view that both pollutants produce significant adverse health effects at currently permitted levels-and hence the new controls are, in the administrator's reasonable judgment, "requisite to protect the public health." The statutory requirement of an "adequate margin of safety" might well be taken to support this view. As I have emphasized, the evidence supporting regulation of ozone seems a good deal weaker than the evidence supporting regulation of particulates, especially if we take into account the fact that ground-level ozone seems to have nontrivial health benefits. But perhaps a court should say that there is much scientific uncertainty here, and that the EPA should be allowed to resolve the doubts as it sees fit. If the court took this route, it would be following the direction established in the Lead Industries case, in which the EPA was given a great deal of room to maneuver.137 There are several advantages to this approach, especially if we consider the institutional role of the courts. A serious problem with intense judicial review of agency action is that it creates delay-and hence ensures a bias in favor of the status quo. 138 In light of the inevitable scientific uncertainties, it should be exceptionally easy for a skillful litigant to challenge a national standard as either too high or too low. 139 On the basis of the EPA's own data, an environmental group would have had a quite plausible argument that the regulation of particulates was insufficiently stringent under the statute.14 In order to allow agencies room to maneuver in the face of scientific uncertainty, it would be reasonable to say that on the basis of minimally plausible evidence, courts should simply uphold the relevant decisions. A special virtue of this approach is that the Bush Administration would be permitted to come to a different conclusion from the Clinton Administration, and viceversa, because different judgments of value could lead to different conclusions about how to proceed in the face of ambiguous science.
2. Evidentiary review. A second possibility, hinted at above, would be to invalidate the ozone regulation while upholding the regulation of particulates. The simple claim here would be that on the evidence given, the new particulates standard was shown to be requisite to protect the public health-but the new ozone standard was not, especially if we take into account the health benefits of ground-level ozone.'41 We have seen the possibility that, all things considered, the regulation would increase rather than reduce health problems.
Justice Breyer's opinion rightly suggests that under the "requisite to protect the public health" language, the EPA is not supposed to remove all risks from the air, or to make the air "riskfree."142 The EPA should take account of context to compare the risk at hand to risks that people face in ordinary life. The statistical risks from low levels of ozone do appear smaller than the statistical risks that people find acceptable in multiple domains (see Appendix, Table A2). Particulates are very different on this count. Here the existing hazards do seem high, on a plausible reading of the evidence. By upholding the particulates standard, and asking the EPA to explain itself more thoroughly with respect to ozone, the court would be contributing to the development of a kind of common law of acceptable risks, of the sort that Justice Breyer seemed to be encouraging.
The approach I am suggesting-upholding the particulates standard as requisite, while invalidating the ozone standard as not shown to be requisite-would certainly be reasonable. The principal objection would be institutional; it would involve the special 141 1 am not attempting here to reach definitive conclusions about the scientific data. I am simply suggesting how a reviewing court might reasonably respond to the data that the EPA has compiled. 142 Whitman, 121 S Ct at 924 (Breyer concurring in part and concurring in judgment), quoting the Benzene Case, 448 US at 642. limitations of judicial review. The evidence shows the possibility of nontrivial health gains from the ozone regulation, and in the face of scientific uncertainty, the agency should be permitted to make whatever (reasonable) policy choices it likes. Especially in view of the risk of status quo bias, perhaps the court should refuse to invalidate a judgment like that in the ozone case, even if the judgment seems wrong.
There is a further set of questions. Suppose that the new particulates standard should be, and is, upheld. At that stage, the EPA will have to decide what, exactly, to regulate, and to do this it will have to decide what fine particulates consist of. This is not a simple question. Currently the EPA is focusing on nitrogen oxides and sulfur oxides, but ignoring mobile sources, even though mobile sources appear to be emerging as the principal source of fine particulates. Ideally, the EPA should be able to create a trading system for the precursors for fine particulates, just as it has for ozone, and just as the Clinton Administration proposed for control of global warming.143 But the EPA has refused to attempt this step, and for one simple reason: It does not know what the precursors are, or at least how they relate to one another. There are many complexities here. Ideally, an agency that is sensitive to the need for high benefits and low costs will try to design a system that promotes regulatory goals at the lowest possible expense. But in view of the technical complexities, it is not clear how much a court can do to require such an approach here.
3. Requiring a clear standard. A third possibility would be to invalidate both regulations as arbitrary or as inadequately justified, not because the risks are too low, but because the EPA did not explain on what grounds it chose these particular regulations, rather than regulations that would be somewhat more strict or somewhat more lenient. This would be the administrative law analogue of what the District of Columbia Circuit took to be a constitutional requirement in ATA.'" The Supreme Court's holding that the nondelegation doctrine does not require this form of specificity says nothing about whether such specificity might be required as a matter of administrative law. If the EPA cannot explain, in concrete terms, why it chose the particular levels it chose, how can courts know that the agency's decision was not arbitrary?
