Avoiding Absurdity? A New Canon in Regulatory Law

Cass R. Sunstein

The author is the Karl N. Llewellyn Distinguished Service Professor of Jurisprudence, Law School and Department of Political Science, University of Chicago. Professor Sunstein is grateful to Adrian Vermeule for extremely helpful discussions.

[32 ELR 11126]

A New Principle

In the last two decades, federal courts have been developing a new and thus far unacknowledged canon of regulatory law: As a general rule, administrative agencies will be taken to have the authority to interpret statutes so as to avoid absurd or patently unreasonable results, even if the interpretation is hard to square with the literal language of the statute. This canon authorizes agencies, and in particular agencies that regulate the environment, far more flexibility in the interpretation of literal language than courts are now permitted to have. My narrow goal in this Dialogue is to describe and to defend this canon. My broader goal is to use the canon as a basis for urging that contemporary theories about interpretation go wrong by emphasizing large claims about democracy and legitimacy at the expense of an inquiry into the real-world capacities of our various institutions, including the U.S. Environmental Protection Agency (EPA) and the federal courts.

The new canon has old roots in the time-honored idea that courts will not construe statutes to produce absurdity.¹ But that notion remains highly controversial, at least in some applications,² in part because of a belief that much mischief might follow if courts attempt to avoid what they consider to be absurdity. For reasons to be elaborated, the canon that I am defending should be endorsed even by those who believe that courts have no business departing from statutory language.³ Compared to courts, regulatory agencies have a high degree of specialized competence and a large measure of political accountability. They are in a good position to know if a departure from literal language will unsettle the regulatory scheme. Because of these characteristics, courts should allow agencies to avoid absurdity, in the face of literal text, even in cases in which courts should not themselves exercise this power—and to allow agencies more freedom to assume this canon than courts.

I will also connect the new canon to some of the largest and most intense of current debates about legal interpretation, especially in the context of regulatory law and environmental protection. In particular, I will urge that it is important to understand interpretive disputes not by reference to abstractions, but in the light of institutional considerations—by asking who is good at what, and by asking about the effects, over time, of one or another approach to interpretation.⁴ Consider, for example, the debate over formalism: the view, defended most prominently by Justice Antonin Scalia, that legitimate interpretation requires fidelity to the ordinary meaning of the relevant text when originally enacted.⁵ Some people, including Justice Stephen Breyer, reject this view, arguing legitimate interpretation includes a variety of devices designed to ensure that sense, rather than nonsense, is made of the law.⁶ Often this debate seems highly abstract and conceptual, involving contests over big ideas like democracy, legitimacy, authority, and constitutionalism.

The last decade, however, has witnessed a kind of "institutional turn,"⁷ in which it is starting to be seen that debates over the best approach to interpretation are debates about consequences and, in particular, about institutional capacities.⁸ The controversy over formalism might itself be seen in these terms. If courts adopt formalism, will the system of regulatory law be better or worse? If courts, abandoning the ordinary meaning of the text, are likely to blunder, the argument for formalism is greatly strengthened. That argument would also be strengthened if our legislatures, faced with a formalist judiciary, would be likely to anticipate problems before they arise and to correct them after the fact. In that event, the problems introduced by formalism, for regulatory law or environmental protection, would be reduced by legislative corrections. But if courts, abandoning formalism, are likely to make the law far more sensible, and if legislatures cannot be expected to anticipate or correct errors, the argument for formalism is greatly weakened. If we emphasize institutional capacities, we will see that in deciding about the appropriate approach to interpretation, two issues are central. The first involves the capacity of the relevant interpreters. The second involves the systemic effects of one or another approach to interpretation.

These general points very much bear on the canon that I am identifying here. In the context of apparent absurdity, reasonable people can disagree about whether and to what extent judges should depart from literal language. But as we shall see, the issue is quite different when agencies are seeking to make that departure. This point has large implications for interpretation in the context of regulation and environmental [32 ELR 11127] protection. It suggests that regulators, simply because of their comparative advantages, should be allowed to repudiate or to endorse formalism as they see fit.

