Massachusetts v. EPA asked the oldest question in statutory law in its newest form. When Congress writes broad words to solve the problems it can see, who decides what those words require for a problem Congress never named? The Clean Air Act directs the Environmental Protection Agency to prescribe standards for the emission of “any air pollutant” from new motor vehicles, and it defines “air pollutant” in section 302(g) with a sweep that reaches any physical, chemical, or biological substance emitted into or otherwise entering the ambient air. When those words were written, the legislators who wrote them were thinking about smog, soot, and lead, not about carbon dioxide accumulating in the atmosphere over decades. In 1999 a coalition of environmental organizations asked the agency to read the old words as covering greenhouse gases, and the agency’s answer, first a long silence and then a formal denial in 2003, set up the litigation that reached the Supreme Court. The decision the Court handed down in 2007 became the most cited environmental ruling of its generation, not because the justices settled the science of climate change, which the administrative record already established, but because they settled the method: how a court reads a capacious statute when the agency charged with enforcing it would rather not.

That method is the reason this account begins three decades before the case itself. The Clean Air Act is the rare statute that generated, in opposite directions, the two most consequential administrative law doctrines of the modern era. In one direction the Act produced Chevron deference: the 1984 holding, born in a dispute about whether an entire factory counts as one “stationary source” for emissions purposes, that courts should accept any reasonable agency reading of an ambiguous statute. That framework governed administrative law for four decades, until the Court discarded it in 2024. In the opposite direction the Act produced the doctrines that restrain agencies: the 1976 holding that infeasibility cannot defeat a state plan that meets the statutory criteria, the unanimous 2001 holding that the statute forbids weighing costs when setting health based air quality standards, and the major questions framework whose most prominent application arrived in a 2022 decision about the Act’s power plant provisions. Massachusetts v. EPA stands at the hinge between those poles. The same capacious language that licensed deference also, read with care, left no ambiguity for deference to fill. The 2007 decision makes sense only against the three earlier landmarks that fixed the statute’s character, and those landmarks are where the explanation properly starts.

The stakes of the reading question explain why the agency fought it so hard, and they had little to do with the science. If greenhouse gases were “air pollutants” under the motor vehicle title, the same definition would apply everywhere else the Act uses the term, including the preconstruction permit programs whose thresholds sweep in any source emitting more than one hundred or two hundred fifty tons per year of a regulated pollutant. Applied to carbon dioxide, those thresholds would have pulled vast numbers of small sources, from apartment buildings to hospitals, into a permitting system built for large factories, a cascade the agency warned would produce absurd results. The agency’s answer to that cascade, years later, was the 2010 Tailoring Rule, which raised the statutory thresholds for greenhouse gases on the ground that Congress could not have intended the literal numbers to apply. That later episode belongs to the aftermath, but it illuminates the 2003 denial: the agency feared that conceding authority over motor vehicle greenhouse gases would commit it to regulating the gas everywhere, all at once. The litigation also broke new ground on who gets to sue. Massachusetts did not come to court as an ordinary injured party; it came as a sovereign state, asserting injury to its coastline and its quasi sovereign interest in the welfare of its citizens, and the majority held that states enjoy special solicitude in the standing analysis. The dissenters protested that this relaxed the constitutional limits on federal judicial power. Both the merits and the standing holding would echo through later climate litigation, but the merits question is the one this account follows, and the merits question runs through the statute’s earlier history.

Massachusetts v. EPA and the Statute the Court Was Asked to Read

The case is Massachusetts v. Environmental Protection Agency, 549 U.S. 497 (2007). It construes two provisions of the Clean Air Act: section 202(a)(1), which provides that the Administrator “shall by regulation prescribe … standards applicable to the emission of any air pollutant from any class or classes of new motor vehicles or new motor vehicle engines,” and section 302(g), which defines “air pollutant” to include “any air pollution agent or combination of such agents, including any physical, chemical, biological, radioactive … substance or matter which is emitted into or otherwise enters the ambient air.” Readers who want the statute’s full architecture, its titles, its programs, and the way its parts fit together, should start with the complete guide to the Clean Air Act before pressing further into the case law.

The Court was asked to decide two questions, and only two. First, whether section 202(a)(1) authorizes the Environmental Protection Agency to regulate greenhouse gas emissions from new motor vehicles, which turns entirely on whether greenhouse gases fall within the section 302(g) definition of “air pollutant.” Second, if the authority exists, whether the agency’s stated reasons for declining to exercise it were so inadequate as to be arbitrary. The litigation posture matters because it shaped everything that followed. A rulemaking petition filed in 1999 asked the agency to set greenhouse gas standards for new vehicles. The agency denied the petition in 2003, offering three reasons: that it lacked statutory authority over greenhouse gases, that it would decline to regulate even if the authority existed, and that voluntary programs and further study were the wiser course. A divided panel of the United States Court of Appeals for the District of Columbia Circuit denied review of that denial in a fractured judgment, and the Supreme Court granted certiorari. Argument took place in November 2006, and the decision issued on April 2, 2007, by a five to four vote. Justice Stevens wrote the majority opinion, joined by Justices Kennedy, Souter, Ginsburg, and Breyer. Chief Justice Roberts wrote the principal dissent, joined by Justices Scalia, Thomas, and Alito, and Justice Scalia wrote a separate dissent as well.

The agency’s 2003 denial is worth reading closely because the Court’s opinion is organized as a point by point rejection of it. The denial offered three reasons. First, the agency claimed it lacked statutory authority over greenhouse gases, reading “air pollutant” as limited to substances that foul the air near the ground, the traditional targets of the Act. Second, the agency said that even if the authority existed, it would decline to exercise it, citing the preference for voluntary programs, the desirability of further scientific study, and the foreign policy dimension of climate negotiations. Third, the agency argued that regulating motor vehicle greenhouse gases would be a piecemeal and ineffective response to a global problem. The majority’s answer to the first reason was textual: the definition’s breadth admits no such limitation. Its answer to the second and third was structural: section 202(a)(1) says the Administrator shall regulate pollutants that in her judgment endanger public health or welfare, and the statute specifies the considerations that may inform that judgment. Policy preferences about voluntary programs and diplomacy are not among them. An agency may decline to regulate for reasons grounded in the statute, such as genuine scientific uncertainty about endangerment, but not for reasons the statute does not recognize. That holding, that an agency’s discretion is bounded by the reasons Congress made relevant, is the decision’s most portable administrative law contribution, and it survived the 2024 overruling of Chevron untouched, because it rests on the statute’s own terms rather than on any deference doctrine.

One clear pass on the question presented will keep the rest of the analysis honest. The case did not ask the Court to determine whether climate change is occurring, whether human activity causes it, or whether regulating motor vehicle emissions would solve it. The agency itself had compiled a record acknowledging the scientific consensus, and the majority took that record as given. The case asked a narrower and in some ways harder question: whether words Congress wrote to control the air pollution of the 1970s, words as broad as “any physical, chemical, biological … substance or matter which is emitted into or otherwise enters the ambient air,” cover gases that trap heat in the atmosphere, and what an agency owes the public when it refuses to use authority the statute appears to grant. The majority answered that the definition is unambiguous, that greenhouse gases fit comfortably within it, and that the agency’s refusal rested on policy preferences the statute does not permit it to consider. The dissent answered that “air pollutant” as used in the motor vehicle title was aimed at substances that pollute the air people breathe, that the agency’s reading deserved deference, and that the majority was forcing regulation through interpretation. Both sides agreed on the stakes: the decision would determine whether the Clean Air Act, a statute written before climate change entered public consciousness, could be the vehicle for the nation’s first federal greenhouse gas regulation.

The textual raw material deserves a close look, because the entire case turns on a few dozen words. Section 302(g) defines “air pollutant” as “any air pollution agent or combination of such agents, including any physical, chemical, biological, radioactive (including source material, special nuclear material, and byproduct material) substance or matter which is emitted into or otherwise enters the ambient air.” Three features of that definition did the work in 2007. First, the repeated “any”: any agent, any combination, any substance or matter. Congress used the broadest available quantifier four times in a single sentence, a pattern the majority read as deliberate capaciousness rather than careless drafting. Second, the definition is organized around the medium rather than the harm. A substance qualifies by being emitted into the ambient air, which is simply the outdoor air surrounding the earth, not by being shown to dirty that air in a particular way. Third, the definition contains no qualifier limiting it to substances Congress had in mind in 1970, when the definition was written, or in 1977 and 1990, when Congress revisited the Act without narrowing it. The agency had long treated the definition as covering whatever the regulatory titles needed it to cover, from the six criteria pollutants subject to national standards to the hazardous pollutants controlled under section 112.

Section 202(a)(1) then attaches a specific regulatory consequence to that broad definition. The Administrator must prescribe standards for emissions of “any air pollutant” from new motor vehicles that “in his judgment cause, or contribute to, air pollution which may reasonably be anticipated to endanger public health or welfare.” The endangerment predicate matters because of how the Act defines “welfare.” Section 302(h) provides that welfare includes effects on “climate” alongside effects on soils, water, crops, vegetation, animals, and visibility. Congress thus wrote the word “climate” into the statute’s definitional section, in the very provision that tells the reader what “welfare” means. The majority in 2007 leaned on that cross reference: a statute whose definition of welfare expressly includes climate can hardly be read to exclude climate related harms from the endangerment inquiry. The motor vehicle title was also the natural entry point for greenhouse gas regulation as a practical matter. Title II had been the engine of the statute’s most visible successes, the tailpipe standards that cut smog forming emissions from new cars by more than ninety percent between the 1970s and the 2000s, and motor vehicles accounted for a large and growing share of the nation’s carbon dioxide emissions. If the definition covered greenhouse gases anywhere, it covered them here. The 1990 amendments, the last major revision of the Act, are telling for what they did not do. Congress rewrote the hazardous pollutant program, created the acid rain trading program, and overhauled the nonattainment provisions, yet it left the section 302(g) definition of “air pollutant” exactly as it had been since 1970. In statutory interpretation, sustained congressional silence of that kind carries weight: Congress legislated repeatedly around the definition without narrowing it, which suggests the breadth was understood and accepted rather than overlooked. The dissenters read the same history differently, as mere inattention to a definition nobody expected to do climate work, but the majority treated the untouched text as evidence that the words meant what they said.

The reason the disagreement ran so deep is that the statute had already taught the Court, three times over, how it wants to be read. Each of those lessons came from a different decade, a different program of the Act, and a different kind of interpretive problem, and together they form the road into the 2007 decision.

The Road to the Decision, 1976 to 2001

Union Electric Co. v. EPA, 427 U.S. 246 (1976)

The first landmark is about who bears the pain when the statute’s demands exceed what is convenient. The Clean Air Act sets national ambient air quality standards for listed pollutants, and then, under section 110, requires each state to adopt an implementation plan that will attain those standards within the statutory deadlines. The federal agency reviews each state plan, and the statute lists the criteria the plan must satisfy: enforceable emission limitations, monitoring, enforcement machinery, and the rest. The question in Union Electric was what happens when a state plan meets the listed criteria but the sources inside the state insist that compliance is economically or technologically infeasible.

Union Electric Company, a Missouri utility, challenged the agency’s approval of Missouri’s plan for controlling sulfur dioxide. The company argued, in substance, that the plan’s requirements could not feasibly be met and that the agency should have weighed that infeasibility before approving the plan. Justice Marshall wrote the opinion for a unanimous Court, Justice Powell concurring, and the Court rejected the argument. The holding is crisp: under section 110, the agency must approve a state plan that satisfies the statutory criteria, and economic or technological infeasibility is not among those criteria. A state may not resist a federally required plan on the ground that compliance will be difficult or expensive, and a source may not block approval of the plan by proving that the controls do not yet exist or cannot be afforded.

The reasoning runs deeper than the holding. The Court examined the structure Congress built and found a deliberate allocation of burdens. The statute gives states wide latitude to choose the mix of controls, the timetables within the federal deadlines, and the distribution of reductions among sources. What the statute does not give states is latitude about the destination: the plans must provide for attainment of the national standards, period. Congress, the majority explained, was aware that the requirements would be demanding, and it chose to place the burden of figuring out how to comply on the states and the sources rather than on the public that breathes the air. Infeasibility arguments belong, if anywhere, to later stages of the process, such as enforcement proceedings or deadline extensions that Congress separately provided, not to the threshold question of whether a plan that meets the statutory checklist gets approved.