In doctrinal terms, judicial invalidation on these grounds might take one of two routes. First, the court might invoke the statutory language and say that it cannot know whether the particular level chosen is "requisite to protect the public health" unless it has a clear sense of why the EPA reached that conclusion. Without numbers and criteria, it is impossible to obtain any such sense. Second, the court might put the statutory language to one side and say that it cannot tell whether the agency's action is arbitrary or capricious, within the meaning of the Administrative Procedure Act,145 unless the EPA has given a more detailed explanation of its choice. Either of these conclusions would be relatively conventional, and neither would mark a huge departure from current law.
But there would be a genuine innovation here. Thus far, courts have not required anything like a quantitative basis for health and safety regulation. Sometimes they have required agencies to show that the costs are not grossly disproportionate to the benefits;146 sometimes they have prohibited agencies from acting when it seems as if there are not benefits at all.147 But the relevant decisions have been more qualitative than quantitative, and when they have been quantitative, the overall judgment has seemed overwhelmingly clear.148 It would be a significant step from these decisions to a holding that agencies must quantify the effects of pollutants at various levels, so as to explain, in specific terms, why one level was chosen rather than another.
Would the step be worthwhile? While I cannot discuss the issue in detail, I believe that it would be, at least for the ozone rule, where the evidence of harm is relatively thin. Such a ruling would not impose an unacceptable informational burden on the EPA. In fact, the EPA routinely gathers enough information to provide the necessary explanation. At the same time, such a requirement would provide a useful spur to the agency, one that would also produce a higher degree of rationality and coherence. The result would be to show when, and why, environmental groups or industry would be able to mount a successful challenge to an ambient air quality standard.
In sum: On remand, the court should probably uphold the particulates standard, on the ground that on a reasonable view of the evidence, the agency had sufficient basis to conclude that that standard was "requisite to protect the public health," even without an attempt to quantify. At the same time, the court should probably remand the ozone rule, on the ground that the EPA has not given an adequate explanation of why that rule is "requisite." The court should encourage the EPA to be as quantitative as possible. And on remand, the EPA should take up the invitation, attempting in the process to give a clear sense, for the first time, of why it has chosen one regulatory "point" rather than another. But my principal goal has not been to urge any particular result, which will obviously turn on close engagement with the record. I have attempted instead to give a sense of the arguments that are available, post-ATA, and a general sense of the grounds on which one might choose among them.

V. CONCLUSION
The Court's principal rulings in ATA represented a return to the status quo-a rejection of some imaginative suggestions about how to read both the Clean Air Act and the Constitution. The Court established, quite correctly under existing law, that national standards should be set without regard to cost, and that the nondelegation doctrine has a small place in constitutional doctrine-or perhaps no place at all.
At the same time, the Court's unambitious opinion offered an almost comically vague interpretation of the key provisions of the Act-an interpretation that will greatly frustrate those who seek to know what, exactly, the EPA is supposed to do. Indeed, the Court left a large number of questions unresolved. There is some ambiguity in the opinion, but it is best taken not to question, but on the contrary to endorse, the cost-benefit default principles developed by lower courts. And while the Court showed no interest in the nondelegation doctrine, its rejection of the approach of the lower court now makes it impossible to invoke an agency's narrowing construction to support an otherwise objectionable delegation. For that reason, the nondelegation doctrine is not quite dead. I have attempted to show that the little life that remains in the nondelegation doctrine might support a plausible constitutional attack on a central provision of the Occupational Safety and Health Act.
ATA also leaves open a number of challenges to national ambient air quality standards, including the very standards at issue in the case. The principal challenges would involve the language of the relevant provision of the CAA ("requisite") and the arbitrary or capricious standard of the APA. On the basis of the evidence before the agency, it would make sense for a court to uphold the particulates standard as having been reasonably judged "requisite" while also invalidating the ozone standard as not shown to qualify as such. For the EPA itself, it would certainly make sense to move in the direction of greater quantification, in which national standards are issued only after an effort to specify the expected benefits, to compare them with the expected benefits of alternatives, and in that way to produce clear standards for choosing appropriate levels of ambient air quality. This is essentially an administrative task, not one for the courts. But it would be entirely appropriate for courts to spur regulatory agencies in this direction. Justice Breyer wrote only for himself, and his pragmatic, consequence-centered concurring opinion attracted no additional justices; but I believe that it will exert an enduring influence on the law of risk regulation. If so, the ATA decision will stand not only as a responsible resolution of the principal questions in the case, but also as a modest step toward a more sensible system of environmental protection. APPENDIX This appendix offers the EPA's own findings about the effects of its proposed particulates and arsenic regulations, and of alternative approaches. The tables are taken verbatim from Innovative Strategies and Economics Group, Regulatory Impact Analysis for Particulates and Ozone Regulations (1998).  c Partial attainment benefits based upon postcontrol air quality as defined in the control cost analysis. a All estimates are measured incremental to partial attainment of the baseline current ozone standard (0.12 ppm, 1 expected exceedance per year). b The results for .08, 5th and .08, 3rd max. are only for the high-end assumptions. The low-end estimates were not calculated for these alternatives.
c Partial attainment benefits based upon postcontrol air quality estimates as defined in the control cost analysis.