The new canon of interpretation is best understood in light of certain pathologies of the largest development in the last century of American law: the rise of the modern regulatory state. In controlling risks to safety, health, and the environment, and in regulating banking and communications, modern legislatures issue countless rules. Some of them are vague; some of them are precise and detailed. In either case, legislatively enacted rules sometimes amount to "thin simplifications" that often extend over decades and that must be engrafted onto complex systems of natural and human interactions.⁹ Simply because of human limitations, these systems cannot be well understood by legislators at the time of enactment. As a result, thin simplifications threaten, on occasion, to produce harm or even disaster, not least from the standpoint of their proponents. The harms can involve safety, health, and the environment itself, as when a statutory enactment, designed to protect air or water quality, has the unintended effect of doing the opposite in one or another application. I will not be able to discuss the point in detail in this space; but there are many ways of reducing the relevant risks. One of the simplest, I suggest, is to authorize agency officials not to reject enacted law, but to abandon a literal approach to interpretation by adapting statutory terms to unanticipated circumstances. By abandoning literalism where it would produce absurdity or patent unreasonableness, agencies, like the courts that do the same, engage in a legitimate form of interpretation—one that occurs in households and workplaces as well as in government agencies and courtrooms.

Some pertinent clarifications before we begin: the canon authorizes agencies to avoid absurdity; it does not compel them to do so. There are systemic reasons why an agency might want to stick to the literal language and to allow absurdity in particular cases. As we shall see, an agency might want to do this if it believes that fidelity to the literal language is the best way of promoting planning and reducing confusion for the future. In addition, the principle is limited to cases of patent unreasonableness or unintended absurdity. These cases usually arise in the context of excessive generality, a problem that occurs when statutory language operates sensibly and as intended in most of its applications, but produces patently unjustifiable and unintended results in a particular case. This case might be too strange or exotic to have been anticipated in advance; or it might be a product of changed circumstances, which makes literal language, never or almost never absurd when enacted, absurd because of new developments. I emphasize that the principle that I am discussing should not be seen as authorization to ignore statutory requirements; it applies only when literal language produces an outcome that was not anticipated and could not plausibly be defended. One of my main goals is to show the difference between ignoring statutes and counteracting this problem through an acceptable form of interpretation.

The Emerging Principle

For regulatory law, the interpretive background is established by Chevron, U.S.A., Inc. v. Natural Resources Defense Council, Inc.,¹⁰ by far the dominant case in the area, which sets out a two-step inquiry for judicial review of agency decisions. The first question (step one) is whether the U.S. Congress has "directly decided the precise question at issue"—more simply, whether Congress has unambiguously banned what the agency proposes to do.¹¹ Under Chevron, agencies are generally permitted to construe ambiguous statutes as they see fit. Under step two, it remains to ask whether the agency's interpretation of the statute is reasonable.¹² I am focusing on step one here: Will the agency be taken to have the power to depart from the literal language of the statute, if its reason for doing so is to avoid absurdity? Many cases suggest an affirmative answer.

Lead and Pipes

The leading case is American Water Works Ass'n v. EPA.¹³ The case involved a creative effort by EPA to regulate lead in drinking water. The details are somewhat technical, but the outcome is of general interest. The court allowed EPA to produce an outcome that flew in the face of statutory language,but that, on a reasonable view of the facts, any sensible person would prefer. The court permitted EPA to stretch the statutory language in order to ensure against an evidently absurd outcome.

The Safe Drinking Water Act (SDWA)¹⁴ requires EPA to produce maximum contaminant level goals (MCLGs) for water contaminants. These goals must "be set at the level at which no known or anticipated adverse effects on the health of persons occur," with an adequate margin of safety.¹⁵ Because no safe threshold had been established, EPA's MCLG for lead was zero. Once an MCLG is established, EPA is required to set a maximum contaminant level (MCL), "as close to the maximum contaminant level goal as is feasible."¹⁶ EPA is authorized not to set an MCL, and to require "the use of a treatment technique in lieu of establishing" that level, if (and only if) it finds "that it is not economically or technologically feasible to ascertain the level of the contaminant."¹⁷

For lead, then, EPA would be expected to set its MCL as close as "feasible" (economically and technologically) to the MCLG of zero, except if it was not "feasible" to ascertain the level of lead contamination (and no one urged that the task of ascertainment was not feasible). But this is not what EPA did, because of some distinctive features of the lead problem. Source water is basically lead-free; the real problem comes from corrosion of service lines and plumbing materials. With this point in mind, EPA refused to set any MCL for lead. EPA reasoned that an MCL would require public water systems to use extremely aggressive corrosion control techniques, which, while economically and technologically "feasible," would be counterproductive, because they would increase the level of other contaminants in the water. What appeared to be the legally mandated solution [32 ELR 11128] would make the water less safe, not more so. EPA therefore chose a more subtle and modest approach. Instead of issuing an MCL, it required all large water systems to institute certain corrosion control treatment, and required smaller systems to do so if and only if representative sampling found significant lead contamination.

