From Statute to Smokestack: The Machinery Between Authorization and Obligation

Consider the sentence Congress placed at section 111(b)(1)(B) of the Clean Air Act, 42 U.S.C. 7411(b)(1)(B): the Administrator of the Environmental Protection Agency “shall” establish standards of performance for new stationary sources. It reads like a command with immediate force. It carries none. No plant manager in 2011 could read that sentence and learn, in tons per year or in pounds per million Btu, what the law demanded of a particular boiler on a particular date. The sentence assigns a task. Converting the task into an obligation requires machinery the statute prescribes in unusual detail, and every link in that machinery can stall, thin out, or fail.

The chain runs through Federal Register notices, a public docket, an opportunity for oral presentation, written comments that can number in the hundreds of thousands, a written response to the serious criticisms, a final rule with an administrative record a court can test, and then an entirely second machine: each state writes an implementation plan translating national decisions into enforceable local limits, permits carry those limits onto individual facilities, inspections test compliance, and enforcement backs the whole structure. A missed statutory deadline, a record that cannot support the final choice, a disapproved state plan, or a permit issued without a required limit each opens the same gap this series keeps returning to: the distance between what a statute authorizes and what actually gets done. The gap is not an abstraction. It is measurable in years of delay and in tons of pollution released while procedure runs its course.

The Clean Air Act is the sharpest illustration of the series thesis because Congress refused to leave the bridging machinery to the spare default of the Administrative Procedure Act. It wrote a bespoke rulemaking procedure into the statute itself at section 307(d), 42 U.S.C. 7607(d), with a docket, an oral hearing opportunity, and a written answer to every major criticism. It built a separate state-plan approval machine at sections 110 and 179, 42 U.S.C. 7410 and 7509, with conditional approvals, disapprovals, sanctions clocks, and a federal backstop plan. The statute took its modern shape in the 1990 amendments, which layered enforceable deadlines, planning duties, and permit programs onto a structure first erected in 1970. To see how a sentence becomes a stack-level obligation, follow the two mechanisms in the order the law walks them.

The mercury story makes the gap concrete. The 1990 amendments directed EPA to study hazardous air pollutants from power plants and to regulate them if the findings warranted it. EPA listed power plants for regulation in 2000, issued a mercury rule in 2005, saw that rule vacated by the D.C. Circuit in 2008 in New Jersey v. EPA, and returned to the docket to build a new rule from the record up. The final standards arrived in December 2011, more than two decades after Congress first pointed at the problem. Every one of those years was procedure: study, listing, proposal, record, vacatur, re-proposal, response, final action. The statute authorized control in 1990; the machinery delivered an enforceable obligation in 2011; the difference between those dates is the subject of this article.

A second illustration comes from the same year. The Act’s good neighbor provision, section 110(a)(2)(D)(i)(I), bars one state’s pollution from significantly contributing to another state’s air quality problems, but that sentence cannot tell any state how much to cut. EPA’s 2005 interstate transport rule was vacated and remanded in 2008; the 2011 replacement drew petitions for review within weeks, a stay before year’s end, and in 2014 the Supreme Court upheld it in EPA v. EME Homer City Generation. The sequence from proposal to stay to Supreme Court decision is the rulemaking chain and the courtroom link operating exactly as designed, slowly. Both examples recur below as the mechanisms that produced them are explained.

The two machines fail in different ways, which is why the article treats them separately. The rulemaking chain fails on paper: a thin record, an unanswered criticism, a statutory reading the text cannot bear, and the court vacates the rule before it ever reaches a state. The plan approval machinery fails in time: a missed submission, a disapproval left uncorrected, clocks running while pollution continues, and the federal backstop arriving years after the harm. One failure is about reasons, the other about deadlines. The Clean Air Act arms both with procedure because Congress learned, across the 1970s, that an agency can be wrong in either dimension.

December 2011 caught both machines mid-cycle. The mercury standards showed the rulemaking chain at its most laborious, nearly a million comments answered on the record. The transport rule showed the plan machinery’s interstate variant, federal plans issued for states whose submissions fell short, then stayed before they operated. A reader in 2011 could watch the entire system, proposal to courtroom to statehouse, in a single season’s Federal Register. The dated later developments noted below, court decisions from 2014 through 2024, show the same machinery still processing the same kinds of problems, which is the strongest evidence that the procedure, not the posture of any season, is the durable subject.

The cast of the machinery is smaller than the stakes suggest. Inside EPA, the Office of Air and Radiation drafts the rules, the regional offices review most state plans, and the Office of General Counsel defends the record in court. The states supply the air agencies that write the plans, issue the permits, and send the inspectors. The D.C. Circuit hears most challenges to national rules, and district courts hear the deadline suits that force missed duties onto a schedule. Industry groups, environmental organizations, and state attorneys general supply the comments and the lawsuits that test each link. No single actor controls the chain; each holds a veto at a different station, which is why the procedure matters more than any participant’s intentions.

That diffusion of veto points is the series thesis in institutional form. Statutes authorize outcomes; they do not produce them. Between authorization and outcome stands a sequence of required acts, each performed by a different actor under a different deadline, each reviewable by a court, and each capable of failing independently. A sound rule dies in a disapproved plan. A sound plan dies in an unissued permit. The Clean Air Act’s distinction is that Congress mapped these failure points in advance and built a named procedure for each: 307(d) for the rule, sections 110 and 179 for the plan, Title V for the permit, sections 113 and 304 for enforcement. The rest of this article walks the first two in detail, because they are where most of the gap opens.

This article walks two mechanisms. The first is the rulemaking chain: how EPA turns a statutory command into a final rule that survives judicial review, under procedures stricter than the APA default. The second is the plan approval machinery: how a final rule becomes enforceable limits inside each state, through state implementation plans, federal backstop plans, and the sanctions that punish delay. Together they carry the sentence from section 111 or section 109 to the permit posted on the facility wall, and each handoff is a place where authorization either becomes action or does not.