Why the holding mattered for the statute’s design is a point worth dwelling on, because it set the template for everything that followed. Union Electric made the Clean Air Act a technology forcing statute. By refusing to let feasibility dilute the plan approval stage, the Court ensured that the standards would drive innovation rather than wait for it. A source that cannot comply with the technology at hand must develop better technology, or shut down, or pay someone else to reduce; what it cannot do is ask the agency to lower the bar at the front end. That design choice explains the statute’s ferocity. It also explains the statute’s political durability: Congress could take credit for clean air goals while the hard choices about who pays fell to the states, the agency, and the courts. The decision confirmed that the Act means what its most demanding words say, and it announced, three decades before Massachusetts v. EPA, that the Court would not soften the statute to spare regulated parties.

What did Union Electric establish about how states must meet federal air standards?

Missouri’s plan had to assure attainment of federal sulfur dioxide standards, and the utility argued the required controls were infeasible. The Court held that infeasibility is no defense: the Act demands that state plans meet the standards, and the state must find a way. The holding made the standards technology forcing, leaving feasibility arguments for the later implementation stage.

The depth behind that answer repays attention because the case is so often cited and so rarely read. The utility’s real complaint was not abstract. Missouri’s plan required substantial sulfur dioxide reductions from coal fired power plants, and in the mid 1970s the available control technology, principally flue gas desulfurization, was expensive, unreliable at scale, and in some configurations unproven for the high sulfur coal the region burned. The company argued that approving a plan built on such controls was arbitrary, and it asked the courts to make the agency consider whether the plan could actually be carried out. The Court’s answer was that Congress had already considered the question and answered it. The statutory criteria for plan approval are stated as requirements, not as factors to be balanced, and the Court refused to insert an additional feasibility inquiry that Congress had omitted. The opinion noted that Congress provided other safety valves, including the possibility of delayed compliance orders and enforcement discretion, but those valves operate after approval, not as grounds for disapproval.

The practical consequence was a shift in where the statute’s battles get fought. After Union Electric, a challenger cannot defeat a state plan at the approval stage by proving hardship. The hardship arguments migrate to enforcement, to variance proceedings, and to Congress itself, which responded to the pressure with the 1977 amendments’ extended deadlines and the elaborate nonattainment machinery. The decision thus did two things at once: it hardened the statute’s front end and it pushed the flexibility to the back end, where political accountability is clearer. For the purposes of the road into Massachusetts v. EPA, the case established the first principle of reading this statute: when Congress wrote demanding words, the Court would not read in escape hatches.

The machinery the Court was interpreting deserves a fuller description, because it remains the backbone of American air pollution control. A state implementation plan is not a vague promise. Section 110(a)(2) lists its required contents: enforceable emission limitations and other control measures, a program for monitoring and analyzing ambient air quality, an enforcement program with adequate personnel and funding, provisions for public participation, and measures to prevent significant deterioration in clean areas, among others. The state holds hearings, adopts the plan under state law, and submits it to the federal agency, which reviews it against the statutory checklist. If the plan satisfies every listed criterion, the agency must approve it; the statute uses mandatory language, and Union Electric confirmed that the list is exhaustive. If the state fails to submit a plan, or submits one the agency disapproves, the backstop in section 110(c) requires the agency to promulgate a federal implementation plan within two years, imposing the controls directly. The design is cooperative federalism with a federal hammer behind it: the state gets the first move and wide discretion over means, but the attainment obligation is non negotiable and the federal government will supply the plan if the state will not.

The aftermath of Union Electric shows how seriously the political branches took the holding. States found the original 1975 attainment deadlines unachievable for many pollutants, and rather than asking courts to soften the statute, they went to Congress. The 1977 amendments extended the deadlines to 1982, with a further extension to 1987 for the most stubborn pollutants, and created the detailed nonattainment program that Chevron would later construe. Congress thus accepted the Court’s reading and bought time instead of weakening the obligation. The episode is the clearest demonstration of the Union Electric principle in action: when the statute’s demands prove too severe, the remedy is legislation, not creative interpretation. That principle would matter in 2007, when the agency argued in effect that regulating greenhouse gases would be too disruptive to attempt. The Court’s answer, foreshadowed in 1976, was that disruption is a question for Congress, while the agency’s job is to read the words it was given.

The decision also shaped the enforcement stage that follows plan approval. Because infeasibility cannot block a plan, sources facing difficult deadlines must seek relief through the channels Congress provided: delayed compliance orders under section 113, which allow extra time on a showing of good faith effort; revisions to the state plan itself, subject to the same approval process; or, ultimately, legislation. Each of those channels keeps the attainment obligation visible and forces the hardship claim into the open, where the agency or Congress can weigh it against the health stakes. What the Court closed off was the quiet route: persuading a reviewing court that the plan was unreasonable because compliance was hard. That closure is why Union Electric is still cited whenever a regulated party asks a court to soften a statutory command on practical grounds.

The holding also fixed the evidentiary posture of later plan litigation. Because the approval question turned on the statutory checklist rather than on a free floating inquiry into reasonableness, challengers could no longer dress feasibility evidence in the language of arbitrariness and expect a court to weigh it. The record that mattered was the state’s demonstration that each listed criterion was met, and a court reviewing the approval asked only whether that demonstration was supported, not whether compliance would be painful. That discipline kept the plan approval docket narrow through the decades that followed, even as the plans themselves grew more elaborate under the 1977 and 1990 amendments, and it is the reason the section 110 machinery could absorb pollutants and programs the 1970 Congress never imagined.

Chevron U.S.A., Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837 (1984)

The second landmark is the most famous administrative law decision of the twentieth century, and it is essential to get its facts right because mythology has blurred them. Chevron was not a climate case, not a case about greenhouse gases, and not a case about the scope of the agency’s power in any grand sense. It was a case about the bubble concept under the Act’s provisions for areas failing to attain the national standards.

The underlying dispute was technical. In the portions of the country designated nonattainment, the Act imposes demanding preconstruction review on new or modified “major stationary sources.” In 1981 the agency revised its regulations to define “stationary source” on a plantwide basis: all of the pollutant emitting activities within the same industrial grouping, on contiguous or adjacent properties under common control, would count as a single source. Under this bubble approach, a plant could modify one unit and increase its emissions so long as it reduced emissions elsewhere within the same bubble, with no net increase for the plant as a whole, and thereby avoid the costly new source review. The Natural Resources Defense Council challenged the rule, arguing that the statute required a narrower, unit by unit definition. The D.C. Circuit agreed with the challengers and struck the rule down. The Supreme Court reversed.

The context of the 1981 rule explains why the dispute reached the Court at all. The bubble had roots in the agency’s offset policies of the late 1970s, but the 1981 rulemaking, issued in the first year of the Reagan administration’s regulatory reform drive, generalized the plantwide definition across the nonattainment permit program created by the 1977 Amendments. Industry welcomed the change as flexibility: a plant modernizing one unit could avoid the expense and delay of new source review by cutting emissions at another unit, with the atmosphere seeing no net increase. Environmental organizations saw a loophole: netting within the bubble allowed real increases at individual smokestacks, and the reductions offered in exchange were often cuts the plant would have made anyway, or cuts in pollutants measured with convenient imprecision. The D.C. Circuit sided with the challengers on purposive grounds. The nonattainment program, the court reasoned, was designed to force steady improvement in the country’s dirtiest air, and a definition that let plants reshuffle emissions without adding controls frustrated that design. The opinion treated “stationary source” as most naturally meaning an individual emitting unit and held that the agency’s plantwide reading contradicted the statutory purpose.

The Supreme Court’s reversal rejected that method of analysis root and branch. Where the D.C. Circuit had asked what reading best served the statute’s goals, the Supreme Court asked only whether Congress had answered the precise definitional question and, finding no answer, whether the agency’s answer was reasonable. The opinion noted that the statute left the precise meaning of “stationary source” open within the nonattainment permit program the 1981 rule implemented, which suggested Congress had left the details of implementation to the agency. The bubble survived not because the Court thought it the best policy, the opinion is notably agnostic on the policy merits, but because the choice belonged to the politically accountable branch within the space Congress left open. That agnosticism is the framework’s signature: the court does not pick the winner of the policy debate, it certifies that the agency was entitled to pick one.

Justice Stevens wrote the opinion for a six justice Court, with three justices not participating, and the decision was unanimous among those who sat. The opinion announced the two step framework that would govern judicial review of agency statutory interpretation for the next four decades. The framework is best stated as a sequence of questions, narrated here in prose because the substance matters more than the outline. A court first asks whether Congress has directly answered the precise question at issue. If Congress has spoken clearly, the inquiry ends; the court and the agency alike must honor the unambiguous command. If instead the statute is silent or ambiguous on the precise point, the court moves to the second question: whether the agency’s construction is a permissible or reasonable reading of the statute. If it is, the court upholds it, even if the court would have read the statute differently on its own. Deference, in this design, is not a reward for agency wisdom. It is a consequence of ambiguity. Where Congress left a gap, the agency fills it; where Congress left no gap, there is nothing to defer to.

That last sentence is the hinge on which Massachusetts v. EPA would later turn, and it is worth pausing over. The 1984 opinion assumed that ambiguity is common in complex regulatory statutes and that Congress, by leaving interstitial questions unresolved, implicitly delegates them to the administering agency. The bubble dispute was the perfect vehicle for the framework precisely because it was so technical and so interstitial. Nothing in the Act’s text dictated whether “stationary source” meant a single smokestack or a whole plant, the legislative history was inconclusive, and the choice between the two readings was a policy judgment about how to balance economic growth against air quality in struggling regions. The Court treated that kind of choice as the agency’s to make, and the framework generalized the point across the whole of administrative law.

What turned a factory definition dispute into the deference era’s landmark?

EPA defined “stationary source” as an entire plant, letting operators offset new emissions against reductions elsewhere on site, the bubble. The Court announced a two step test: first ask whether Congress answered the precise question; if the statute is silent, accept any reasonable agency reading. That test governed administrative law from 1984 until the Court discarded it in 2024.

The depth behind that answer starts with why a modest dispute about factory definitions became the most cited administrative law opinion in American history. Part of the answer is timing. By the early 1980s the federal courts were drowning in challenges to agency rulemaking, the D.C. Circuit in particular was deeply divided over how aggressively judges should second guess agency readings of statutes, and the Supreme Court needed a unifying statement. Part of the answer is the opinion’s own ambition. Justice Stevens did not write a narrow decision about bubbles; he wrote a general theory of the judicial role in the administrative state, and lower courts seized on the two step sequence because it gave them a clean, repeatable method. Within a decade the framework was the starting point for nearly every case about agency interpretation, cited thousands of times, taught in every administrative law course, and treated by agencies as a shield for their most creative readings.

The framework also generated the deepest normative debate in modern public law, and the debate deserves to be stated at full strength on both sides because the 2024 overruling did not settle the underlying disagreement. The case for deference runs as follows. Regulatory statutes are technical, and agencies possess scientific and engineering expertise that generalist judges lack. Agencies answer, however indirectly, to an elected President, which gives their policy choices a democratic pedigree that judicial policy choices lack. Congress legislates against the background of agency administration and leaves gaps precisely because it wants the expert body to fill them over time as circumstances change. And a single national agency interpretation promotes uniformity, where fifty different judicial readings would produce chaos. The most sophisticated modern defenders of this view, Cass Sunstein and Adrian Vermeule, argued that deference reflects a sound allocation of institutional competence: courts are good at policing boundaries, agencies are good at making the technical and political tradeoffs inside those boundaries, and the two step test keeps each institution in its lane.