EPA did not contend that an MCL was not "feasible" to implement, nor did it argue that it was not "feasible," in the economic or technological sense, to monitor lead levels in water. Nonetheless, the court upheld the Agency's decision.¹⁸ The court accepted EPA's seemingly implausible suggestion that the word "feasible" could be construed to mean "capable of being accomplished in a manner consistent with the Act."¹⁹ The court said that "case law is replete with examples of statutes the ordinary meaning of which is not necessarily what the Congress intended." It added, pointedly and controversially, that "where a literal meaning of a statutory term would lead to absurd results," that term "has no plain meaning."²⁰ Because an MCL would itself lead to more contamination, "it could lead to a result squarely at odds with the purpose of the SDWA."²¹ The court therefore accepted EPA's view "that requiring public water systems to design and implement custom corrosion control plans for lead will result in optimal treatment of drinking water overall, i.e., treatment that deals adequately with lead without causing public water systems to violate drinking water regulations for other contaminants."²²

It should be clear that the court permitted a quite surprising and even countertextual interpretation of the SDWA. The critical point in the court's opinion is the suggestion that a statutory term lacks a plain meaning when the literal language would produce "absurd results." The statutory terms seem to make no room for EPA's refusal to issue an MCL. In upholding EPA's refusal, the court authorized the Agency to avoid absurdity.

Trifles

Sometimes regulation appears to require government to control small or trivial harms. To the extent that this is so, two problems might arise. The first is that significant private and public expense might be imposed for no real gain—itself an absurdity. The second is that people might be less safe on balance—if, for example, the product or process that causes a trivial risk is less hazardous than a product or process that is already on the market, and that will not be regulated in the future. In that case, regulation of a small or trivial risk would seem to threaten to produce absurdity from the standpoint of the statute's proponents. Can agencies provide correctives?

In a series of cases, the U.S. Court of Appeals for the District of Columbia Circuit has developed a principle authorizing (not requiring) agencies to make de minimis exceptions to regulatory requirements. The initial case was Monsanto Co. v. Kennedy.²³ There, the Agency banned acrylonitrile on the ground that it counts as a "food additive," migrating in small amounts from bottles into drinks within bottles. The Food and Drug Administration (FDA) concluded that the ban was justified on safety grounds, a conclusion that the court found inadequately justified. But what is more important in the case is the general language with which the court remanded the case to the FDA. The court stressed that the Agency had discretion to exclude a chemical from the statutory definition of food additives if "the level of migration into food . . . is so negligible as to present no public health or safety concerns."²⁴

A related case presented the question whether EPA was permitted to make categorical exemptions under the prevention of significant deterioration program of the Clean Air Act (CAA).²⁵ Here, the court spoke in far more ambitious terms, showing considerable enthusiasm for de minimis exemptions. It announced that

categorical exemptions may be permissible as an exercise of agency power, inherent in most statutory schemes, to overlook circumstances that in context may fairly be considered de minimis. It is commonplace, of course, that the law does not concern itself with trifling matters, and this principle has often found application in the administrative context. Courts should be reluctant to apply the literal terms of a statute to mandate pointless expenditures.²⁶

In fact the court expressly connected this principle with the idea that the court should "look beyond the words to the purpose of the act" to avoid "absurd or futile results."²⁷ Thus the court concluded, in its broadest statement on the point, that "most regulatory statutes, including the [CAA]," permit de minimis exemptions upon an adequate factual showing.²⁸

Here, then, is an explicit recognition of agency authority to prevent patent unreasonableness by exempting tiny risks from regulatory controls—even if the exemption seems to fly in the face of the statutory terms. The recognition is best seen as permission to agencies to avoid regulatory absurdity, and in particular the sort of absurdity that comes from "thin simplifications" in the form of rigid regulatory commands.

Costs

Suppose that a regulatory statute appears to ban agencies from considering costs. Suppose that any such ban would produce serious problems, not because the ban is itself absurd, but because the ban prevents the agency from adopting a creative approach that would, in a sense, involve more aggressive regulation.

This appeared to be the problem in Michigan v. EPA.²⁹ At issue there was an EPA decision to approve a state implementation plan (SIP) for the regulation of ozone. The key statutory phrase provided that SIPs must contain provisions adequately prohibiting "any source or other type of emissions activity within the state from emitting any air pollutants in amounts which will . . . contribute significantly to [32 ELR 11129] nonattainment in, or interfere with maintenance by, any other State with respect to any such national primary or secondary ambient air quality standard."³⁰ At first glance, this provision might well be read as a kind of absolute ban on "significantly contributing" pollutants. But EPA did not understand it that way. Instead, EPA reached an apparently ingenious conclusion. It would adopt a low threshold for deciding whether a contribution was "significant." In this way, many states would count as significant contributors. But the significant contributors would be required to reduce their ozone only by the amount achievable via "highly cost-effective controls,"³¹ meaning those that could produce large reductions relatively cheaply. In states with high control costs, then, relatively low reductions would be required. In states with low control costs, substantial reductions would be mandatory.