The Rulemaking Chain: From Statutory Command to Final Rule

The chain, in order

Most Clean Air Act rulemakings begin before the proposal, with an advance notice of proposed rulemaking. The advance notice is optional and carries no legal force; it is the agency asking for data before it commits to a direction. In July 2008, after the Supreme Court held in Massachusetts v. EPA (2007) that greenhouse gases qualify as air pollutants under the Act, EPA published an advance notice soliciting information on how the statute’s machinery might address them. The document was a long Federal Register notice full of questions. It bought the agency information and time, and it showed how the chain adapts to pollutants Congress never named.

The proposal, called a notice of proposed rulemaking, is where legal obligation starts to take shape. EPA publishes the proposed regulatory text in the Federal Register with a preamble explaining the factual basis, the statutory authority, and the choices the agency is weighing, and it poses specific questions to commenters. The comment period usually runs sixty to ninety days and is often extended for complex rules. During that window anyone may submit data, studies, economic critiques, or legal objections; trade associations, state agencies, public health groups, and individual citizens all use the same docket. The proposal is a draft the public is invited to break, and the agency expects it to be tested.

The preamble does double duty that the APA never imagined. It must persuade the public during comment and persuade the court after, which is why EPA preambles for major rules run hundreds of pages with technical support documents behind them. The preamble states the statutory authority for each provision, summarizes the factual basis with docket citations, explains why the agency chose its approach over the alternatives commenters proposed, and identifies the data it relied on. A preamble that merely announces the rule’s content satisfies the APA; a preamble that survives 307(d) review reads like a brief written before the lawsuit exists, because in substance that is what it is.

Section 307(d)(5) adds an opportunity the APA does not require: EPA must offer oral presentation of data, views, and information on the proposal. The hearing is not a trial; there is no cross-examination and no judge. Its value is access. A small-town mayor, a plant worker, or a tribal representative who will never hire a Washington law firm can put testimony into the same record that carries the industry studies, and the statute requires the transcript to join the docket. The requirement dates to the 1977 amendments, when Congress concluded that paper comments alone filtered out voices the pollution rules would reach.

The response to comments is where rulemakings are won or lost in court. Section 307(d)(6) commands a statement of basis and purpose that answers each major comment, criticism, and item of new data submitted during the comment period. For the 2011 mercury and air toxics rule, nearly one million comments arrived, and the response document ran thousands of pages, grouping repetitive submissions and answering each distinct criticism with record citations. An agency that waves past a serious technical objection does not merely look careless; it builds the record on which a court will later vacate the rule. The drafters of section 307(d) understood that the final rule’s real author is often the commenter who found the flaw the agency missed.

The final rule publishes in the Federal Register with three components: a preamble explaining what the agency decided and why, the binding regulatory text codified in the Code of Federal Regulations, and a summary of the response to comments. The effective date usually lands thirty to sixty days after publication; major rules face a sixty-day delay under the Congressional Review Act, which also gives Congress a window to disapprove the rule by joint resolution. From the effective date the rule binds, and permits, plans, and enforcement actions cite it as authority. But binding is not the same as surviving: petitions for review routinely follow within weeks, and the court reads the record the agency built.

The administrative record is the thread connecting every step. Section 307(d)(4) requires EPA to establish a docket containing the data, information, and documents the agency considered, including drafts exchanged with the White House Office of Information and Regulatory Affairs during interagency review. Section 307(d)(7)(A) then defines the record for judicial review: the promulgated rule, the proposal, the docket materials, the hearing transcript, and the response document. Courts test the final rule against that record and that record only; an explanation invented after promulgation, sometimes called post hoc rationalization, cannot save a rule the record does not support. The practical discipline is blunt: if a fact mattered to the decision, it belongs in the docket before the rule issues, because the courtroom has no door for late evidence.

Two friction points slow the chain from outside the statute’s text. First, nationally important regulatory actions undergo review by the Office of Information and Regulatory Affairs under Executive Order 12866, a process that can add months of interagency negotiation over costs, benefits, and alternatives. Second, the Act’s own deadlines routinely outrun the agency’s capacity; when a nondiscretionary duty passes unmet, section 304(a)(2) lets any person sue to compel performance, and the resulting consent decrees set court-enforceable rulemaking schedules. Both forces are institutional, not partisan: review tempers haste, deadline suits punish drift, and the chain absorbs the delay either way.

The comment period as a contest

The comment period looks democratic and functions adversarial. The docket for a major rule receives submissions from regulated industries with engineering studies attached, from environmental groups with epidemiological data, from state agencies with implementation concerns, from small businesses describing compliance costs, and from organized campaigns generating hundreds of thousands of nearly identical letters. EPA must log all of it. The form letters count as participation, but they rarely move the rule; what moves the rule is the submission that attaches data, cites the record, and shows the proposal’s numbers do not add up.

Ex parte communication gets special treatment under 307(d). Contacts between EPA and outside parties during the comment period that bear on the rulemaking must be memorialized in the docket, so that no argument reaches the decisionmaker without reaching the other commenters too. The rule against secret data is absolute in one direction: the agency may not rely on material that never entered the public docket, because reliance on hidden material denies commenters the chance to rebut it and denies the court a record to review. A study the agency commissioned, a dataset it ran, a model it tweaked after the proposal, all of it must be docketed in time for response, or the final rule cannot rest on it.

The strategic consequence is that sophisticated commenters write for two audiences at once: the agency staff who might fix the proposal, and the judges who will later read the docket. A comment that says the cost estimate is wrong accomplishes little; a comment that says the cost estimate omits retrofit downtime shown in the agency’s own docketed survey forces a written answer and builds the record for vacatur if the answer never comes. Environmental groups learned this discipline in the 1970s; industry learned it in the 1980s; by 2011 the docket for a major rule read like a trial record assembled before any trial, because everyone involved knew the trial would come.

Timing matters at the margins. Comments filed after the deadline may be considered at the agency’s discretion but carry less weight, and a commenter who discovers a flaw only after the period closes faces the forfeiture rule the companion section covers. EPA sometimes reopens a comment period when it adds important new material to the docket, a practice courts encourage because it cures the secret-data problem, and sometimes extends the period under pressure from Congress or from the sheer volume of submissions. The 2011 mercury rule’s comment period drew extensions and still closed with the docket full; the agency’s choice to answer the distinct criticisms rather than the mass signatures is the reason the rule’s record survived as long as it did.