The case against deference is equally serious and draws on older constitutional soil. The Constitution vests the judicial power in the courts, and the classic statement of that power, from Marbury v. Madison, is that it is the province and duty of the judicial department to say what the law is. Deference, on this view, asks judges to abdicate that duty whenever a statute is unclear, which is exactly when judging is hardest and most necessary. It also permits agencies to be judges in their own cause, interpreting the limits of their own power, which the separation of powers forbids. And it raises concerns about fair notice and liberty: regulated parties should be able to know what the law requires from the statute and the courts, not from the shifting reasonable readings of the enforcer. The most forceful modern critic, Philip Hamburger, argued in his study of administrative law’s history that deference revives extralegal prerogatives the Constitution was written to prevent, while Justice Gorsuch, in a series of opinions culminating in his concurrence in the 2024 overruling decision, argued that Chevron traded the rule of law for bureaucratic convenience. The historian Thomas Merrill, in his account of the doctrine’s rise and fall, showed how a framework born as a modest tiebreaker hardened into a license for agencies to claim ever larger powers, which is why the debate outlived the doctrine.

In 2024 the Court overruled the Chevron framework in Loper Bright Enterprises v. Raimondo. The majority, written by Chief Justice Roberts over a dissent by Justice Kagan, held that the Administrative Procedure Act requires courts to exercise independent judgment in determining whether an agency has acted within its statutory authority, and that Chevron’s presumption of deference to reasonable agency readings cannot be reconciled with that command. The vote was six to two in Loper Bright itself, Justice Jackson having recused after participating in the case below, and six to three in the companion Relentless case, and the opinion expressly preserved the holdings of cases previously decided under Chevron. The underlying 1984 holding about the bubble, the plantwide definition of stationary source, was left intact; what fell was the general framework, not the result it once supported. Courts may still respect agency views for their persuasiveness, particularly on technical matters within agency expertise, but they may no longer defer as a matter of course.

The four decades between the framework’s announcement and its overruling were not static, and the qualifications the Court added along the way help explain why the doctrine fell when it did. In 2001, in United States v. Mead Corp., the Court added what scholars call step zero: deference applies only where Congress delegated authority to the agency to act with the force of law, and the agency exercised that authority in a qualifying format such as notice and comment rulemaking or formal adjudication. Informal agency pronouncements, opinion letters, and manuals get only respectful consideration, not Chevron deference. In 2013, in City of Arlington v. FCC, a divided Court held that the framework applies even to an agency’s interpretation of the scope of its own jurisdiction, over a dissent warning that this let agencies define their own powers. And across the same period, the major questions doctrine grew up alongside Chevron as a kind of threshold exception: for decisions of vast economic and political significance, courts demanded clear congressional authorization rather than deferring to a reasonable reading of ambiguous text. Each qualification chipped at the framework’s simplicity, and by the time Loper Bright arrived, the doctrine’s defenders were defending a structure full of carve outs.

The framework’s Clean Air Act origins shaped how the agency wrote rules for decades. Rulemaking preambles under the 1990 amendments routinely structured their statutory analysis in two steps, first arguing that Congress had left the precise question open, then defending the chosen reading as reasonable, a habit visible across the hazardous pollutant standards, the interstate transport rules, and the greenhouse gas program itself. That drafting discipline was rational under the regime: an agency that could place its reading inside the deference space survived judicial review even when judges disagreed with the policy. The habit outlived its usefulness only when the Court began deciding the ambiguity question against agencies, narrowing the space step one left open and, in 2024, eliminating the second step entirely.

The 2024 majority’s reasoning rested on two pillars. The first was statutory: section 706 of the Administrative Procedure Act directs reviewing courts to “decide all relevant questions of law,” language the majority read as a command for independent judicial judgment that Chevron had effectively rewritten. The second was constitutional: the judicial power includes saying what the law is, and a doctrine that systematically transfers that function to the executive in close cases cannot stand. On stare decisis, the majority concluded that Chevron was unworkable in practice, pointing to the proliferating exceptions and the difficulty lower courts had in applying the two steps consistently, and that no serious reliance interests supported keeping it, since regulated parties adapt to whatever interpretive regime courts use. Justice Kagan’s dissent answered each point: Congress legislates against a background understanding that agencies will fill statutory gaps, agencies possess comparative expertise on the technical questions embedded in regulatory statutes, Chevron promoted national uniformity and political accountability, and discarding a precedent cited thousands of times would flood the courts with challenges to long settled agency interpretations. The disagreement, in other words, reprised the original normative debate at a higher level of abstraction, and neither side persuaded the other.

For the road into Massachusetts v. EPA, the significance of Chevron lies in its first step. The 2007 majority did not defer to the agency’s reading of “air pollutant” because it found nothing to defer to: the statutory definition was, in the majority’s view, unambiguous, and unambiguous text ends the inquiry. The dissenters, notably, invoked deference principles to argue that the agency’s narrower reading deserved respect. The case is thus a study in how the same framework can cut in opposite directions depending on where a court finds the ambiguity line, and it previews the post 2024 world in which courts draw that line without any thumb on the scale.

The step one analysis in 2007 repays a closer look because it shows how much work the ambiguity finding does. The majority did not simply announce that the text was clear; it walked through the definition’s components, the repeated “any,” the absence of limiting language, the welfare cross reference to climate, and the consistency of a broad reading with the Act’s structure, and concluded that no reasonable reader could confine “air pollutant” to the substances the agency preferred. The dissenters performed the same exercise and reached the opposite conclusion, emphasizing that “air pollutant” in ordinary usage connotes dirty air rather than a colorless, odorless gas that is harmless to breathe at ambient concentrations. Both sides were doing step one analysis: fighting over whether Congress had answered the question. That is the point worth carrying forward. Deference doctrines, whether Chevron’s or its successors, matter only in the space that textual analysis leaves open. Massachusetts v. EPA is remembered as an environmental decision, but as an administrative law decision it is a demonstration that the size of the deference space is itself a question of statutory interpretation, and that a court convinced the text is clear owes the agency nothing.

Whitman v. American Trucking Associations, Inc., 531 U.S. 457 (2001)

The third landmark is, by the judgment of most specialists, the single most consequential interpretive holding about the Clean Air Act ever issued, and it arrived unanimously. The case concerned the national ambient air quality standards, the centerpiece of the statute. Section 109 directs the agency to set those standards at levels “requisite to protect the public health,” allowing “an adequate margin of safety.” In 1997 the agency revised the standards for ozone and particulate matter, tightening them on the basis of an extensive scientific record. Industry groups and several states challenged the revisions, and the D.C. Circuit handed them a startling victory: it held that section 109, as construed by the agency, effected an unconstitutional delegation of legislative power, because the statute supplied no intelligible principle to guide the agency’s choice among the range of possible standards.

The Supreme Court reversed, unanimously, in an opinion by Justice Scalia. The decision contains two holdings, and each deserves separate attention. The first rejected the nondelegation challenge. The Court held that the “requisite to protect the public health” standard, qualified by the margin of safety language, supplies an intelligible principle that is, if anything, more determinate than delegations the Court had previously upheld. The opinion surveyed the history of nondelegation doctrine and found that Congress routinely delegates with far vaguer guidance; by comparison, a directive to set standards at the level required to protect health, with a safety margin, gives the agency a genuine, judicially reviewable criterion. The D.C. Circuit’s contrary conclusion, the Court explained, mistook the difficulty of the scientific judgment for the absence of a legal standard.

The second holding is the one that shapes the statute to this day. The challengers argued that the agency must or at least may consider the costs of implementation when setting the standards. The Court held that it may not. Section 109(b)(1) instructs the agency to set standards requisite to protect public health, and it says nothing about costs. Elsewhere in the Act, Congress expressly directed the agency to consider costs, feasibility, and economic impact, which showed that Congress knew how to write such a directive when it wanted one. The contrast was dispositive. The opinion reasoned that Congress does not make major structural choices in obscure corners of a statute; having built cost considerations into other provisions explicitly, Congress would not have hidden a cost mandate, or even a cost permission, in the silence of section 109. The standards are therefore set on health grounds alone, and costs enter the picture only later, when states choose the means of attaining the standards through their implementation plans.

Why the cost holding is the single most consequential interpretive holding about the Act requires spelling out, because its consequences are easy to underestimate. First, it locks in the statute’s moral structure. The national standards reflect a judgment about what the air must be like for people to breathe safely, and that judgment is not discounted by the price of achieving it. A standard that permitted cost balancing would be a different statute, one in which health is one value among others; the statute Congress wrote, as the Court read it, treats health as the value and cost as a problem for the implementation stage. Second, it explains the Act’s stringency. Because the standards are set without regard to cost, they can demand reductions that are enormously expensive, and the political and economic fights then migrate to the state plans, the technology standards, and Congress itself, exactly the migration Union Electric had mapped a quarter century earlier. Third, it disciplines the agency in both directions. The agency cannot weaken a standard because compliance would be costly, but it also cannot strengthen a standard beyond what health requires in order to pursue other goals; the health criterion cuts both ways.

The Court had taken the same interpretive posture under a different statute a generation earlier, and the parallel is instructive. In TVA v. Hill, the snail darter decision, the Court read the Endangered Species Act as demanding that federal agencies ensure their actions do not jeopardize listed species, even when a nearly completed dam stood in the balance, because the statutory command left no room for cost balancing. Whitman does the same work for the Clean Air Act’s health standards. In both cases the Court treated Congress’s omission of cost language as a deliberate choice rather than an oversight, and in both cases the result was a statute with sharper teeth than its regulated community had hoped.

The 1997 rulemaking that produced the litigation shows the cost holding’s practical bite. The agency’s Administrator at the time, Carol Browner, proposed replacing the one hour ozone standard of 0.12 parts per million with an eight hour standard of 0.08 parts per million, reflecting evidence that longer exposures at lower concentrations harmed health, and establishing the first ever standard for fine particulate matter, particles smaller than 2.5 micrometers, which the scientific record linked to premature mortality. The rulemaking record ran to thousands of pages of epidemiological studies, staff papers, and advisory committee reviews. Industry commenters argued that the health benefits were uncertain and the compliance costs enormous, with estimates running into the tens of billions of dollars annually. The agency’s response, sustained by the Supreme Court, was that the statute made those arguments legally irrelevant to the standard setting decision. The numbers could be debated when states wrote their implementation plans, when the agency set technology based requirements for specific source categories, and when Congress considered amending the law, but not when the agency answered the section 109 question of what the air must be like.

The D.C. Circuit’s interlude deserves attention because of how unusual it was. Judge Williams, writing for the panel, did not simply remand for a better explanation. He held that the agency’s construction of section 109 effected an unconstitutional delegation of legislative power, the first time in decades a federal court had invalidated agency action on nondelegation grounds, and he sent the standards back for the agency to adopt a narrower, constitutionally adequate interpretation of its own authority. The move was inventive: rather than striking the statute down, the court invited the agency to save it by reading in the limiting principle the court thought missing. The Supreme Court’s unanimous reversal treated the invention as unnecessary. Section 109 already contained the limiting principle, in the words “requisite to protect the public health” and the margin of safety clause, and the Court measured that language against the delegations it had previously sustained, from “public interest” to “fair and equitable,” finding section 109 at least as determinate. The episode stands as a caution about lower courts using constitutional avoidance to rewrite regulatory statutes: the Supreme Court preferred to read the statute as written and uphold it.

Why does the Clean Air Act forbid cost considerations when EPA sets air quality standards?

Section 109 orders EPA to set standards at levels requisite to protect public health, never mentioning cost, while other provisions mention cost expressly. The Court read that silence as deliberate: standard setting is a health judgment, and costs enter only when states choose how to meet the standards.

The depth behind that answer includes an important qualification that is often missed. Whitman does not hold that costs are irrelevant to the Clean Air Act as a whole. It holds that costs are irrelevant to one specific decision: the setting of the national ambient air quality standards under section 109. Everywhere else in the statute, the cost question is answered provision by provision, on the text of each provision. When Congress wrote technology based standards for new sources, it told the agency to consider cost explicitly. When Congress wrote the hazardous air pollutant program, it built cost and feasibility into the standard setting machinery. The Court’s method is thus granular rather than global: read each provision for what it says about costs, and do not borrow a cost permission from one section to soften another. That granularity was confirmed in later decisions. In Michigan v. EPA, decided in 2015, the Court held that the agency must consider cost when deciding whether regulation of power plant mercury emissions is “appropriate and necessary” under section 112, precisely because that phrase, unlike section 109’s health language, invites a broader judgment. The contrast vindicates Whitman’s approach: the statute is not cost blind in general, but section 109 is cost blind by design, and the design is Congress’s.