Was this unlawful? Challenging EPA's interpretation, the plaintiffs urged that the statute banned any consideration of costs at all. In their view, "contribute significantly" made no room for an inquiry into the costs of compliance. The court rejected the argument, finding no "clear congressional intent to preclude consideration of costs."³² But the court obviously had a difficult time with the statutory phrase "contribute significantly," which seems to refer to environmental damage, not to environmental damage measured in light of cost. In upholding EPA's decision, and in ruling that EPA would be allowed (not required) to do what it did, the court insisted that significance should not "be measured in only one dimension," that of "health alone." In fact, in some settings, the term "begs a consideration of costs."³³ In the court's view, EPA would be unable to determine "'significance' if it may consider only health," especially in light of the fact that ozone causes adverse health effects at any level. If adverse effects exist on all levels, how can EPA possibly choose a standard without giving some weight to cost?³⁴ In any case "the most formidable obstacle" to a ban on consideration of cost "is the settled law of this circuit,"³⁵ which requires an explicit legislative statement to preclude consideration of cost. The best way to understand this decision is to see it as an effort to authorize the Agency to adopt a solution that would avoid patent unreasonableness.

Here, as in American Water Works, the court permitted the Agency to read the statutory text aggressively, on the theory that the Agency's approach was so much more sensible, and so much less unreasonable, than the approach that would be required by textualism. It seems clear that the Congress that enacted the SDWA could not foresee the special problems created by removing lead from water, problems that, EPA plausibly argued, would make an MCL counterproductive. So too for the nonattainment program: from every point of view, EPA's effort to require only cost-effective controls seemed better than an effort to define "contribute significantly" in a cost-vacuum.³⁶ Thus the Agency was allowed to depart from the ordinary meaning of the statutory language, as court perhaps would not be; and the reason was to enable the Agency to avoid an outcome that seemed absurd in light of the statutory variables and common sense.

The Canon Defended

I now explore the legitimacy of the new canon. Why, exactly, are agencies being authorized to avoid absurdity? Why are they being authorized to approach statutory text more aggressively than courts? To approach these questions, it is necessary to know something about the theory of legal interpretation. No one should deny that a sensible legal system might well allow interpreters—courts or administrators—to abandon literal language in cases of absurdity. This is a familiar and conventional aspect of legal interpretation in the United States, England, and elsewhere. Such a system would abandon textualism in some cases. But whether to abandon the text, and how often to do so, are questions that reasonable people can dispute. My basic claim here is that a decision to follow literal language, or instead to allow exceptions for absurdity or patent unreasonableness, calls for an inquiry into (a) the capacities of the interpreter, and (b) the effects, on the relevant system of law, of one or another approach. These issues, typically neglected in theories of interpretation, help show why agencies might be permitted to make exceptions for absurdity even if courts are not. To see this point, some jurisprudence is necessary.

Jurisprudence as Pragmatic, Not Metaphysical

I have emphasized that according to a time-honored principle of interpretation, courts should not construe statutes so as to produce absurd results. In a way this old principle is entirely familiar. In ordinary life, as in law, the literal meaning of words will often produce unintended absurdity. A parent tells a child: "Don't leave the house under any circumstances"; but if the house is on fire, the child had better leave. Someone tells his best friend: "Do not tell anyone where I am going this weekend"; but under emergency conditions, disclosure might well be justified. Building on an understanding that literal language is not always the best guide to meaning, courts have made unintended absurdity a central theme in two of the most controversial and celebrated cases in American law: Riggs v. Palmer³⁷ and Church [32 ELR 11130] of the Holy Trinity v. United States.³⁸ In Riggs, a court prohibited a murderer from inheriting from the estate of the murder victim, even though the language of the governing law made no exception for murderers, and seemed to require the murderer to inherit. In Church of the Holy Trinity, the U.S. Supreme Court refused to ban a church from paying for the importation of a rector into the United States, even though the literal language of the statute seemed to ban the importation.