Why does the Clean Air Act add procedures on top of the APA?

Congress distrusted the ordinary notice and comment default for pollution rules that reach hundreds of facilities at once. Section 307(d) layers on a docket, an oral hearing opportunity, and a written answer to every major criticism, so that the final rule carries a record a court can test against what participants actually said.

The longer answer is scale and distrust. By 1977 Congress had watched a decade of pollution rules that reached thousands of facilities at once, built on technical records no generalist judge could reconstruct, and defended with explanations written after the fact. The APA’s default, notice plus a concise general statement of basis and purpose, was designed for rules of modest scope. Section 307(d) replaces the concise statement with a full written answer to every major criticism and replaces the agency’s private file with a public docket, because a rule that can idle a power plant should carry its reasons on its face.

The distrust had its roots in the case law of the early 1970s. In the early 1970s the D.C. Circuit, reviewing the first generation of Clean Air Act rules, began demanding more than the APA minimum: in Portland Cement Association v. Ruckelshaus (1973) the court required EPA to disclose the technical basis for its new source standards so that commenters could test it. The lesson stuck. When Congress rewrote the Act’s procedures in the 1977 amendments, it codified what the judges had been improvising: a public docket, an oral hearing opportunity, and a written response to the serious criticisms. Section 307(d) is thus a statute written by a legislature that had watched its agency lose cases for thin records and decided to make the thick record mandatory.

The provision’s coverage is deliberate. Section 307(d)(1) lists the rulemakings it governs: national ambient air quality standards, new source performance standards, hazardous air pollutant standards, and the other national-scale actions the Act assigns to EPA, and it displaces the APA’s informal rulemaking requirements for those actions except as the subsection expressly incorporates them. Congress did not impose the heavier procedure on every minor amendment; it reserved the full machinery for the rules with national reach. The health-based limits at the center of this machinery are explained in the ambient standards article, and the 307(d) procedure is what turns those limits from a Federal Register notice into obligations a court will enforce.

Each added procedure answers a specific failure mode. The docket answers the secret record: nothing the agency relied on may hide from the commenters who must critique it or from the judges who will review it. The oral hearing answers the access gap: written comment favors organized interests with technical staff, and the statute keeps a channel for everyone else. The response requirement answers the pretense of consideration: an agency must show its work against the strongest objections, not merely announce that it considered them. Together they make the rule’s defense contemporaneous with the rule rather than an exercise in hindsight.

The cost is time and volume, and Congress accepted the trade openly. A 307(d) rulemaking with a million comments consumes years of staff work; the response document alone can exceed the rule’s preamble by an order of magnitude. The agency sometimes narrows the pain by grouping repetitive comments and answering the distinct technical claims once, a practice courts accept so long as no serious criticism goes unanswered. The alternative, a thin record and a quick rule, buys speed at the price the next section describes: vacatur.

One boundary deserves emphasis because it is easy to misread. Section 307(d) supplements the statute’s own commands; it does not excuse the agency from them. A perfect procedure cannot save a rule the statute does not authorize, and a court that finds the Act forbids the agency’s reading will vacate the rule no matter how immaculate the docket. The chain’s lesson is narrower and still demanding: where the statute permits a choice, only a choice explained against the record survives.

The record and the courtroom

Judicial review under section 307(d)(9) applies the familiar arbitrary and capricious standard, but the record rule gives it teeth. The court asks two questions: did the evidence in the docket support the agency’s choice, and did the agency answer the major criticisms leveled during comment. The 2008 vacatur of the Clean Air Mercury Rule in New Jersey v. EPA turned on the second kind of failure compounded by the first: the D.C. Circuit held EPA had used an improper delisting procedure to avoid the hazardous pollutant standards the statute required. The agency returned to the docket and built the 2011 replacement rule on a new record, which is why the mercury timeline runs through a courtroom between proposal and final action.

The 2015 decision in Michigan v. EPA shows the standard operating at the final rule stage. The Supreme Court held that EPA had acted unreasonably in deciding to regulate power plants’ hazardous emissions without considering cost at the threshold listing decision, remanding the 2011 mercury standards without vacating them. The rule stayed in effect while the agency addressed the defect on remand, a common judicial compromise that keeps pollution controls running while the record is repaired. The episode illustrates the chain’s central discipline: the defect was not a missing comment period or a thin docket but a reasoning gap the record could not fill, and the remedy was more procedure, not less.

For the 2011 reader, one background assumption about courtroom review deserves dating, because it later changed. In 2011 courts applied Chevron deference, born in a 1984 Clean Air Act case about the meaning of “stationary source”: where the statute was ambiguous, judges accepted EPA’s reasonable reading. In 2024 the Supreme Court overruled Chevron in Loper Bright Enterprises v. Raimondo, directing courts to exercise independent judgment on questions of statutory meaning. The technical record still decides whether the agency’s facts hold up, but the agency’s reading of the Act no longer borrows deference from the judiciary. A 2011 rulemaking record built to earn Chevron’s shelter faces, after the 2024 overruling, a court that reads the statute for itself.

A second later development belongs in the same dated frame. In 2022 the Supreme Court decided West Virginia v. EPA, holding that section 111(d) did not authorize the generation-shifting approach of the 2015 Clean Power Plan and announcing that transformative agency actions need clear congressional authorization. The decision did not alter the 307(d) procedure; it narrowed what the procedure may produce. Rulemakings since have been drafted with that boundary in view, another reminder that the chain transmits only as much authority as the statute carries.

Two recurring errors about challenges need brief correction. First, venue is not a free choice: section 307(b)(1) sends petitions for review of nationally applicable final actions to the D.C. Circuit and regionally applicable actions to the regional circuit, with a sixty-day filing window from Federal Register notice. A challenger cannot shop for a sympathetic panel; the statute assigns the forum. Second, guidance is not a shortcut around the chain: documents that EPA labels guidance but that impose binding obligations are treated by courts as rules issued without required procedure and set aside. The machinery has no express lane.

The sixty-day filing window shapes litigation strategy as much as venue does. Potential challengers must decide quickly whether to sue, and trade associations often file protective petitions while negotiations with the agency continue, because a missed deadline forfeits review entirely. Courts, for their part, routinely sever the challenged provisions from the rest of the rule, vacating only the unlawful portion where the remainder can stand alone. Severability is another gap-management device: it lets the lawful parts of the machinery keep operating while the defective parts go back through the chain.