The margin of safety concept, which the nondelegation discussion treated as part of the intelligible principle, has its own substantive importance. Section 109 requires standards that protect public health with “an adequate margin of safety,” which the agency has long read as authority to guard against scientific uncertainty and to protect sensitive subpopulations such as children, the elderly, and people with respiratory disease. The margin is not a license to set standards at zero; the word “adequate” constrains it, and the agency must explain why the chosen margin fits the evidence. But it does mean the standards need not wait for perfect science, a point that mattered enormously for pollutants like fine particles, where the epidemiological evidence was strong but the biological mechanisms were still being worked out. After Whitman, the 1997 standards went into effect, states began the long work of writing implementation plans for the eight hour ozone and fine particle standards, and the agency revisited the standards on the statute’s five year review cycle, tightening the ozone standard to 0.075 parts per million in 2008 and to 0.070 in 2015. Each revision replayed the same legal structure: health science in, cost arguments out at the standard setting stage, costs addressed in implementation.

The implementation stage, where costs finally enter, shows the other half of the design. Once the health based standards are set, each state must produce a plan demonstrating attainment, choosing among controls on factories, vehicles, fuels, and area sources, and here cost is not only permitted but central: states routinely pick the cheapest mix of measures that achieves the required reductions, and the agency approves cost effective strategies. The technology based standards for new sources, under sections 111 and 112, likewise build cost into their definitions, requiring the best demonstrated systems with explicit attention to economic impact. The statute is thus a machine with two chambers: a health chamber where costs are forbidden, and an implementation chamber where costs are decisive. Whitman’s achievement was to keep the chambers separate, so that neither the agency nor the courts could use the practical difficulties of the second chamber to dilute the health judgment of the first.

The chambers metaphor also explains why later cost disputes under the Act kept returning to Whitman for their vocabulary. Litigants challenging technology based standards argued that Congress’s explicit cost language in those provisions proved costs were always relevant; the agency answered, citing Whitman, that each provision’s text governs its own decision, and that borrowing a cost inquiry from one section to soften another repeats the error the 2001 Court rejected. That provision by provision method became the standard grammar of Clean Air Act cost litigation, invoked by both sides through the mercury, interstate transport, and greenhouse gas rulemakings, and it is the reason the Michigan majority could demand threshold cost consideration under section 112 without disturbing the health only rule of section 109.

The nondelegation half of Whitman deserves a final word because it closed a door that the D.C. Circuit had tried to open. Had the nondelegation challenge succeeded, every health based standard under the Act would have been vulnerable, and Congress would have faced the impossible task of specifying numerical pollution limits in statutory text. The unanimous rejection reaffirmed that broad delegations guided by intelligible principles are constitutional, and it did so in a way that left the agency’s scientific role intact while keeping the ultimate policy choice, the health only criterion, firmly with Congress. The decision thus completed the triangle that the road from 1976 to 2001 had been building. Union Electric said the statute’s demands cannot be softened by pleading hardship. Chevron said that where the statute is genuinely unclear, the agency’s reasonable reading governs. Whitman said that where the statute speaks in health only terms, neither the agency nor the courts may smuggle costs into the standard. With those three principles in place, the stage was set for the question Massachusetts v. EPA would put to the Court: whether the broadest words in the statute, the definition of “air pollutant” itself, unambiguously cover the gases warming the planet, and what follows when they do.

Massachusetts v. EPA: The 2007 Decision, in Depth

The earlier sections of this article traced the statute from its identity through three landmark rulings on feasibility, deference, and cost. What follows picks up where that account leaves off, with the decision that carried the Clean Air Act into the climate era. Massachusetts v. EPA, 549 U.S. 497 (2007), is the most cited environmental ruling of the modern Supreme Court, and it is also among the most miscited. The opinion did two large things and pointedly declined to do a third, and keeping those boundaries straight is the difference between understanding the case and merely invoking it.

The dispute began with a petition, not a rule. In 1999, the International Center for Technology Assessment joined by other organizations asked the Environmental Protection Agency to regulate greenhouse gas emissions from new motor vehicles under section 202(a)(1) of the act. That provision directs the Administrator to set emission standards for “any air pollutant” from new motor vehicles which, in the Administrator’s judgment, causes or contributes to air pollution that may reasonably be anticipated to endanger public health or welfare. The petition sat for years while the agency gathered comment. In 2003 the agency denied it, giving two reasons. First, the agency said it lacked statutory authority to regulate greenhouse gases at all, because carbon dioxide and its kin were not “air pollutants” within the meaning of the law. Second, the agency said that even if it possessed the authority, it would decline to exercise it, citing a set of policy considerations: the desirability of a comprehensive approach rather than piecemeal vehicle standards, the existence of voluntary programs, ongoing scientific uncertainty, and the risk that unilateral American action would complicate foreign policy. A divided panel of the D.C. Circuit upheld the denial in 2005, and the Supreme Court granted review.

The appellate decision the justices agreed to review had itself been fractured. In July 2005, a divided panel of the D.C. Circuit rejected the petitioners’ challenge, with Judge Randolph concluding that the agency had permissibly declined to make the endangerment judgment and Judge Sentelle concluding that the petitioners lacked standing altogether. Judge Tatel dissented at length, arguing that the agency’s policy rationales were no substitute for the scientific determination the statute demanded. The Tatel dissent functioned as a draft of the Supreme Court majority opinion two years later, and several of its formulations reappeared in Justice Stevens’s reasoning, a reminder that appellate dissents often write the first version of the law to come.

Before reaching the substance, the Court had to decide whether the challengers belonged in court at all. Standing doctrine requires a plaintiff to show a concrete injury, a causal connection to the defendant’s conduct, and a likelihood that a favorable decision will redress the harm. Climate change would seem to defeat all three, because its causes are diffuse and its effects are shared by the entire planet. Justice Stevens, writing for a five justice majority, held that Massachusetts cleared the bar, and the reasoning turned on the identity of the plaintiff. “States are not normal litigants for the purposes of invoking federal jurisdiction,” the opinion observed, and Massachusetts came to court asserting what the majority called a quasi-sovereign interest in protecting the territory within its borders. When the commonwealth entered the Union it surrendered certain sovereign prerogatives, including the power to negotiate emissions treaties with foreign nations or to invade a neighboring state to stop pollution at its source. In exchange, the majority reasoned, it gained the right to invoke the federal courts, and that exchange entitled the commonwealth to what the opinion termed “special solicitude” in the standing inquiry.

The injury itself was concrete rather than abstract. Massachusetts owns a substantial stretch of coastal land, and the majority credited affidavits showing that rising seas had already begun to swallow portions of it, with the prospect of far greater losses as warming continued. That was not a generalized grievance about the state of the world; it was damage to the plaintiff’s own property, the classic form of injury that courts have recognized since the founding. Causation followed from the agency’s own concession that motor vehicle emissions in the United States contribute to the atmospheric concentration of greenhouse gases. The majority acknowledged that American vehicles are a fraction of the global total, but it refused to treat that fraction as legally nothing. Agencies, the opinion explained, do not get to avoid obligations simply because a problem is large; a small step toward addressing a massive harm still counts as a step. Redressability was satisfied on the same incremental logic. No one claimed that regulating new American vehicles would reverse sea level rise by itself. The claim was only that such regulation would slow the pace of loss, and slowing the pace of an ongoing injury is redress enough. The majority distinguished the Court’s earlier restrictive standing decision in Lujan v. Defenders of Wildlife, where individual plaintiffs had asserted an undifferentiated interest in the lawful administration of the laws. Massachusetts was asserting damage to its own coastline, and that difference decided the question.

The standing holding’s afterlife proved narrower than either side predicted. Later litigants invoked “special solicitude” to open courthouse doors for a range of state led challenges, and lower courts divided over whether the solicitude was a thumb on the scale for all state plaintiffs or a description of the quasi-sovereign interest at stake in that particular case. The Supreme Court itself never repudiated the language, but neither did it extend the reasoning to private plaintiffs asserting diffuse climate injuries; individuals and organizations continued to face the full rigor of the traditional three part test. The practical result was a two track standing regime in environmental litigation: states suing over harms to their territory entered with a recognized advantage, while everyone else had to show the kind of particularized injury the Massachusetts majority had found in coastal erosion. That asymmetry shaped the plaintiff lineup of nearly every major climate case that followed, with state attorneys general rather than private groups carrying the jurisdictional weight.

Why could Massachusetts get into court over rising seas?

The Court held 5-4 that Massachusetts had standing to challenge EPA’s refusal to regulate greenhouse gases. States receive “special solicitude” in standing analysis, and the Commonwealth showed concrete injury through coastal land loss to rising seas, a causal link to the agency’s inaction, and redressability through incremental emission cuts.

On the merits, the majority confronted the agency’s claim that it lacked authority over greenhouse gases, and here the statutory text did the decisive work. Section 302(g) defines “air pollutant” to mean “any air pollution agent or combination of such agents, including any physical, chemical, biological, radioactive … substance or matter which is emitted into or otherwise enters the ambient air.” The majority read that language as deliberately capacious. Carbon dioxide, methane, nitrous oxide, and fluorinated gases are physical and chemical substances emitted into the ambient air; nothing in the definition excludes them, and the repeated use of the word “any” signaled a definition meant to sweep broadly rather than to enumerate. The majority rejected the agency’s argument that Congress could not have meant to include climate gases because the regulatory machinery of the act was designed with conventional pollutants in mind. Awkward fit, the opinion reasoned, is not the same as no fit, and it is the text that controls, not assumptions about what the drafters envisioned. These provisions operate within the broader architecture of the statute, whose ambient standard machinery is examined in the companion account of the NAAQS framework.

Having established that greenhouse gases fall within the agency’s jurisdiction, the majority turned to the agency’s second reason for inaction, the policy arguments, and found them legally weightless. Section 202(a)(1) frames the regulatory judgment in scientific terms: the Administrator is to decide whether the pollutant causes or contributes to air pollution that may reasonably be anticipated to endanger public health or welfare. The statute lists the considerations that may support declining to regulate, and foreign policy preferences, voluntary industry programs, and a desire for a comprehensive approach are not among them. The majority did not say that policy can never matter; it said that when Congress writes a decision criterion into a statute, the agency must answer in those terms or explain why the science is too uncertain to permit a judgment. What the agency may not do is substitute its own sense of wise governance for the inquiry Congress assigned.

The agency’s policy reasons deserve a closer look, because the majority’s treatment of each one set the template for reviewing discretionary inaction. The first reason, that piecemeal vehicle standards would interfere with a comprehensive approach, failed because the statute commands a pollutant by pollutant, source category by source category inquiry; comprehensiveness is not a listed criterion. The second, that voluntary programs were already addressing the problem, failed because Congress had not made voluntarism a substitute for the endangerment judgment. The third, scientific uncertainty, was the closest to a legitimate answer, and the majority left the door open: had the agency said the science was too uncertain to form a judgment, that would have been a statutorily grounded response. What the agency had said instead was that uncertainty counseled caution as a policy matter, which is a different claim. The fourth, foreign policy, failed because the statute assigns foreign affairs to no one in section 202 and the agency offered no analysis connecting its vehicle standards to diplomatic outcomes. Together the four rejections established a durable rule of administrative law: when Congress writes the question, the agency must answer the question.

The judgment was therefore reversed and the matter remanded, with instructions that the agency either make the endangerment determination the statute calls for or offer a reasoned explanation, grounded in the statute, for why it could not do so.

This is the point where the miscitation begins, and it deserves emphasis because the entire downstream history turns on it. The 2007 ruling did not order the regulation of greenhouse gases. It did not set an emission standard, did not command the agency to find endangerment, and did not hold that any particular control technology was required. What it ordered was an answer: the agency had to decide the scientific question Congress posed, in the terms Congress posed it, rather than declining on policy grounds the statute never authorized. The obligation to regulate followed later, and only because of what the agency itself concluded when it finally made that determination. In December 2009, the Administrator issued the endangerment finding, concluding that greenhouse gases from motor vehicles do endanger public health and welfare, and that finding triggered the mandatory duties the statute attaches to such a conclusion, including the joint vehicle emission standards finalized the following year. The distinction between the 2007 remand and the 2009 finding decides how the case may honestly be cited. Massachusetts v. EPA is authority for the proposition that the agency has jurisdiction over greenhouse gases, and it is authority for the proposition that discretionary refusals to act must be grounded in statutory criteria. It is not authority for the proposition that the Court itself imposed climate regulation, because the Court did the opposite of imposing: it sent the question back and insisted that the agency’s own scientific judgment, not the justices’ policy preferences, supply the answer. Every later litigant who cites the decision for more than that is borrowing a weight the opinion declined to carry.