These are famous cases, but they are hardly uncontroversial.³⁹ Is it so clear that courts should be permitted to allow exemptions for absurd outcomes? How might we answer that question? There is a conventional answer, and we can take the analysis of the legal philosopher H.L.A. Hart as both classic and representative. Hart appears to welcome judicial corrections for unintended absurdity, finding them necessary in light of our inevitable "inability to anticipate"⁴⁰ the countless contexts to which statutory terms will have to be applied. In Hart's account, a "feature of the human predicament (and so the legislative one) is that we labour under two connected handicaps whenever we seek to regulate, unambiguously and in advance, some sphere of conduct by means of general standards to be used without further official direction on particular occasions."⁴¹ These handicaps are "our relative ignorance of fact" and "our relative indeterminacy of aim."⁴² Mechanical jurisprudence, involving simple application of law to fact, would be possible only if "the world in which we live were characterized only by a finite number of features, and these together with all the modes in which they combine could be known to us."⁴³ But "plainly this world is not our world."⁴⁴ The problem with a mechanical approach is that it secures

a measure of certainty or predictability at the cost of blindly prejudging what is to be done in a range of future cases, about whose composition we are ignorant. We shall thus indeed succeed in settling in advance, but also in the dark, issues which can only be reasonably settled when they arise and are identified.⁴⁵

The problem with that kind of decision is that it forces us "to include in the scope of a rule cases which we would wish to exclude in order to give effect to reasonable social aims. . . ."⁴⁶ In the context of environmental protection, the analysis seems straightforward. The literal language of congressional enactments will often include cases that Congress could not anticipate, and hence interpreters who follow that language will produce judgments "in the dark," and unreasonable, even absurd ones to boot. The cases discussed above seem to fall in this category.

Of course much of Hart's account is sensible and right. But Hart neglects some key issues, and he is able to do so only because of his apparently unself-conscious use of the word "we" to identify the interpreting authority. Why should we think that "we" are interpreting the statute? If "we" are engaging in interpretation, it might well make sense to insist that "we" should settle the relevant issues in the light rather in the dark. But if the question of how to interpret a statutory text is approached in institutional terms, it will become clear that Hart fails to attend to two crucial points. The first involves the risk of judicial blunder; the second involves the dynamic effects of one or another approach to interpretation.

In cases involving unanticipated problems, the question is not what "we" will do; it is what courts will do. Perhaps they will make nonsense rather than sense of the law. The possibility is far from remote in the context of regulatory law and environmental protection, where courts may lack the information that would enable them to answer the complex question about what should be done "in the light." And it is important to investigate the effects of the courts' choice on future litigants and on legislative behavior. A sensible system of interpretation understands the inevitability of dynamic effects; it sees that the judges' choice will not be limited, in its consequences, to the immediate parties. Suppose that courts, abandoning the text because it will produce nonsense, will introduce a high degree of uncertainty into the law. Suppose that if they proceed blindly and in the dark, they will create good incentives for the legislature, which will promptly correct the problems that arise. In these circumstances, perhaps a refusal to correct absurdities is not so bad after all. In the environmental context, these issues can be quite difficult. A decision to abandon the text, on the ground that it produces an absurd result in a particular case, might introduce a great deal of uncertainty into the future

Ironically, Ronald Dworkin, often taken to be Hart's adversary, shares the same blindness. Dworkin also rejects literalism, believing that judges should make the best constructive interpretation of the existing legal materials.⁴⁷ At first glance, the belief seems unexceptionable; but like Hart, Dworkin fails to address the imperfections of judges and the systemic effects of one or another approach to interpretation. It is not false to say, with Dworkin, that the controversy in Riggs "was a dispute about what the law was, about what the real statute the legislators enacted really said."⁴⁸ But it is hopelessly inadequate to cast the issue in those terms. The dispute was about the appropriate approach to a statute whose literal text produces absurdity. If courts ask about the meaning of "the real statute the legislators enacted," they will be neglecting some of the crucial issues.

What are those issues? Three are central: (a) the capacity of courts to make sound decisions if they make such exceptions; (b) the problems, in terms of increased uncertainty, that would be produced by the granting of exceptions; and (c) the likelihood that a legislature, faced with a rigid, rule-following judiciary, would correct the situation.⁴⁹ It is not at all clear, in the abstract, how the empirical issues should be resolved. In fact, they cannot be resolved in the abstract. Everything depends on context and on institutional capacities. In a legal system in which legislatures are able to anticipate many potential absurdities, the need for exemptions will decrease; so too in a system in which legislatures will promptly correct absurdities as they arise. In the environmental context, courts should probably take a literalist [32 ELR 11131] approach if Congress will correct absurdities and if courts lack the tools to identify situations in which literalism does not in fact make for absurdity. But in a legal system in which absurdities are common and unlikely to be anticipated or corrected, and in which judges are able to exempt absurdities without creating systemic harm, the case for literalism is severely weakened. In this light, those who debate the role of "plain meaning" in interpretation might well be taken to be disagreeing over some highly empirical questions about the real-world consequences of one or another approach. Those questions might not be easy to resolve; but if we focus on them, we will know what we are disagreeing about, and some quasi-theological disputes about how to interpret statutes might appear to be relatively mundane.