One foreshadowing sentence, one mechanism: a companion section of this series takes up the forfeiture rule that punishes silence during the comment period; at this stage, the chain’s lesson is that an objection never raised with reasonable specificity while the docket was open is an objection the court will generally refuse to hear.

The court’s role in technical disputes is review, not replication. Judges do not rerun EPA’s models or second-guess which study the agency credited; they ask whether the agency examined the relevant data, considered the important aspects of the problem, and drew a rational connection between the facts found and the choice made. That posture gives the docket its decisive weight: an agency that documented its reasoning step by step will survive disagreement over the science, while an agency that announced a conclusion and gestured at the record will not. The hard-look standard, as judges call it, is less about being right than about showing the work, which returns every courtroom fight to the quality of the rulemaking chain.

Stays, remands, and second chances

A final rule does not always take effect on its effective date, because challengers routinely ask courts to stay the rule while the case proceeds. The court weighs the challenger’s likelihood of success, the irreparable harm of letting the rule operate, the harm a stay inflicts on others, and the public interest. The Cross-State Air Pollution Rule won its stay in December 2011, five months after issuance, which meant the interstate transport program it created never operated while the litigation ran. A stay is not a decision on the merits, but for a rule whose compliance investments are irreversible, the stay can decide the practical outcome before the merits are reached.

Courts have developed compromises for rules that fail review but whose sudden disappearance would cause harm. In 2008 the D.C. Circuit vacated the Clean Air Interstate Rule in North Carolina v. EPA, then on rehearing left the vacated rule in place until EPA issued a replacement, reasoning that yanking the program would sacrifice air quality gains while the agency rebuilt the record. Remand without vacatur, used again for the mercury standards in Michigan v. EPA in 2015, keeps the rule operating while the agency repairs the identified defect. Both devices show the judiciary managing the gap directly: the rule was unlawful as written, yet the court declined to restore the pre-rule world while the machinery ran again.

The statute also provides a second chance inside the agency before the courtroom. Section 307(d)(7)(B) generally limits judicial review to objections raised with reasonable specificity during the comment period, but it excuses objections whose grounds arose after the period closed or that could not practicably have been raised earlier, and EPA’s reconsideration practice lets petitioners present new information for agency reconsideration before or alongside judicial review. The provision rewards diligence rather than perfection: a commenter who raised the issue early preserves it; a commenter blindsided by the final rule’s new reasoning gets a door back in. Either way, the docket remains the arena, and the record remains the evidence.

The Plan Approval Machinery: From National Rule to State Obligation

The state’s job under section 110

The Act’s central structural bet is cooperative federalism: EPA sets the national requirements, and each state writes the plan that delivers them. Section 110(a)(1) gives states three years after EPA issues or revises a national ambient air quality standard to submit an infrastructure plan demonstrating the legal authority, monitoring network, enforcement resources, and interstate transport provisions the standard demands. Section 110(a)(2) lists the required elements: enforceable emission limits and control measures, air quality monitoring, an enforcement program, adequate personnel and funding, and the good neighbor provisions barring one state’s pollution from fouling another’s air. For areas that violate a standard, Part D adds attainment plans with reasonably available control measures, reasonable further progress milestones, and a modeled demonstration that the area will reach attainment by its deadline.

The infrastructure elements reach beyond emission limits into the state’s capacity to run the program. Section 110(a)(2) requires monitoring networks that measure actual air quality, legal authority to enforce the plan’s limits, personnel and funding adequate to the task, a prevention-of-significant-deterioration program for clean areas, and consultation with local governments. EPA reviews these elements on paper but tests them in practice: a state with elegant limits and no inspectors has submitted a plan that cannot function, and the completeness and adequacy findings are the tools for saying so. The infrastructure plan is thus less a pollution budget than a demonstration that the state owns a working regulatory apparatus.

The state plan is not a promise; it is a set of federally enforceable limits. Once EPA approves a state implementation plan, its emission limits, compliance schedules, and control measures become federal law alongside state law, enforceable by the state, by EPA, and by citizens. A facility’s obligations therefore arrive in two layers: the national rule sets the direction, and the approved plan writes the numbers that the facility’s permit will carry. This is why the plan approval step, bureaucratic as it sounds, determines whether the national rule changes anything at any particular smokestack.

Timing discipline comes from section 110(k). EPA must first determine whether a state submission is complete, a check on paperwork rather than substance, and must act on a complete submission within twelve months: approve it in full, approve it in part, conditionally approve it, or disapprove it in whole or in part. If EPA finds no submission at all, or finds an existing plan substantially inadequate to meet the Act’s requirements, separate findings start the clocks the next sections describe. The twelve-month action duty is nondiscretionary, which makes it a frequent target of the deadline suits described earlier.

The substantial inadequacy finding deserves its own mention because it shows the machinery correcting course years after approval. In 2015 EPA issued a SIP call finding that startup, shutdown, and malfunction provisions in the plans of thirty-six states were substantially inadequate, because they shielded excess emissions during equipment transitions from enforcement. The affected states had to revise their plans to remove the exemptions. A plan approved in one decade can thus be reopened in the next when its provisions prove incompatible with the Act’s enforcement design, and the call mechanism keeps the state-plan layer honest over time.

The good neighbor problem, four times over

No problem has cycled through the plan approval machinery more often than interstate transport. The good neighbor provision forbids upwind states from sending pollution that prevents downwind states from attaining the standards, but the sentence supplies no method for dividing the required cuts among the upwind states. Each attempt to supply the method has run the full chain: proposal, record, comments, final rule, state plan submissions, disapprovals, federal plans, petitions for review, and a court decision sending the agency back to start.

The first cycle began with the NOx SIP Call of 1998, which found that nitrogen oxide emissions from twenty-two states and the District of Columbia significantly contributed to ozone problems downwind and required those jurisdictions to submit revised plans with power plant budgets. States that missed the submission duties received federal implementation plans around 2000. The second cycle was the Clean Air Interstate Rule of 2005, a cap-and-trade approach to sulfur dioxide and nitrogen oxides that the D.C. Circuit vacated in 2008 and then left in place pending replacement. The third cycle was the 2011 Cross-State Air Pollution Rule, issued with federal plans for states lacking adequate good neighbor submissions, stayed that December, and upheld by the Supreme Court in 2014.