The remand’s narrowness also explains the standing fight’s lasting bitterness. Because the Court ordered only an answer rather than a rule, the practical stakes of the jurisdictional holding were lower than the dissenters’ rhetoric suggested: even a plaintiff who cleared the standing bar would receive, at most, a reasoned agency explanation, not an emission standard. Yet the dissenters understood that the answer, once compelled, would be hard for any administration to give in the negative, since the scientific record the agency had already compiled pointed one way. The standing holding thus functioned as a gate whose real significance lay in what waited beyond it, and the fight over the gate was fierce precisely because both sides could see the corridor.

The chain that ran from the remand shows why the jurisdiction versus obligation distinction is not academic. The 2009 finding made greenhouse gases “subject to regulation,” and that phrase in turn pulled the permitting programs into the picture, producing the Tailoring Rule and the Utility Air Regulatory Group litigation over whether the agency could rewrite the statute’s thresholds to accommodate a pollutant the thresholds were never designed to measure. None of those downstream fights was ordered by the 2007 Court; each arose from the agency’s own subsequent determinations under standards the statute supplied. This is why careful lawyers cite Massachusetts v. EPA for the proposition that the courthouse door is open and the statute reaches climate pollutants, while citing the endangerment finding, not the decision, for the proposition that regulation was required.

The two dissents deserve presentation in their own terms, because each captured a genuine cost of the majority’s approach. Chief Justice Roberts, joined by Justices Scalia, Thomas, and Alito, dissented on standing, and his objection went to the structure of the federal courts. In his telling, the majority had converted an injury shared by every coastal landowner on earth into a ticket to federal court by the expedient of calling the plaintiff a state. The “special solicitude” language had no pedigree in prior decisions, the dissent argued; it was invented for this case to reach a desired result. Generalized grievances about the atmosphere belong to the political branches, where all citizens are represented, not to courts, where the remedy for a global phenomenon will always be partial and symbolic. On redressability, the dissent pressed the arithmetic the majority had minimized: even eliminating every American vehicle emission would leave the overwhelming bulk of global greenhouse gases untouched, because the growth in emissions from developing economies dwarfed any plausible American reduction. A court order that cannot meaningfully change the plaintiff’s situation, the dissent concluded, is an advisory opinion dressed in standing’s clothing, and the majority’s incrementalism licensed lawsuits over any problem too big for any one actor to solve.

Justice Scalia, joined by the Chief Justice and Justices Thomas and Alito, dissented on the merits, and his argument was about who gets to read ambiguous statutes. The majority had treated the definition of “air pollutant” as unambiguous, but the dissent insisted that the phrase takes its meaning from the company it keeps. Throughout the act, “air pollutant” appears in provisions built around conventional pollutants subject to ambient standards, state implementation plans, and technology based controls; reading the term to sweep in every climate gas, the dissent argued, would produce absurdities, subjecting countless small sources to permitting programs designed for smokestacks. Where a statute can reasonably be read in two ways, the dissent continued, the agency’s reasonable reading deserves deference under the framework the Court itself had established in Chevron, and the agency’s conclusion that Congress had not delegated climate policy to it through a definitional subsection was at minimum reasonable. The majority’s real error, in this telling, was not misreading a definition but misunderstanding restraint: the decision to defer a rulemaking for prudential reasons is the ordinary stuff of administration, and courts that second guess such timing judgments turn themselves into a council of revision over the executive branch. The two dissents together framed the lasting critique of the decision, that it opened the courthouse to sweeping policy litigation and narrowed the space in which agencies could say no. The majority’s answer, embedded in the opinion’s structure, was that the courthouse was already open to states defending their territory and that saying no was always permitted, provided the no was spoken in the statute’s own language.

The Consequences, in Order

The 2007 ruling settled two questions, whether greenhouse gases were within the agency’s jurisdiction and whether policy discomfort could substitute for a statutory answer, and in doing so it opened four more. Each of the decisions that followed tested a different boundary of the authority Massachusetts v. EPA had confirmed, and taken together they trace a complete arc: from displacement of judge made law, through the limits of agency self help, through the demand that costs be faced openly, to the doctrine that polices the outer edge of regulatory ambition.

Why did the Court channel climate claims from judges to the agency?

The Court held 8-0 that the Clean Air Act displaces federal common law nuisance claims seeking judicial caps on power plant carbon dioxide. Because Congress assigned EPA the job of deciding whether and how to limit those emissions, judges may not set limits through tort suits. State law claims were left undecided.

American Electric Power Co. v. Connecticut, 564 U.S. 410 (2011), reached the Court only four years after Massachusetts, and it answered the question the earlier decision had left hanging: if the statute empowers the agency to address greenhouse gases, may courts address them too, through the older and more flexible instrument of federal common law? Eight states and New York City had sued five large power companies under a federal common law theory of public nuisance, asking a federal judge to cap the companies’ carbon dioxide output and then ratchet the caps downward year by year. The theory was straightforward. Nuisance law had long allowed courts to abate unreasonable interferences with public rights, and before Congress built the modern regulatory state, federal courts routinely filled gaps in pollution control with judge made standards. The plaintiffs argued that the gap here was real, because the agency had not yet set any emission limits for existing power plants, and that courts should act until it did.

Justice Ginsburg, writing for a unanimous Court of eight, with Justice Sotomayor recused because she had sat on the Second Circuit panel below, rejected the invitation. Federal common law is displaced, the opinion explained, when Congress has “spoken directly” to the question at issue, and Congress had spoken directly here by assigning the Environmental Protection Agency the task of deciding whether and how to regulate carbon dioxide from power plants. The displacement test does not ask whether the agency has finished the job; it asks whether Congress gave the agency the job. Massachusetts v. EPA had already confirmed that greenhouse gases are air pollutants under the act, and the act’s provisions for new and existing stationary sources supplied the machinery for addressing them. That Congress had entrusted the field to expert administration meant courts could not run a parallel nuisance regime alongside it, with different judges setting different caps for different defendants on different records. The opinion was careful about what it did not decide. It left open whether state common law nuisance claims might survive, a question that would occupy lower courts for years, and it said nothing about whether the agency’s eventual standards would be adequate. On the threshold question of standing the Court divided evenly, four to four, which left the Second Circuit’s jurisdictional ruling in place without creating precedent. The practical effect was a clean channeling rule: climate grievances against emitters would henceforth be addressed through the statute’s administrative process, with judicial review of agency action, rather than through direct suits asking judges to write emission standards from the bench.

The displacement holding drew on a lineage older than the environmental statutes. Federal common law survives only in the interstices of congressional action, and the Court had held a generation earlier, in cases concerning water pollution, that comprehensive legislation displaces judge made remedies when it speaks directly to the problem. Justice Ginsburg’s opinion applied that test without sentimentality: the question was not whether the agency had solved climate change but whether Congress had given the agency the tools and the responsibility. It had. The opinion also noted that the agency was already moving, the endangerment finding issued and the vehicle standards final, with existing source standards under study, which made the displacement especially clean; courts would not build a parallel track while the designated track was in use. The four to four division on standing left the appellate court’s jurisdictional analysis standing but created no precedent, a formal curiosity that underscored how contested the courthouse door remained even as the substantive holding commanded unanimity. And the reservation concerning state law proved prescient: with the federal common law door closed, later plaintiffs tried state nuisance theories in state courts, producing a second wave of litigation over preemption and removal that the 2011 decision had deliberately left for another day.

What stopped the agency from rewriting the permitting thresholds?

The Court held 5-4 that EPA could not rewrite the act’s permitting thresholds of 100 and 250 tons per year to avoid sweeping millions of small sources into the program. Statutory numbers are not the agency’s to revise, even when literal application creates administrative strain. Sources already in the program must still address greenhouse gases.

Utility Air Regulatory Group v. EPA, 573 U.S. 302 (2014), tested what happens when a lawful assertion of jurisdiction collides with statutory machinery that was never built for the pollutant in question. The sequence mattered. After the 2009 endangerment finding, the agency issued emission standards for new motor vehicles, and under the act’s structure those vehicle standards automatically made greenhouse gases “subject to regulation.” That phrase in turn triggered two permitting programs: prevention of significant deterioration review for new and modified major sources, and the Title V operating permit program. Both programs keyed their coverage to numerical thresholds written into the statute decades earlier, 100 tons per year for certain listed source categories and 250 tons per year for others. Applied to conventional pollutants, those numbers captured a manageable universe of large industrial facilities. Applied to carbon dioxide, they captured millions of sources, from apartment buildings to small commercial kitchens, an administrative impossibility the agency itself acknowledged. To avoid that result, the agency issued what it called the Tailoring Rule, rewriting the statutory thresholds upward to 75,000 and 100,000 tons per year for greenhouse gases, phasing in coverage over time. The agency’s candor was striking: it conceded that the statute as written could not sensibly be applied, and it asked the Court to bless a pragmatic rewrite.

Justice Scalia, writing for a five justice majority on the central question, refused. The opinion’s core principle was simple and, in the majority’s telling, nonnegotiable: an agency may not rewrite unambiguous statutory terms to suit its own policy needs, even when the literal application of those terms produces absurd results. The numbers 100 and 250 were Congress’s numbers, enacted through the legislative process, and the executive branch has no power to amend them by rulemaking. The majority acknowledged the agency’s dilemma but located the fault in the agency’s own prior step. Greenhouse gases became subject to the permitting programs only because the agency had chosen to treat the vehicle standards as the trigger, and having created the collision, the agency could not resolve it by editing the statute. The opinion added a warning that would echo through the later cases: when an agency claims to discover in a long extant statute an unheralded power to regulate a significant portion of the American economy, courts greet the announcement with skepticism, because Congress is expected to speak clearly before assigning decisions of vast economic and political significance. That sentence, planted in a case about permitting thresholds, became the seedbed of the major questions doctrine.

The majority’s interpretive method in the case carried implications beyond the thresholds. Justice Scalia’s opinion treated the phrase “air pollutant” as a chameleon, taking its color from the surrounding provisions: capacious in the definitional section construed in Massachusetts, narrower in the permitting programs where Congress had written numerical thresholds aimed at large industrial sources. That move answered the dissent’s charge of inconsistency. The 2007 decision had read the definition broadly because the definition’s text was broad; the 2014 decision read the same words narrowly in a different title because the surrounding machinery showed what Congress meant there. Whether that reconciliation persuades depends on one’s theory of interpretation. Defenders called it context sensitive textualism, the ordinary work of reading words in their setting. Critics called it result oriented, a way to expand jurisdiction in 2007 and contract it in 2014 while claiming fidelity to text both times. The debate over which description fits continues to structure arguments about how the act’s general language should meet its specific programs.

The judgment was more nuanced than the headline suggests, and the nuance mattered for what the agency could still do. While five justices held that greenhouse gases alone could not trigger the permitting programs, seven justices agreed that sources already subject to review for conventional pollutants, the so called anyway sources, could be required to apply best available control technology to their greenhouse gas emissions as well. The agency thus kept a meaningful, if narrower, foothold: any large new facility already in the permitting system would still have to address its carbon output. Justice Breyer, joined by Justices Ginsburg, Sotomayor, and Kagan, concurred in part and dissented in part, defending the Tailoring Rule as a reasonable accommodation of conflicting statutory commands. In his telling, the agency faced a genuine impossibility, apply the thresholds literally and drown the program, or decline to regulate greenhouse gases at all and defy the statute’s command, and choosing a phased middle path was exactly the kind of reasonable gap filling that deference doctrines were built to protect. Justice Thomas, joined by Justice Scalia, concurred to argue the agency should have gone further and disclaimed greenhouse gas authority over the permitting programs entirely. The competing opinions framed a permanent tension in administrative law: whether fidelity to text means enforcing every number Congress wrote even when the numbers misfire, or permitting the expert agency to adapt old machinery to new pollutants without pretending the adaptation is what Congress enacted.