It is tempting to think this argument neglects the whole issue of legitimacy. Does it? I do not believe so. Almost everyone agrees that judges should not ignore statutory terms, even if the relevant judges are superb and even if the relevant legislatures are hopelessly inattentive and confused.⁵⁰ But we are speaking here not of ignoring terms, but of preventing unintended absurdity and, much of the time, of handling the particular interpretive problem of excessive generality. This is Hart's focus as well. The problem is pervasive in ordinary life, where language sometimes sweeps more broadly than any reasonable person would like. Legislatures, even the most attentive, are unable to avoid the problem. In such circumstances, judges who make sense rather than nonsense of the law are interpreting it, not ignoring it. If judicial interpreters, like other sensible people, disregard literal language to avoid a result that (in all likelihood) was not intended or sought, it is ludicrous to object that courts have committed some abstract sin called "illegitimacy." But it is not ludicrous to object; perhaps courts will make the situation worse rather than better, and perhaps they do better to interpret the statute literally. Whether the objection is reasonable depends on the pragmatic and institutional factors that I have identified here.

The Case of Agencies

Now let us turn to administrative agencies, giving particular attention to EPA and other agencies involved in protecting safety, health, and the environment. Recall that for courts, the case for literalism, even in the case of absurdity, depends on a belief that courts will not always make sound decisions about what counts as absurd; a judgment that literalism will create good incentives for legislatures, to anticipate problems before the fact and to correct them after the fact; and a fear that efforts to prevent absurdity will create systemic harm, by making planning more difficult and by increasing the level of uncertainty. We have seen that in the abstract, reasonable people might well disagree about how to do this calculus in the context of judges. But the calculus changes, in major ways, when agencies are involved.

It changes for two fundamental reasons. First, agencies are in a good position to know whether a particular outcome is, in fact, absurd. In the environmental setting, a court may be unlikely to have a clear sense of whether one or another approach would produce a patently unreasonable outcome; but EPA has far more information. To say the least, courts would have a hard time in deciding whether the language of the SDWA, taken literally, would compromise drinking water safety in the context of regulatory controls on lead. Confronted by an argument to that effect from regulated companies, courts might well hesitate to resolve the underlying issues; they might sensibly choose to follow the text. By contrast, agencies face no such disabilities. They are likely to know whether a literal interpretation really would cause environmental harm.

Second, agencies are in a far better position to know whether systemic harm will be created by departing from the text and trying to produce sense rather than nonsense in particular cases. A serious problem with abandoning literalism is that the abandonment might undermine planning and make it hard for the agency to enforce the law, in a simple and predictable way, in the future. But it is difficult for courts, and other generalists, to know whether this risk is real or fanciful. Because of their specialized competence, agencies are in a far better position to resolve that question. Compared to courts, they are more likely to know (for example) whether a decision to allow de minimis exceptions to a regulatory program would undo the regulatory scheme, or whether a decision to proceed as EPA did under the SDWA would make for unacceptable confusion in the future.

But this is not the only point. In resolving these questions, judgments of value might also be involved, and they can be extremely important. To the extent that value judgments are being made, agencies have advantages over courts too, precisely because of their political accountability. It is not inappropriate to conclude that EPA, subject as it is to presidential oversight, should be authorized to decide whether to make such exemptions, to the extent that the statute might fairly be described as ambiguous. (Recall that excessive generality is a form of ambiguity, and that where a statute produces absurdity, it is reasonable to say, as did the American Water Works court, that it lacks a plain meaning.) This point means that the Bush Administration has the authority to resolve certain issues, under Chevron step one, differently from the Clinton Administration. So be it. In the ordinary run of cases, it is a virtue, not a vice, of Chevron that the decision allows different administrations to resolve statutory ambiguities in different ways. Chevron itself, allowing the Reagan Administration to reject the view of the Carter Administration, is the most salient case in point. What I am adding here is a suggestion that under step one, the relevant administration should also be authorized, most of the time, to accept or to reject literalism when that approach would lead to an absurd or patently unreasonable outcome. As the recent cases suggest, agencies should have significantly more discretion, on this score, than federal courts.