The fourth cycle arrived in 2023, when EPA issued a Good Neighbor Plan, itself a federal implementation plan covering more than twenty states for the 2015 ozone standard, after disapproving or finding inadequate the states’ own submissions. In June 2024 the Supreme Court stayed the plan in Ohio v. EPA while challenges proceeded, questioning whether EPA had adequately considered the effect of partial stays on the plan’s design. Four cycles, one statutory sentence, twenty-six years. The machinery does not resolve the good neighbor problem so much as process it, each pass producing a more elaborate record and a narrower set of legally survivable choices.

The lesson for the plan approval machinery is that disapproval and the federal backstop are not aberrations; for the hardest problems they are the normal path. States submit, EPA disapproves, the clocks run, federal plans issue, courts review, and the next cycle begins with a thicker docket. The gap between authorization and action narrows with each pass but never quite closes, because each pass teaches the next round of commenters exactly where the last record failed.

EPA’s four answers

Full approval is the quiet outcome and the most common: EPA finds the submission meets every applicable duty of the Act, publishes the approval, and the plan’s limits become federally enforceable. Nothing about approval is ceremonial; EPA’s review checks each element of section 110(a)(2) and the applicable Part D duties, and states routinely negotiate revisions with the regional office before the formal submission to avoid a disapproval.

Conditional approval handles the plan that is almost adequate. EPA approves on the condition that the state adopt specific corrective measures within one year; if the state delivers and EPA approves the correction, the conditional approval converts to a full approval, and if the state misses the commitment, the conditional approval converts to a disapproval by operation of law. The device lets imperfect plans take effect without pretending they are complete, and the automatic conversion keeps the state honest about the deadline.

Partial approval with partial disapproval splits the submission along severable lines: the adequate portions take effect as federal law while the deficient portions trigger the consequences of disapproval. EPA uses this where a plan’s control measures for one pollutant satisfy the Act but its interstate transport provisions do not, a common pattern in good neighbor submissions. The split decision reflects the plan’s actual content rather than forcing an all-or-nothing verdict, and each portion then follows its own procedural track.

Full disapproval is the outcome that sets the machinery’s penalties in motion. EPA publishes the disapproval with its reasons, the state’s submission fails to become federal law, and two clocks begin to run: an eighteen-month sanctions clock and a twenty-four-month clock for EPA to impose a federal implementation plan. Disapproval is not the end of the state’s role; the state may submit a corrected plan at any time, and EPA’s approval of the correction stops both clocks. But until the correction arrives, the statute’s backstops engage on their fixed schedule.

Approval is not a one-time event. States revise their plans as standards tighten, as courts vacate old approvals, and as their own laws change; each revision runs the same 110(k) review. A plan approved for the 1997 ozone standard may need a fresh submission for the 2008 revision and another for 2015, and each submission restarts the twelve-month clock. The plan layer is therefore a moving surface, and EPA’s regional offices spend more time reviewing revisions than reviewing first-time submissions.

Behind the four formal answers sits an informal process the statute never names. State air agencies circulate draft submissions to EPA’s regional office months before the formal filing, and regional staff flag the provisions likely to draw disapproval, suggest curative language, and negotiate timelines for the pieces the state cannot finish. Most disapprovals that reach the Federal Register are the residue of these negotiations, the provisions on which the two sides could not agree. The formal machinery of approval and disapproval is thus the visible tip of a continuous bargaining relationship, and the clocks and sanctions are the leverage that keeps the bargaining honest.

What actually happens when EPA disapproves a state implementation plan?

Two clocks start. Eighteen months after a disapproval or a failure-to-submit finding, the first mandatory sanction applies: new or modified major sources must offset emission increases at a two-to-one ratio. The highway-funding cutoff follows six months later, at the twenty-four-month mark, when EPA must also impose a federal implementation plan.

The sanctions clock runs under section 179. Eighteen months after EPA’s disapproval or its finding that a required submission never arrived, the first sanction applies in the affected area unless the deficiency has been corrected: new or modified major stationary sources must offset their emission increases at a ratio of two to one, meaning a new facility must secure twice as many tons of reductions elsewhere as it adds. The offset sanction raises the price of building new polluting capacity in a jurisdiction whose plan has failed, directing investment pressure at the state’s planning process.

If the deficiency persists, the second sanction follows at the twenty-four-month mark: a cutoff of federal highway funding for most projects in the area, with exceptions for safety improvements, transit, and a narrow set of other categories. Highway dollars matter to every governor, which is why the sanction works as leverage rather than as punishment for its own sake. Both sanctions lift when the state submits a correction and EPA approves it, so the pressure lasts exactly as long as the failure does.

On the same twenty-four-month schedule, section 110(c) imposes a duty on EPA itself: the agency must promulgate a federal implementation plan for the state or the portion of the state affected, unless the state corrects the deficiency and wins approval first. A FIP is EPA writing the plan the state did not: federal emission limits, compliance schedules, and sometimes trading programs that bind sources directly, administered by the federal government rather than the state agency. The NOx SIP Call at the turn of the century produced federal plans for states that missed their submission duties; the 2011 Cross-State Air Pollution Rule included federal plans for states without adequate good neighbor submissions; and in 2023 EPA issued a Good Neighbor Plan, itself a FIP covering more than twenty states, which the Supreme Court stayed in June 2024 in Ohio v. EPA while challenges proceeded. The federal backstop is real, and states bargain against its arrival.

A federal plan does not run itself. EPA must administer the FIP’s limits, process the permits or trading allowances it creates, and enforce against violators, all with federal staff in place of the state agency. The arrangement is deliberately uncomfortable for both sides: the state loses control over requirements imposed on its own sources, and EPA inherits administrative burdens it would rather the state carried. That mutual discomfort is the design. A state can reclaim its program at any point by submitting a corrected plan and winning EPA’s approval, which withdraws the FIP and stops the sanctions clocks, and most FIP episodes end that way, with the federal plan serving as the prod rather than the permanent regime.