The practical footprint of the split judgment deserves attention, because the case is often remembered only for the rebuke. The Tailoring Rule had proceeded in phases, first covering only the largest sources already in the permitting system, then stepping down to the rewritten thresholds; the majority’s invalidation of the rewrite left the first phase’s logic intact through the anyway sources holding. In practice, that meant every major new industrial facility already undergoing prevention of significant deterioration review for conventional pollutants would also face best available control technology analysis for its greenhouse gases. For the power sector and heavy industry, the economic heartland of the dispute, the agency’s leverage survived in substantial measure. What died was the agency’s claim to set its own jurisdictional boundaries by editing numbers Congress had enacted. The Breyer opinion in partial dissent warned that the majority’s formalism would punish precisely the agency behavior courts should encourage: honest confrontation of statutory awkwardness through transparent, phased implementation. The majority’s reply, embedded in the opinion’s structure, was that transparency about rewriting the law does not make rewriting the law lawful, and that the remedy for an unworkable statute lies with the legislature that wrote it.

Michigan v. EPA, 576 U.S. 743 (2015), moved the fight from the scope of jurisdiction to the honesty of the regulatory calculus. The provision at issue was section 112(n)(1)(A), part of the hazardous pollutant program created by the 1990 amendments, which directed the agency to regulate emissions of toxic pollutants from electric power plants if the Administrator found such regulation “appropriate and necessary.” The agency made that finding without considering cost at all, reasoning that costs would be weighed later, when the actual emission standards were written. The resulting rule, the Mercury and Air Toxics Standards finalized in late 2011, imposed roughly $9.6 billion per year in compliance costs on the power sector, while the agency’s own analysis put the direct, quantifiable benefits from reducing the targeted hazardous pollutants at only a few million dollars per year. The agency pointed to tens of billions in co-benefits from incidental reductions in fine particulate matter and sulfur dioxide, but the challengers, led by the state of Michigan and industry groups, argued that the threshold decision itself had been made with eyes closed to price.

Justice Scalia, writing for five justices, held that the phrase “appropriate and necessary” necessarily includes consideration of cost. The opinion’s logic was deliberately commonsensical: no reasonable decisionmaker calls an action appropriate without weighing what it costs, and a statute that uses the word “appropriate” imports that ordinary understanding. “One would not say that it is even rational, never mind ‘appropriate,’ to impose billions of dollars in economic costs in return for a few dollars in health or environmental benefits,” the majority observed, and it rejected the agency’s argument that cost could be deferred to a later stage. The threshold finding was the gateway to the entire regulatory edifice; a gate that ignores price is not a gate but a formality. The majority did not demand any particular methodology, formal cost benefit analysis was not required, but it insisted that cost be on the table from the beginning rather than smuggled in after the commitment to regulate had been made.

The cost benefit arithmetic behind the rule generated its own controversy, one the majority opinion acknowledged without resolving. The agency’s estimate of a few million dollars in direct benefits from reducing mercury and other toxics looked indefensible beside $9.6 billion in costs, but the agency’s full accounting claimed $37 billion to $90 billion in co-benefits, chiefly from the incidental reduction of fine particulate matter as plants installed controls or retired. Defenders argued that counting co-benefits was simple honesty: a rule that saves lives through cleaner air should get credit for those lives regardless of which pollutant label attaches to them. The counterargument ran that the maneuver allowed the agency to justify any expensive rule by pointing to particulate reductions it could have pursued directly, turning the hazardous pollutant program into a vehicle for ends Congress had assigned elsewhere. The majority did not rule the co-benefit methodology unlawful; it held only that the threshold finding could not ignore cost. The deeper question, how honestly an agency must match its stated rationale to its real one, was left for future rulemakings and future lawsuits. The decision sent the rule back to the agency without vacating it outright, leaving the standards in place while the agency repaired the reasoning.

Justice Kagan, joined by Justices Ginsburg, Breyer, and Sotomayor, dissented with equal force and a different theory of how regulation actually works. In her telling, the majority had fixated on a single preliminary finding while ignoring the full rulemaking record, in which the agency had considered costs exhaustively at every subsequent stage, from the setting of the emission floors to the compliance timelines to the cost effectiveness of control technologies. The word “appropriate,” she argued, is capacious by design, and Congress’s choice of that word rather than a cost benefit formula signaled trust in the agency’s judgment about when and how to weigh expense. The dissent also stressed a practical reality the majority had minimized: by the time the Court ruled, the power sector had already spent most of the compliance money and the great bulk of the regulated plants were meeting the standards, which made the majority’s demand for threshold cost consideration feel less like a guardrail and more like a belated lecture. The two opinions thus offered competing visions of rationality in government. For the majority, rationality meant facing the price before deciding to act. For the dissent, rationality meant considering costs where they could actually shape outcomes, across a multi year rulemaking, rather than performing a ritual calculation at a preliminary stage. How the agency internalized that lesson, building cost analysis into the earliest phases of its rulemakings, is examined in the companion account of EPA rulemaking practice.

The rule at the center of Michigan had been among the most consequential the agency ever issued. The Mercury and Air Toxics Standards set the first national limits on mercury, arsenic, chromium, and other hazardous pollutants from coal and oil fired power plants, requiring controls such as activated carbon injection and scrubbers across a fleet that had never faced such limits. Industry compliance spending ran into the billions, and dozens of aging coal units retired rather than retrofit, a transformation the agency counted among the rule’s achievements and its critics counted among its costs. On remand, the agency issued a supplemental finding in 2016 that considered cost and reaffirmed the appropriate and necessary determination, this time with a formal accounting of compliance expense against monetized benefits. The rule itself was never vacated, and by the time the supplemental finding issued, the great majority of regulated units had already installed the required controls, which meant the litigation’s practical legacy lay less in any ton of mercury avoided than in the procedural template it imposed. Every subsequent major rulemaking would carry a threshold cost discussion in its opening pages, a direct inheritance of the majority’s insistence that price be faced before commitment.

How did the major questions doctrine decide the Clean Power Plan fight?

The Court held 6-3 that section 111(d) did not authorize EPA’s Clean Power Plan. Under the major questions doctrine, an agency claiming power over matters of vast economic and political significance must show clear congressional authorization, and a phrase about the “best system of emission reduction” was not clear enough to support generation shifting.

West Virginia v. EPA, 597 U.S. 697 (2022), is the decision that gave the modern limit its name and its shape. The rule at issue was the Clean Power Plan, finalized in 2015, which set emission guidelines for existing power plants under section 111(d), the provision governing existing sources of pollutants not covered by the ambient standards or the hazardous pollutant program. The heart of the plan was the agency’s determination of the “best system of emission reduction,” the statutory phrase that anchors section 111. Historically, that phrase had meant measures a plant could apply to its own operations: efficiency upgrades, better maintenance, cleaner burning techniques, improvements inside the fence line of the facility. The Clean Power Plan kept one such measure, heat rate improvements at coal plants, as its first building block, but added two more of a different character: shifting electricity generation from coal plants to natural gas plants, and shifting it further to renewable sources. The plan thus set state level goals that could be met not by cleaning up individual plants but by changing the mix of plants that generated the nation’s electricity. The rule never took effect; the Court stayed it in February 2016 before the lower courts had even reviewed it, an extraordinary intervention, and it was later repealed and replaced before the 2022 decision. The Court nevertheless reached the merits, reasoning that the agency’s claimed authority remained live.

Chief Justice Roberts, writing for six justices, held that the plan exceeded the agency’s lawful power, and the reasoning proceeded through the major questions doctrine stated in the Court’s own terms. In certain extraordinary cases, the majority explained, separation of powers principles and a practical understanding of how Congress legislates make courts reluctant to read a transformative delegation into ambiguous text. Where an agency asserts authority of vast economic and political significance, the Court applies a demanding standard: the agency must point to “clear congressional authorization” for the power it claims. The Clean Power Plan, in the majority’s telling, was exactly such a case. The agency had discovered in an ancillary provision, one it had used only a handful of times in fifty years, the power to set energy policy for the entire nation: deciding how much coal generation should exist, effectively restructuring an industry that Congress had repeatedly declined to restructure through cap and trade legislation. The phrase “best system of emission reduction” was not, the majority concluded, the kind of clear statement that could bear that weight. Read in context, the provision concerned emission reductions achievable by the regulated sources themselves, and generation shifting, which achieves reductions by moving production to different sources altogether, lay outside its natural meaning. The majority emphasized the unprecedented character of the claim: never before had the agency used section 111 to force a sector wide transition, and the absence of any prior assertion of such power counted against the agency rather than for it.

Justice Gorsuch, joined by Justice Alito, concurred to systematize the doctrine, describing the circumstances that trigger its application and the reasons, rooted in the constitutional allocation of lawmaking power, that justify it. His opinion treated the doctrine not as a novel invention but as the mature form of a longstanding judicial skepticism toward agency claims of extravagant statutory authority.

Justice Kagan, joined by Justices Breyer and Sotomayor, dissented, and her opinion merits full presentation because it offered the strongest version of the losing position. In her telling, the majority had the text exactly backwards. Congress wrote “best system of emission reduction” in deliberately capacious terms precisely because it wanted the agency to identify the most effective means of cutting pollution as technology and circumstances evolved, and generation shifting was, as a factual matter, how the power sector actually reduced emissions: utilities routinely shifted generation among units in response to price signals, and the plan simply harnessed that existing practice. The majority’s demand for a clear statement, the dissent argued, was a judge made overlay with no basis in the statute, one that allowed the Court to displace the agency’s expert judgment whenever the stakes were high enough to make the justices uncomfortable. The dissent warned of the doctrine’s asymmetry: it would reliably frustrate regulation of major problems while leaving deregulation untouched, because only ambitious agency action triggers the clear statement demand. And it closed with a rebuke aimed at the institutional stakes, that the Court was appointing itself the decisionmaker on climate policy in place of Congress, which wrote the broad language, and the expert agency, which Congress charged with applying it. The majority’s answer, implicit in its structure, was that broad language is not the same as clear language, and that when an agency claims the power to reorder an industry, the Constitution’s assignment of lawmaking to the legislature requires more than a capacious phrase. The two positions remain the poles of the contemporary debate over the administrative state: whether generality in a statute is a delegation to be honored or a gap the courts must police.

The justiciability choice amplified the decision’s reach. By deciding the scope of section 111(d) after the specific rule had been stayed, repealed, and replaced, the majority ensured that no future administration could revive generation shifting without confronting the clear statement demand. Later rulemakings took the lesson: the replacement rules that followed stayed inside the fence line, framing their emission guidelines around measures individual plants could adopt. The doctrine thus did its work less through the invalidation of one rule than through the channeling of every rule that came after.

The Clean Power Plan’s architecture explains why the majority treated it as transformative rather than incremental. The first building block, heat rate improvements, was the traditional inside the fence line measure, worth only small emission reductions. The second and third blocks did the real work: the plan calculated each state’s emission goal on the assumption that generation would shift from coal to natural gas and then to renewables, effectively requiring states to reshape their electricity mix to comply. The agency estimated compliance costs in the billions and projected the retirement of dozens of coal plants, figures the majority cited as evidence of the plan’s vast economic significance. The February 2016 stay had already signaled the Court’s skepticism; stays of agency rules before appellate review are vanishingly rare, and the five justice vote to freeze the plan suggested where the merits would land. After the change of administration, the agency repealed the plan and replaced it with the Affordable Clean Energy rule of 2019, which returned to the inside the fence line reading, only for the D.C. Circuit to vacate that replacement in January 2021. The 2022 decision thus arrived after the specific rule had been through a full cycle of promulgation, suspension, repeal, and vacatur, and the majority decided it anyway, treating the underlying claim of authority as a live controversy capable of repetition. That justiciability choice was itself contested, and it underscored the majority’s determination to settle the scope of section 111(d) rather than let the question cycle indefinitely through successive administrations.