In my view, the problem in Holy Trinity should be seen very differently in the context of the modern regulation, which will pose as a central question: What are the views of any agency charged with implementation of this law? As against occasional pleas for judicial "updating" of obsolete statutes, offered most influentially by Judge Guido Calabresi,⁵¹ it might be claimed that the argument needs to be updated: for the most part, solutions to the problem of legal obsolescence should come from administrative agencies, immersed in the problems at hand and having both technocratic and democratic virtues as compared to courts. In modern Holy Trinity-type cases, courts should not do the [32 ELR 11132] work on their own, but should permit agencies to engage in a degree of statutory adaptation. This is not at all on the theory that agencies are authorized to violate the statute (no agency is authorized to do that), but after a conclusion that agencies reasonably concluded that despite the generality of the text, it is legitimate to interpret it to avoid absurdity, at least if there is no considered legislative judgment that the text should be applied in a way contrary to the agency's view.

To say this is not to deny the possibility of counterarguments. In defending the new canon of construction, I have not discussed the possibility of congressional correction. If the legislature is highly responsive, then it will, by hypothesis, be less important to allow agencies to have the authority to avoid absurdity. And if agencies themselves should be distrusted, the new principle will seem much less appealing. If agencies deserve distrust, we might suspect that they are not applying their expertise, and we might be skeptical about their judgments of value. Suppose, for example, that agencies will seek to exempt risks not because those risks are genuinely trivial, but because of political pressures, imposed by regulated industries with self-serving agendas. If so, agency authority to exempt de minimis risks, as a way of avoiding apparent absurdity, seems far less appealing. Or suppose that in American Water Works, EPA was not really adopting the most sensible approach, but attempting to ease regulation in a way that allowed unacceptable health risks from lead. More generally, endorsement of the new canon that I have identified here depends on the judgment that when agencies claim that a certain outcome would be absurd, we have good reason to believe that they are likely to be right.

There is no algorithm to prove that the new canon is justified; everything depends on context and on the reliability of our various institutions. In a world in which legislatures correct absurdities and agencies are confused or untrustworthy, formalism—an insistence on adherence to the literal language—would make as much sense for agencies as for courts. In such a world, American Water Works, and the other cases I have discussed, would be wrong. But this is not our world. Agencies must apply statutory language in countless settings, many of them new and unanticipated and often in the face of changing circumstances. In the environmental context, agencies must work with statutes that are frequently quite rigid, that must of course be obeyed, but that in some applications produce outcomes that Congress could not possibly have wanted. It is far too much to expect Congress to be closely attentive whenever excessive generality produces absurdity. Indeed a well-functioning system of interpretation would make it unnecessary for Congress to spend its time on such comparative trivia; it would allow corrections through the interpretive methods discussed here.

Of course regulatory agencies often blunder. But in the modern state, statutory terms must be engrafted onto count-less situations and problems that even the most reasonable and far-sighted legislatures could not easily have foreseen. The "thin simplifications"⁵² of enacted law will inevitably fit poorly with at least some of the contexts to which they apply. Indeed this is a pervasive problem—and an inevitable product of the movement from a common-law system to one pervaded by statutory commands. In areas ranging from safety in drinking water to telecommunications and banking, statutory law cannot possibly fit well with the range of situations to which, over decades or more, to which it must be adapted. This is an extremely serious problem for which it is important to find correctives.⁵³ The new canon that I am discussing is a modest means of reducing that problem.

Will agencies abuse their new authority? Can EPA be trusted to use the new canon so as to avoid real absurdity, rather than bowing to powerful private interests? I think that the underlying risks are minimal. In the cases I am discussing here, there is no need for alarm; the usual pattern is one in which sensible regulators are attempting to counteract excessive generality or palpable unreasonableness, neither intended by Congress. In allowing agencies to make corrections, courts are merely permitting agencies to do what courts themselves have long done. Once we debate interpretive questions in institutional terms—in terms of who is good at what—we might reasonably dispute whether courts should be authorized to interpret statutory language so as to avoid absurdity. But even if courts should be denied that authorization, we should welcome agency power to make sense rather than nonsense of regulatory law. The new canon of interpretation thus fits well with the displacement of the common law by regulatory programs—and it is likely to find an increasingly receptive judiciary in the coming decades.

1. Church of the Holy Trinity v. United States, 143 U.S. 457 (1892).

2. See Frederick Schauer, The Practice and Problems of Plain Meaning: A Response to Aleinikoff and Shaw, 45 VAND. L. REV. 715 (1992).

3. See id. at 734-36; Edward Rubin, Modern Statutes, Loose Canons, and the Limits of Practical Reason, 45 VAND. L. REV. 579 (1992). Some of these issues, and the jurisprudential questions, are discussed in some detail in CASS R. SUNSTEIN & ADRIAN VERMEULE, INTERPRETATION AND INSTITUTIONS (Univ. of Chicago Working Paper, July 2002); related issues of regulatory law are discussed, from a different angle, in CASS R. SUNSTEIN, RISK AND REASON: LAW, SAFETY, AND THE ENVIRONMENT (2002).