Partial disapproval deserves separate attention because it is the most common trigger of the clocks. EPA frequently approves the bulk of a state’s submission while disapproving a severable element, often the interstate transport provisions, which means the sanctions and FIP clocks run only for the disapproved portion while the approved portions take effect as federal law. The state then faces a focused correction task rather than a wholesale rewrite, and the federal plan, if it comes, covers only the gap the state left. This surgical quality is why the machinery tolerates imperfection: the clocks punish only the deficiency, and the rest of the plan keeps working.

The practical sequence, then, is unforgiving but legible. Disapproval starts both clocks at once. The state has eighteen months to correct before the first sanction bites, and twenty-four months before the second sanction and the federal plan arrive together. Most disapprovals never reach the sanctions stage because the state negotiates a correction with EPA’s regional office inside the window; the clocks function as a timetable for settlement rather than as a punishment schedule. The states that let the clocks run out discover that the statute meant what it said: the federal plan displaces state choices the state declined to make.

One institutional note closes the loop on disapproval. EPA’s regional offices, not headquarters alone, manage most plan reviews, and the negotiation between a state air agency and its regional office is where the machinery usually succeeds or fails. The Federal Register notices record the outcomes; the years of technical exchange before them decide the content. A disapproval published in the Federal Register is typically the visible end of extended exchange between the state and the regional office.

The pipeline: from rule to stack

With both mechanisms in view, the full pipeline runs in a fixed order. Congress writes the authorizing sentence. EPA runs the 307(d) chain and produces a final rule with a record: a revised ambient standard, a performance standard, a hazardous pollutant rule. Each state then writes its implementation plan translating the national rule into enforceable limits for its sources, and EPA approves, conditionally approves, or disapproves each submission. The approved plan’s limits flow into permits: Title V operating permits compile every applicable duty for a major facility into a single document, and preconstruction permits under new source review impose technology and offset duties before new capacity is built. Inspectors test compliance through stack tests, continuous emission monitors, and records review. Enforcement, by the state, by EPA under section 113, or by citizens under section 304, backs every prior step.

Each handoff is a place where the gap opens or closes. A rule vacated for a thin record never reaches any state plan. A disapproved plan never delivers limits to any permit. A permit issued without a required limit binds no one to that limit. An inspection program without monitors finds no violations. The machinery’s redundancy is the point: the 307(d) record protects the rule step against courtroom failure, the sanctions and FIP clocks protect the state step against delay, and the permit and enforcement steps convert paper limits into measured tons. When the chain holds, the statutory sentence reaches the smokestack; when any link fails, authorization remains ink on paper.

The permit step deserves a closer look because it is where the public first meets the obligation. A Title V operating permit gathers every applicable duty for a major source, federal and state, into a single document with monitoring, recordkeeping, and reporting terms, typically issued for five-year terms by the state agency under EPA-approved programs. Preconstruction review adds a second permit layer: new or modified major sources must obtain permits before building, installing best available control technology in attainment areas or the lowest achievable emission rate in nonattainment areas, and offsetting new emissions where the air already violates the standard. A source that builds without the required preconstruction permit violates the Act from the first day of construction, which is why permit challenges are a favored enforcement target.

Inspection and enforcement close the chain with measured facts. State inspectors conduct stack tests and review continuous emission monitoring data that major sources must collect and report; the monitors, required by many rules and permits, create a compliance record the facility cannot easily dispute. When violations surface, the state usually acts first, but section 113 lets EPA step in with its own enforcement where state action is absent or inadequate, a practice known as overfiling, and section 304 lets citizens sue violators directly and sue EPA for missed nondiscretionary duties. Each enforcement path cites the permit, the permit cites the plan, the plan cites the rule, and the rule cites the statute: the chain read backward is the enforcement case.

The citizen suit is the chain’s third enforcement leg and its most distinctive. Section 304 authorizes any person to sue a violator of an emission standard or limitation, including limits in an approved state plan or a Title V permit, and to sue EPA for failing to perform a nondiscretionary duty, after sixty days’ notice to the agency, the state, and the alleged violator. The notice period gives the government a chance to act first; when it does not, private plaintiffs litigate the same chain backward, from violated limit to the plan and rule behind it. The provision makes the public a permanent auditor of the machinery, which is why regulated sources treat the docket and the permit as litigation exhibits from the day they are written.

The series thesis, restated through this article’s evidence: statutes authorize, but procedure delivers. The Clean Air Act’s drafters understood this better than most, which is why they wrote the procedure into the statute rather than trusting the APA default or the states’ good intentions. The 307(d) chain and the plan approval machinery are the Act’s answer to the gap between authorization and action, and four decades of dockets, disapprovals, sanctions clocks, and federal backstop plans show the answer working, slowly, in public, on the record.

Mechanism Three: The Obligations That Reach Individual Facilities

The statute speaks to the public in the language of health and welfare, but it speaks to a plant manager in the language of limits, deadlines, and permits. No enforcement action has ever cited a facility for violating the Clean Air Act in the abstract. Every citation names a particular duty that a particular unit failed to perform: a performance standard ignored, a preconstruction permit never obtained, a toxic limit exceeded, an operating permit condition breached. The four instruments described in this section are the machinery that turns congressional sentences into those particular duties. Each aims at a different population of sources, each rests on a different theory of what technology can be demanded, and each carries its own history of litigation. A practitioner who can keep the four straight can read any facility’s obligations the way a physician reads a chart.

On a real facility the four instruments overlap. A petroleum refinery may face performance limits for a dozen of its units under section 111, toxic limits for its process vents and storage vessels under section 112, a nonattainment permit with offsets for a proposed expansion under Part D, and a Title V permit gathering all three into one enforceable document with its monitoring terms. The overlap is deliberate. Congress built redundancy into the source level machinery on the theory that different instruments catch different failures: the performance measure sets the technology floor, the preconstruction permit controls growth, the toxics program addresses the pollutants the other programs miss, and the operating permit makes the whole collection legible and enforceable. The cost of the redundancy is complexity, and the complexity is why the facility’s environmental staff, its counsel, and its permitting authority spend so much of their effort on the question this section answers: which duties apply, to which units, under which instrument.