The dissent’s textual case rested on a simple observation about how the provision works. Section 111(d) instructs the agency to set guidelines reflecting the best system of emission reduction that has been adequately demonstrated, and then leaves the states to write the actual plans. The phrase “system,” the dissent argued, naturally encompasses approaches that operate across the electricity grid, because emission reduction in the power sector has always been a system level enterprise: grid operators dispatch plants in order of cost, shifting generation continuously, and a plant’s emissions depend as much on how often it runs as on how cleanly it burns. On that understanding, generation shifting was not a departure from the provision’s history but its fulfillment, the demonstrated system by which the industry actually cuts carbon. The majority’s inside the fence line limitation, the dissent charged, appeared nowhere in the text and reflected the justices’ own conception of proper regulation rather than Congress’s. The disagreement thus reduced to a familiar interpretive divide: whether “system” means a technology installed at a smokestack or a method by which an interconnected industry reduces its total output.

The Verdict: The Doctrine Factory

No other American statute has generated two doctrines of comparable reach in opposite directions. In 1984, a dispute over the Clean Air Act’s definition of a pollution source produced the deference framework under which agencies won reasonable interpretation fights for four decades. In 2022, a dispute over the same act’s existing source provision produced the major questions doctrine, under which agencies lose ambitious claims of power unless Congress has spoken with unmistakable clarity. Both doctrines grew from the same feature of the same law: sweeping delegations joined to enormous economic stakes. Congress wrote phrases like “best system of emission reduction” and “appropriate and necessary” and “any air pollutant” because it could not foresee every pollutant, every technology, or every crisis, and it wanted the expert agency to adapt. That drafting choice is the engine of the entire litigation history recounted here. In one era it expanded agency power, as courts honored the breadth of the delegation. In another era it contracted that power, as courts demanded that breadth be proven rather than presumed.

The earlier sections of this article showed the pattern forming before climate entered the picture. Union Electric taught that feasibility arguments cannot override clear text. Chevron taught that ambiguous text belongs, within reason, to the agency. Whitman taught that cost cannot enter where Congress excluded it. Each of those rulings priced a different clause of the same statute, and each moved in the direction of honoring the legislative bargain as written. The climate cases continued that project under greater pressure, because the stakes of honoring the bargain kept rising. Massachusetts v. EPA honored the definition Congress wrote even though the regulatory machinery strained to contain it. American Electric Power honored the congressional choice of an administrative forum over a judicial one. Utility Air Regulatory Group honored the numbers Congress enacted even though the agency’s program depended on changing them. Michigan honored the ordinary meaning of “appropriate” even though the agency preferred to weigh costs later. West Virginia honored the limits of a capacious phrase even though the agency’s plan addressed the central environmental challenge of the era. Read together, the eight decisions form a single sustained inquiry into what broad statutory language can carry, and the answer changes with the load. That is the doctrine factory’s product: not a rule for all seasons but a pricing mechanism, adjusting the exchange rate between legislative generality and executive ambition as the Court’s understanding of the constitutional stakes evolves.

This is the series thesis in its purest form: litigation is the stage where a statute’s breadth gets priced. The price is not set when Congress votes; it is set decades later, case by case, as courts decide what each capacious phrase can bear. Massachusetts v. EPA sits at the hinge of that pricing history. It confirmed the broadest possible reading of the act’s jurisdiction, bringing the defining environmental challenge of the century within the agency’s reach, while simultaneously insisting that the agency’s discretion be exercised in the statute’s own terms rather than the Administrator’s policy preferences. The decision thus did double work that later cases would pull apart. American Electric Power took the jurisdictional half and channeled it, displacing judge made law in favor of the administrative process the statute created. Utility Air Regulatory Group took the textual half and hardened it, refusing to let the agency edit the statute’s numbers even to save its own program. Michigan v. EPA took the reasoning half and deepened it, demanding that the economic consequences of regulation be faced at the threshold rather than deferred. West Virginia v. EPA took the delegation half and bounded it, holding that the broadest phrases cannot support the most transformative claims without clearer warrant from Congress.

The symmetry is worth stating plainly, because it is the namable claim this article exists to make. The Clean Air Act is the doctrine factory of American administrative law. Its breadth made Chevron possible, because only a statute this ambitious could present the interpretive questions deference was built to resolve. Its breadth made the major questions doctrine necessary, in the majority’s telling, because only a statute this ambitious could tempt an agency to reorder an industry on the strength of a phrase. The same drafting virtue, generality that endures across decades of scientific and economic change, reads as wisdom in 1984 and as danger in 2022. Nothing about the words changed. What changed was the judiciary’s theory of what broad words authorize, and that theory is itself a product of the stakes the statute’s success created. A law that governed smokestacks could be read generously without frightening anyone. A law that governs the climate cannot.

In June 2024, the Court overruled the 1984 deference framework itself, closing the circle that began with a Clean Air Act dispute forty years earlier. That development belongs to the Chevron half of this article’s story and is recounted there; what matters for the litigation history told here is the pattern it completes. Every generation of judges has found in this statute the occasion to restate the terms of the administrative state, because every generation has found the statute’s ambitions colliding with the limits of its text. The verdict, then, is not that any one of these decisions was right or wrong in isolation. It is that the act’s greatest strength, language broad enough to reach problems its drafters never imagined, guarantees a permanent docket. As long as the phrases stay broad and the stakes stay vast, the courts will keep pricing the breadth, one doctrine at a time, in each direction in turn.

The Construction Ledger

Decision	Year	Section construed	Holding in one sentence	Practical consequence for what the agency may do
Union Electric Co. v. EPA	1976	Section 110, state implementation plans	The agency must approve a state plan that satisfies the statutory criteria and may not reject it for technological or economic infeasibility.	Feasibility objections go to Congress, not to the Administrator; approved plans are judged against the text alone.
Chevron U.S.A. v. NRDC	1984	Nonattainment permit program, meaning of “stationary source”	Courts defer to reasonable agency readings of ambiguous statutory language under a two step framework.	The agency wins close interpretive fights, a latitude that endured until the framework was overruled in June 2024.
Whitman v. American Trucking Assns.	2001	Section 109(b)(1), standard setting	The agency may not consider implementation costs when setting national ambient air quality standards.	Health based standards are set on health alone, with cost entering only at the implementation stage.
Massachusetts v. EPA	2007	Sections 202(a)(1) and 302(g)	Greenhouse gases are air pollutants within the agency’s jurisdiction, and refusals to regulate must rest on statutory criteria rather than policy preference.	The agency may regulate climate pollutants but must ground every yes or no in the statute’s scientific inquiry.
American Electric Power Co. v. Connecticut	2011	The act as a whole, displacement	The statute displaces federal common law nuisance claims seeking judicial caps on power plant carbon dioxide.	Climate disputes go through the agency’s rulemaking process with judicial review, not through tort suits before judges.
Utility Air Regulatory Group v. EPA	2014	Prevention of significant deterioration and Title V thresholds	The agency may not rewrite the statute’s numerical permitting thresholds, though it may apply greenhouse gas controls to sources already in the program.	Statutory numbers are fixed by Congress; the agency keeps authority over climate pollutants from facilities it already permits.
Michigan v. EPA	2015	Section 112(n)(1)(A), “appropriate and necessary”	The agency must consider cost at the threshold decision to regulate power plant toxic pollutants.	Cost analysis moves to the front of every major rulemaking record rather than arriving after the commitment to act.
West Virginia v. EPA	2022	Section 111(d), existing source guidelines	The provision does not clearly authorize a generation shifting scheme, and transformative claims require clear congressional authorization.	Sector restructuring plans need explicit legislative warrant; source focused emission measures remain on firmer ground.

Study

Readers working through this case arc will get the most from it by reading the eight decisions in chronological order and asking the same question of each: where did the Court locate the line between what the statute says and what the agency may add? Union Electric, Chevron, and Whitman fix the statute’s character before climate enters the picture; the five climate decisions then test that character under rising pressure. A useful exercise is to take the construction ledger above as a map and, for each row, identify the sentence in the opinion where the Court states its interpretive rule, then compare how the dissent states the competing rule. The majority and dissent in Massachusetts v. EPA disagree about the size of the deference space; the majority and dissent in West Virginia v. EPA disagree about what generality authorizes. Naming the precise point of disagreement, rather than the outcome, is the skill this series is built to teach.

A second exercise is citation discipline. Each decision is authority for its holding and its reasoning, not for the policy outcome a reader wishes it supported. Massachusetts v. EPA is authority for jurisdiction and for the rule that refusals to act must be statutorily grounded; it is not authority for any particular emission standard. Practice stating each holding in one sentence, as the ledger does, and check the statement against the opinion’s own words before repeating it.

Readers who want a structured place to keep these notes can use the VaultBook legislation study notebook, which is built for tracking holdings, dissents, and statutory cross references across a case series like this one.

Frequently Asked Questions

Q: What did Massachusetts v. EPA decide about the Clean Air Act?

In April 2007, a 5 to 4 Supreme Court held that greenhouse gases fall within the Clean Air Act’s definition of “air pollutant” under section 202(a)(1), which covers any physical or chemical substance emitted into the ambient air. Justice Stevens, writing for the majority, rejected the Environmental Protection Agency’s claim that it lacked authority to regulate motor vehicle greenhouse gases and its refusal to decide the 1999 rulemaking petition on policy grounds. The Court also recognized that states receive “special solicitude” in standing analysis because of their quasi-sovereign interests, such as Massachusetts’s threatened coastline. Justice Scalia’s dissent would have deferred to the agency’s reading, while Chief Justice Roberts’s dissent argued the state lacked standing. Crucially, the Court did not order greenhouse gas regulation; it ordered the agency to answer the petition on statutory grounds, either by making an endangerment finding or explaining why the science could not support one, and that finding followed in December 2009.

Q: Are greenhouse gases air pollutants under the Clean Air Act?

Yes, under the reasoning of Massachusetts v. EPA, decided in April 2007. The Clean Air Act defines “air pollutant” at 42 U.S.C. 7602(g) as any air pollution agent or combination of agents, including any physical, chemical, biological, or radioactive substance emitted into or otherwise entering the ambient air. The majority found that definition capacious enough to include carbon dioxide, methane, nitrous oxide, and hydrofluorocarbons. Justice Scalia’s dissent argued the term should cover only substances that pollute in the traditional sense, not ubiquitous compounds such as carbon dioxide, but the majority’s textual reading prevailed. The holding directly concerned section 202(a)(1) motor vehicle authority; later the agency extended greenhouse gas regulation to other programs, including the 2009 endangerment finding for mobile sources and subsequent stationary source rules, each of which drew separate legal challenges decided in 2014 and 2022.

Q: How did the Clean Air Act produce the Chevron doctrine?

The Chevron doctrine was born inside a Clean Air Act dispute. In Chevron U.S.A. v. Natural Resources Defense Council, decided in 1984, the Court reviewed the Environmental Protection Agency’s 1981 “bubble” rule, which treated an entire industrial plant as a single stationary source so that new equipment could be offset by retiring old equipment elsewhere on the site. Justice Stevens, writing for a unanimous Court, upheld the agency and announced the two-step deference framework: courts ask first whether Congress spoke clearly, and if not, whether the agency’s construction is reasonable. That framework governed administrative law for four decades. In 2024, Loper Bright Enterprises v. Raimondo overruled Chevron and ended mandatory deference, directing courts to exercise independent judgment in determining statutory meaning. The 1984 bubble policy outcome itself was not relitigated; what fell was the deference regime the Clean Air Act case had created.

Q: What did West Virginia v. EPA hold about the Clean Air Act?

In June 2022, the Court held 6 to 3 that the Clean Power Plan’s “generation shifting” approach exceeded the Environmental Protection Agency’s authority under section 111(d). The 2015 rule had required existing coal and gas plants to meet emission limits premised on shifting electricity production to cleaner sources, a restructuring of the power sector the majority said Congress had not clearly authorized. Chief Justice Roberts applied the major questions doctrine, holding that transformative regulatory claims need clear congressional authorization. The decision did not invalidate section 111(d) itself, which remains available for source-specific standards. Justice Kagan’s dissent argued the statute’s broad language, “best system of emission reduction,” plainly covered generation shifting and warned that the ruling would constrain the agency’s ability to address climate change.

Q: What is the major questions doctrine in Clean Air Act cases?