4. See Adrian Vermeule, Interpretive Choice, 75 N.Y.U. L. REV. 74, 103 (2000).

5. See ANTONIN SCALIA, A MATTER OF INTERPRETATION (1990).

6. See RONALD DWORKIN, LAW'S EMPIRE 242-44 (1985).

7. See Vermeule, supra note 4; Cass R. Sunstein, Must Formalism Be Defended Empirically?, 66 U. CHI. L. REV. 636 (1999).

8. See Schauer, supra note 2; Vermeule, supra note 4.

9. See JAMES SCOTT, SEEING LIKE A STATE (2000).

10. 467 U.S. 837, 14 ELR 20507 (1984).

11. Id. at 842, 14 ELR at 20508.

12. Id.

13. 40 F.3d 1266, 25 ELR 20335 (D.C. Cir. 1994).

14. 42 U.S.C. §§ 300f to 300j-26, ELR STAT. SDWA §§ 1401-1465.

15. Id. § 300g-1(b)(4)(A), ELR STAT. SDWA § 1412(b)(4)(A).

16. Id. § 300g-1(b)(4)(B), ELR STAT. SDWA § 1412(b)(4)(B).

17. Id. § 300g-1(b)(6)(D), ELR STAT. SDWA § 1412(b)(6)(D).

18. 40 F.3d at 1271, 25 ELR at 20336.

19. Id.

20. Id.

21. Id.

22. Id., 25 ELR at 20337.

23. 613 F.2d 947 (D.C. Cir. 1979).

24. Id. at 955.

25. Alabama Power Co. v. Costle, 636 F.2d 323, 10 ELR 20001 (D.C. Cir. 1979).

26. Id. at 360, 10 ELR at 20011.

27. Id. at 360 n.89, 10 ELR at 20011 n.89.

28. Id. at 360, 10 ELR at 20011.

29. 213 F.3d 663, 30 ELR 20407 (D.C. Cir. 2000).

30. 42 U.S.C. § 7410(a)(2)(D)(i)(I), ELR STAT. CAA § 110(a)(2)(D)(i)(I) (emphasis added).

31. 213 F.3d at 675, 30 ELR at 20408.

32. Id. at 678, 30 ELR at 20412.

33. Id.

34. Id.

35. Id.

36. Many cases fall in the same category as those discussed here. See, e.g., Natural Resources Defense Council v. EPA, 824 F.2d 1146, 17 ELR 21032 (D.C. Cir 1987). But there are counterexamples. Consider, for example, American Mining Congress v. EPA, 824 F.2d 1177, 17 ELR 21064 (D.C. Cir. 1987). Congress had not clearly dealt with the problem of how to handle materials held for recycling, and the relevant EPA regulation defined certain materials involved in recycling as "solid waste." In particular, it said that spent materials, sludges, scrap metal, and the like would be treated as solid waste if they were not directly reused but were instead held as part of an industry's ongoing production process. EPA reasoned that materials that were stored, transported, and held for recycling were associated with the same kinds of environmental harms as materials that were abandoned or disposed of in some final way. The court of appeals struck down EPA regulation on the ground that the governing statute defined solid waste as "garbage, refuse, sludge . . . and other discarded material"; for the court, material held for recycling was not "discarded." Citing the dictionary, the court thought that the "ordinary plain-English meaning" was decisive. Cf. Public Citizen v. Young, 831 F.2d 1108, 18 ELR 20173 (D.C. Cir. 1987) (the court found it necessary or at least relevant to emphasize that the rigid text was not, in fact, absurd).

37. 115 N.Y. 506, 22 N.E. 188 (1889).

38. 143 U.S. 457 (1892).

39. See SCALIA, supra note 5 (criticizing Holy Trinity).

40. See H.L.A. HART, THE CONCEPT OF LAW 125 (1972).

41. Id.

42. Id.

43. Id. at 126.

44. Id.

45. Id.

46. Id.

47. See his discussion of Tennessee Valley Auth. v. Hill, 437 U.S. 153, 8 ELR 20513 (1978), in DWORKIN, supra note 6, at 345-48.

48. Id. at 20.

49. See SUNSTEIN & VERMEULE, supra note 3, for a detailed discussion.

50. Actually, I believe the underlying issues are more complicated than the agreement suggests, but I cannot address the complications here.

51. See GUIDO CALABRESI, THE COMMON LAW IN AN AGE OF STATUTES (1989).

52. See SCOTT, supra note 9.

53. The movement toward economic incentives, as opposed to command-and-control regulation, can be understood in these terms. See A. DENNY ELLERMAN ET AL., MARKETS FOR CLEAN AIR (2000).