Standards for new sources under section 111

Section 111 of the act, codified at 42 U.S.C. 7411, gives the EPA authority to set performance standards for categories of stationary sources whose emissions the agency finds endanger public health or welfare. The paired citation, section number plus United States Code location, is the standard form in federal environmental practice; readers new to the convention may consult how to read a federal statute. The mechanism works in two steps. First the agency lists a source category, a formal finding that the category contributes meaningfully to dangerous air pollution. Then it sets a standard of performance for new, modified, and reconstructed sources in that category, defined as the degree of emission limitation achievable through the best system of emission reduction the agency determines has been adequately demonstrated, taking cost, energy requirements, and nonair quality health and environmental effects into account.

That definition carries the whole philosophy of the provision. The limit is technology based rather than health based: the question is what the best demonstrated system can achieve at a cost the agency deems reasonable, not what the ambient air requires. The first standards, issued in December 1971, covered five categories: fossil fuel fired steam generators, incinerators, Portland cement plants, nitric acid plants, and sulfuric acid plants. The list grew steadily from there, eventually reaching more than seventy categories that sweep from petroleum refineries and municipal waste combustors to stationary gas turbines and equipment leaks at chemical plants. Once a limit exists, it applies automatically to any new source in the category, and to any existing source that undergoes a modification, defined as a physical or operational change that increases emissions. That modification trigger is where much of the litigation has lived, because owners and the agency have fought for decades over which plant changes count as modifications and which count as routine maintenance.

The listing decision itself follows a statutory test. Section 111(b)(1)(A) directs the agency to list a category where, in its judgment, the category’s emissions cause or contribute significantly to air pollution that may reasonably be anticipated to endanger public health or welfare. The endangerment finding is the jurisdictional key: without it, no limit follows. Once listed, a category stays on the regulatory agenda because section 111(b)(1)(B) requires the agency to review each limit at least every eight years and revise it as appropriate, a rolling duty that has produced successive generations of limits for major categories as control technology advanced. The eight year review is one of the statute’s quiet engines of stringency. It converts a limit that reflected the best demonstrated system of one decade into the floor for the next decade’s inquiry, and it gives litigants a recurring lever: a missed review deadline is a discrete, suable failure, unlike the vaguer complaint that a program has grown stale.

The D.C. Circuit’s early encounters with the provision set the tone for everything that followed. In Portland Cement Association v. Ruckelshaus, 486 F.2d 375 (D.C. Circuit 1973), the court upheld the cement limit while insisting that the record demonstrate the limit’s achievability, a demand that taught the agency to build technical records thick enough to survive review. In Sierra Club v. Costle, 657 F.2d 298 (D.C. Circuit 1981), the court sustained the 1979 power plant standards and accepted the agency’s consideration of cost alongside technology, while also policing the ex parte contacts through which interested parties had lobbied the agency off the record. Together the two decisions established the provision’s working rules: the agency enjoys wide latitude in choosing among demonstrated technologies and in weighing cost, but it must show its work in a record that a court can examine.

Section 111 also reaches existing sources, though by a different route. Subsection (d) directs the agency to issue guidelines for pollutants that are neither criteria pollutants nor hazardous air pollutants, and directs states to write plans imposing standards of performance on the existing sources in listed categories. The provision slept through most of the statute’s history and then became the vehicle, in 2015, for the agency’s carbon dioxide guidelines for existing power plants. In 2022 the Supreme Court held in West Virginia v. EPA, 597 U.S. 697 (2022), that the provision as written did not authorize the generation shifting approach those guidelines employed, applying the major questions doctrine to require clear congressional authorization for agency action of vast economic and political consequence. The holding left the new source program untouched and left subsection (d) available for measures that fit its text, but it confirmed that the provision’s most aggressive uses would face searching review.

Two different permits for building something new

A company that wants to build or substantially modify a major emitting facility faces one of two preconstruction permitting regimes, and which one applies depends entirely on the quality of the air where the facility will stand. The two programs share a name, new source review, and almost nothing else.

In areas meeting the ambient standards, Part C of the act imposes prevention of significant deterioration review. The program applies to major emitting facilities, defined as sources with the potential to emit one hundred tons per year if they belong to one of twenty eight listed industrial categories, or two hundred fifty tons per year for everyone else. A covered source must obtain a permit before construction demonstrating four things, each with its own technical apparatus. It must apply best available control technology, a case by case determination of the maximum reduction achievable for each regulated pollutant, taking energy, environmental, and economic impacts into account. It must show through air quality modeling that the project will not cause or contribute to a violation of any ambient limit. It must show that the project will not consume more than the statutory increment, the small additional amount of pollution the act allows in clean air areas beyond a baseline date, with the tightest increments reserved for Class I areas such as large national parks and wilderness areas. And it must analyze additional impacts including visibility, soils, and vegetation. The program’s premise is that clean air is a resource worth spending carefully: growth is permitted, but each project must justify its share of the remaining clean air budget.

The best available control technology determination follows a disciplined method the agency has refined through guidance. The permitting authority works from the top down: it identifies every available control option for the pollutant, eliminates the technically infeasible, ranks what remains by effectiveness, and then evaluates the top option against energy, environmental, and economic impacts before either selecting it or documenting why it was rejected and moving to the next. The method is demanding on applicants, who must produce the engineering analysis, and on permitting authorities, who must defend each elimination step in a record the courts can review. Prevention of significant deterioration also carries a distinctive consultation duty. For projects near Class I areas, the permitting authority must notify the federal land manager, the official charged with protecting the parks and wilderness areas Congress placed in the strictest increment class, and must consider the manager’s analysis of adverse impact on air quality related values including visibility. A source can satisfy every modeling test and still face a difficult permit proceeding if the land manager’s visibility analysis is adverse.