The major questions doctrine holds that an agency claiming power over matters of vast economic and political significance must point to clear congressional authorization. Its roots run through MCI v. AT&T in 1994 and FDA v. Brown and Williamson in 2000, but the phrase and its modern force come from Clean Air Act cases: Utility Air Regulatory Group v. EPA in 2014, which rejected the agency’s Tailoring Rule, and West Virginia v. EPA in 2022, which rejected generation shifting under section 111(d). Defenders describe the doctrine as a safeguard against agencies rewriting statutes through creative interpretation. Justice Kagan’s 2022 dissent called it a judicial invention that lets courts override Congress’s delegation to expert agencies. After Loper Bright ended Chevron deference in 2024, the doctrine remains a separate constraint that applies even when ordinary statutory interpretation would leave room for agency judgment.

Q: Can citizens sue under the Clean Air Act?

Yes. Section 304 of the Clean Air Act, codified at 42 U.S.C. 7604, authorizes any person to sue a violator of an emission standard or limitation and to sue the Administrator for failure to perform a nondiscretionary duty, such as missing a statutory deadline. A plaintiff must give 60 days’ notice to the Administrator, the state, and the alleged violator before filing. The suit is barred if the agency or the state has already commenced and is diligently prosecuting a civil action. Prevailing parties may recover attorney’s fees. Courts have read limits into the provision: a citizen suit cannot be used to challenge a nationally applicable rule, which must instead be attacked through a section 307 petition for review in the D.C. Circuit within 60 days of the rule’s publication. Plaintiffs must still satisfy Article III standing, so a generalized concern about air quality cannot substitute for a concrete, particularized injury.

Q: Which court hears Clean Air Act challenges?

Section 307(b)(1) of the Clean Air Act centralizes review. Petitions challenging nationally applicable rules, such as national ambient air quality standards, new source performance standards, and emission guidelines, must be filed in the United States Court of Appeals for the District of Columbia Circuit within 60 days of the rule’s notice in the Federal Register. Actions that are locally or regionally applicable go to the regional circuit where the source sits. The 60-day deadline is strict, and section 307(b)(2) bars raising the validity of a rule later as a defense in an enforcement action, so regulated parties must challenge promptly or lose the argument. This channeling explains why the D.C. Circuit has decided most of the landmark Clean Air Act cases, from the lead and particulate disputes of the 1980s through the greenhouse gas fights decided in 2007, 2014, and 2022.

Q: Did Whitman v. American Trucking limit the Clean Air Act?

Whitman v. American Trucking Associations, decided unanimously in 2001, answered two questions with lasting force. First, section 109(b)(1) requires national ambient air quality standards at levels “requisite to protect the public health” with an adequate margin of safety, and the agency may not consider implementation costs when setting those standards; Justice Scalia’s opinion held the text unambiguously bars cost considerations. Second, the provision supplies an intelligible principle guiding agency discretion, so the statute does not unconstitutionally delegate legislative power. The decision limited industry’s cost-based attacks on the standards themselves while channeling cost analysis to the implementation stage, where states and the agency choose control measures. Its no-cost-at-setting rule was later balanced by Michigan v. EPA in 2015, which required cost consideration when the agency decides under section 112 that regulating power plants is “appropriate and necessary.”

Q: Did Massachusetts v. EPA give states special standing in climate litigation?

The 2007 decision is the leading statement of state standing in environmental cases. The majority held that Massachusetts, suing in its quasi-sovereign capacity to protect its coastline and territory, deserved “special solicitude” in the standing analysis, combining its proprietary interest in state-owned coastal land with Congress’s procedural grant of a right to challenge agency action under section 307. Chief Justice Roberts’s dissent objected that the majority had relaxed Article III’s injury, causation, and redressability requirements for a favored class of litigants. Lower courts have applied the decision cautiously: it supports state challenges to federal regulatory inaction, but private plaintiffs still face ordinary standing rules, and in American Electric Power v. Connecticut in 2011 the Court split 4 to 4 on whether states could bring federal common law climate claims at all.

Q: What happened to the “bubble concept” after Chevron was overruled?

The bubble was never a statute; it was the Environmental Protection Agency’s 1981 policy of treating all emission points within one industrial plant as a single stationary source, so that a modification adding pollution could be offset by reductions elsewhere under the same roof. That policy supplied the facts for Chevron U.S.A. v. Natural Resources Defense Council in 1984, and the deference doctrine born there governed interpretation of the bubble provision for decades. When Loper Bright Enterprises v. Raimondo overruled Chevron in 2024, the bubble policy itself was not struck down; instead, the standard of review changed. Whether “stationary source” means a single device or an entire plant is decided by a court’s independent reading of the Clean Air Act’s text and structure, without deference to the agency’s preference.

Q: What does “requisite to protect public health” mean after Whitman?

Section 109(b)(1) directs the Environmental Protection Agency to set national ambient air quality standards at levels “requisite to protect the public health,” allowing an adequate margin of safety. In Whitman v. American Trucking Associations in 2001, the Court read “requisite” to mean sufficient but not more than necessary, and held unanimously that the provision bars cost considerations at the standard-setting stage. The phrase identifies a health-based ceiling and floor: standards must be protective, including a margin of safety for sensitive groups such as children and the elderly, but the agency may not tighten them beyond what health protection requires. Costs enter later, when states design implementation plans and choose control technologies under sections 110 and 111, a division the Court preserved by requiring cost analysis at the regulatory stage in Michigan v. EPA in 2015.

Q: How did American Electric Power v. Connecticut displace federal common law climate claims?

In 2011, the Court held 8 to 0 that the Clean Air Act displaces federal common law public nuisance claims seeking court-ordered limits on greenhouse gas emissions from power plants. Justice Ginsburg’s opinion reasoned that when Congress has delegated the question to an expert agency and the agency is regulating the subject, judge-made federal common law has no room to operate; courts should not set emission standards through nuisance litigation when the statute assigns that job to the Environmental Protection Agency. The displacement test asks whether the statute speaks directly to the question. On the separate question of whether the state plaintiffs had standing, the Court split 4 to 4, leaving the lower court’s affirmative ruling intact without precedential value.

Q: What did UARG v. EPA do to the Tailoring Rule?

In Utility Air Regulatory Group v. EPA, decided in 2014, the Court rejected the agency’s attempt to rewrite the Clean Air Act’s numerical permitting thresholds. After the 2007 Massachusetts decision and the 2009 endangerment finding made greenhouse gases regulated pollutants, the statutory thresholds of 100 and 250 tons per year would have swept millions of small sources into the prevention of significant deterioration program. The agency’s 2010 Tailoring Rule raised those thresholds by orders of magnitude, but the 5 to 4 majority held that an agency may not revise clear statutory terms to fit its policy preferences. The Court did uphold one part of the program: sources already subject to permitting for conventional pollutants could be required to install best available control technology for greenhouse gases, the so-called “anyway sources” holding.

Q: What did Michigan v. EPA say about cost in Clean Air Act rulemaking?

In 2015, the Court held 5 to 4 that the Environmental Protection Agency acted unreasonably when it decided that regulating mercury and other hazardous emissions from power plants was “appropriate and necessary” under section 112 without considering cost. Justice Scalia’s majority opinion treated cost as a centrally relevant factor that the word “appropriate” naturally includes, calling it irrational to impose billions in compliance costs for a few dollars in quantified benefits without weighing them. Justice Kagan’s dissent argued the agency had considered costs extensively when setting the actual emission standards, even if not at the threshold finding. The decision stands alongside Whitman v. American Trucking in 2001: costs are barred at the national ambient air quality standard-setting stage but required when the statute’s “appropriate and necessary” language opens the door.

Q: What is the endangerment finding sequence under the Clean Air Act?

The sequence began with a 1999 rulemaking petition asking the Environmental Protection Agency to regulate motor vehicle greenhouse gases, which the agency denied in 2003 on the ground that it lacked authority and that regulation would be unwise. Massachusetts v. EPA in 2007 reversed the denial, holding greenhouse gases are air pollutants and requiring a statutorily grounded answer. The agency responded in December 2009 with the endangerment finding, concluding that greenhouse gases from motor vehicles endanger public health and welfare, which triggered a mandatory duty to set emission standards. Joint vehicle greenhouse gas standards with the Department of Transportation’s fuel economy rules followed in 2010. The finding is the legal linchpin for later mobile source rules, though later administrations revised or rescinded standards built on it in 2020, 2021, and 2024.

Q: What notice and venue rules govern Clean Air Act citizen suits?

Section 304 requires a would-be plaintiff to give 60 days’ notice to the Environmental Protection Agency, the state, and the alleged violator before filing suit, giving authorities a chance to act first. The suit must then be brought in the district court for the district where the violation occurred, not as a free-floating challenge to national policy. If the agency or state has commenced and is diligently prosecuting its own civil action, the citizen suit is barred. Attempts to use section 304 to attack a nationally applicable rule are misdirected: such challenges belong in the D.C. Circuit under section 307 within 60 days of the rule’s publication, and courts have dismissed citizen suits that were really collateral attacks on rulemaking. The structure channels policy disputes to the specialized appellate forum and reserves district courts for enforcement against specific sources.

Q: What makes D.C. Circuit review of Clean Air Act rules distinctive?

Beyond exclusive jurisdiction over nationally applicable rules under section 307, the D.C. Circuit has developed a dense body of precedent on how closely courts should scrutinize the Environmental Protection Agency’s technical records, from the lead and particulate matter battles of the 1980s through the greenhouse gas cases decided in 2007, 2014, and 2022. Its review is confined to the administrative record and applies the arbitrary and capricious standard, but the court’s repeated encounters with the same statutory provisions have produced detailed doctrines on rulemaking procedure, cost-benefit analysis, and statutory deadlines. The Supreme Court takes comparatively few of these cases, so D.C. Circuit panel decisions often supply the working law for industry and regulators for years at a time, making the court’s composition and precedents a practical factor in regulatory strategy.

Q: Did Whitman v. American Trucking settle nondelegation challenges to the Clean Air Act?

Largely yes. In 2001, the D.C. Circuit had suggested that section 109(b)(1)’s instruction to set standards “requisite to protect the public health” might delegate legislative power without an intelligible principle. The Supreme Court unanimously disagreed: Justice Scalia’s opinion held that the provision’s health-based directive, combined with the adequate margin of safety requirement, gave the agency a clear, determinate standard to apply. The Court reaffirmed the intelligible principle test from J.W. Hampton in 1928 and declined to revive a stricter nondelegation doctrine. Industry and advocacy groups have since challenged other Clean Air Act provisions on delegation grounds, but Whitman has remained the controlling precedent, and the major questions decisions of 2022 and 2024 constrained agency action on statutory interpretation grounds rather than through nondelegation.

Q: What does “generation shifting” mean in Clean Air Act cases?

The term comes from the Environmental Protection Agency’s 2015 Clean Power Plan, which set section 111(d) emission guidelines for existing power plants using three “building blocks”: improving heat rates at coal plants, shifting generation from coal to existing natural gas plants, and shifting generation to new renewable sources. Chief Justice Roberts’s majority opinion in West Virginia v. EPA in 2022 used “generation shifting” to describe the second and third blocks, treating them as a system-wide restructuring of the electricity sector rather than a pollution control measure applied at individual sources. The distinction mattered because the Court held that such sector-wide restructuring required clear congressional authorization, which section 111(d)’s reference to the “best system of emission reduction” did not supply.

Q: How should these cases be cited in legal writing?

Under the Bluebook, the leading cases are cited as Massachusetts v. EPA, 549 U.S. 497 (2007); Chevron U.S.A., Inc. v. Nat. Res. Def. Council, Inc., 467 U.S. 837 (1984); Whitman v. Am. Trucking Ass’ns, 531 U.S. 457 (2001); Am. Elec. Power Co. v. Connecticut, 564 U.S. 410 (2011); Util. Air Regulatory Grp. v. EPA, 573 U.S. 302 (2014); Michigan v. EPA, 576 U.S. 743 (2015); West Virginia v. EPA, 597 U.S. 697 (2022); and Loper Bright Enters. v. Raimondo, 144 S. Ct. 2244 (2024). Short forms such as Massachusetts, West Virginia, and Loper Bright are standard after a full citation. When quoting statutory language, cite the United States Code section, for example 42 U.S.C. 7409(b)(1) for the national ambient air quality standard provision, alongside the Clean Air Act section number familiar to practitioners.