In areas violating the standards, Part D imposes nonattainment new source review, and the premise reverses. Here the air is already too dirty, so new growth must pay for itself in cleaner air. The control technology test becomes the lowest achievable emission rate, defined as the most stringent limit achieved in practice by any source in the category or contained in any state’s plan, whichever is stricter, with no consideration of cost. A source that meets the technology test must still obtain emission offsets, real and enforceable reductions from other sources in the same nonattainment area, or from another area of equal or higher classification whose emissions contribute to the violation, at ratios greater than one to one that grow stricter as the area’s classification worsens. A new factory in a severe ozone area might have to secure one and three tenths tons of reductions for every ton it will emit. The offset market this requirement created is one of the statute’s quieter inventions: a price on new emissions in dirty air areas, paid in retired pollution from old ones.

The trigger for both programs turns on quantitative thresholds the agency set by regulation. A physical or operational change at an existing major source triggers review as a major modification only if the resulting emissions increase exceeds the significance level for the pollutant: forty tons per year for nitrogen oxides, sulfur dioxide, and volatile organic compounds, one hundred tons for carbon monoxide, with lower levels for fine particles and their precursors. Below those levels, the change escapes review even in nonattainment areas, which is why project planning so often turns on emissions accounting at the margin. The boundary between a modification and routine maintenance produced one of the fiercest fights in the program’s history. The agency’s 2003 equipment replacement provision would have excluded a defined class of replacements from review; the D.C. Circuit vacated it in New York v. EPA, 443 F.3d 880 (D.C. Circuit 2006), holding that the agency had stretched the statutory term past its meaning. The decision returned the program to case by case determinations and confirmed that the modification trigger, dull as it sounds, is where the two permitting regimes most often succeed or fail in court.

The two programs collided memorably in the litigation that produced Chevron deference. In Chevron U.S.A. v. Natural Resources Defense Council, 467 U.S. 837 (1984), the Supreme Court upheld the agency’s bubble policy, which treated an entire plant as a single stationary source so that decreases at one unit could offset increases at another without triggering new source review. The decision’s famous two step approach for reviewing agency interpretations of ambiguous statutes would dominate administrative law for four decades, as the judicial review section below explains. For the practitioner the episode’s lesson is narrower: the definition of the regulated unit, source versus plant, decides how often the permit trigger fires, and that definitional fight has continued ever since.

The toxics program Congress rebuilt in 1990

For its first twenty years, section 112 was the act’s great disappointment. The provision directed the agency to set limits for hazardous air pollutants at levels protecting public health with an ample margin of safety, a health based command that required the agency to determine safe exposure levels for substances whose risks were poorly understood. The agency managed to issue limits for only a handful of pollutants across two decades, and the program became the standard example of a health based mandate the science could not support. Congress responded in the 1990 amendments by replacing the provision’s philosophy. The rewritten section 112, codified at 42 U.S.C. 7412, is technology based from the ground up.

The rewrite begins with a list. Congress wrote nearly one hundred ninety hazardous air pollutants directly into the statute, from benzene and formaldehyde to mercury and dioxins, and directed the agency to list the categories of sources that emit them. The agency eventually listed more than one hundred seventy categories, from petroleum refineries and chemical plants to dry cleaners and chromium electroplaters. For major sources in listed categories, defined as sources emitting ten tons per year of any single listed pollutant or twenty five tons of any combination, the agency must set limits reflecting the maximum achievable control technology. The statute defines the minimum stringency of those limits with unusual precision: for existing sources, the limit cannot be less stringent than the average emission limitation achieved by the best performing twelve percent of sources in the category, or the best five sources where the category has fewer than thirty; for new sources, the floor is the best controlled similar source. The agency may go beyond the floor where it finds further reductions achievable considering cost, energy, and nonair quality impacts, and it must revisit each limit at least every eight years to assess the residual risk remaining after the technology measure applies. Smaller area sources face generally available control technology, a less demanding test suited to the dry cleaner and the autobody shop.

The program’s scale is what distinguishes it. Where the old section 112 produced a handful of limits, the rewritten provision produced a steady stream of technology based rules across the 1990s and 2000s, each built on an industry survey of what the best performers already achieved. The boiler standards, finalized in March 2011 and reconsidered later that year, applied the methodology to the country’s vast population of industrial, commercial, and institutional boilers. The mercury and air toxics standards for power plants, signed in December 2011, extended it to the electric utility sector for the first time, requiring controls for mercury, acid gases, and other hazardous pollutants from coal and oil fired generating units. Both rulemakings illustrated the provision’s characteristic discipline: the limit follows demonstrated performance rather than leading it, and the floor calculation, built from data on real facilities, is where the stringency is actually decided.

Two backstop provisions keep the program moving when the main mechanism stalls. Section 112(j), the provision practitioners call the hammer, provides that where the agency misses the statutory deadline for a category’s limit, the state must set case by case maximum achievable limits for the major sources in that category through the Title V permit process. The hammer converts agency delay into state action, which gives the agency a powerful incentive to meet its own schedule. Section 112(f) addresses what the technology limits leave behind: within eight years after a technology measure issues, the agency must assess the residual health risk to the most exposed individual and, where the risk remains unacceptable, impose additional controls. The residual risk program has been slower and more contested than the technology program, because it returns the agency to the health based judgments the 1990 rewrite was designed to avoid, but it supplies the provision’s long run health guarantee. A third element, the urban air toxics strategy of section 112(c)(3) and (k), directed attention to the area sources, the small businesses and neighborhood operations whose individual emissions are modest but whose cumulative contribution in cities is large.

One permit that gathers every obligation

Until 1990 the act imposed its duties through a scattered collection of rules, plans, and orders, and no single document told a facility what the law required of it. Title V, added by the 1990 amendments as sections 501 through 507 and codified at 42 U.S.C. 7661 to 7661f, created the operating permit program to fix that. A major source, generally one with the potential to emit one hundred tons per year of any regulated pollutant or ten and twenty five tons of hazardous pollutants, along with any source subject to a performance or toxics limit and any source in the acid rain program, must obtain a comprehensive operating permit from its state or, where the state lacks an approved program, from the agency itself. The permit compiles every applicable requirement, the performance limits, the toxics limits, the new source review conditions, the plan provisions, into one document with monitoring, recordkeeping, and reporting terms sufficient to assure compliance. The public gets notice and an opportunity to comment on each draft permit, and the agency reviews each proposed permit and may object.

The permit’s most discussed feature is the permit shield. Section 504(f) provides that compliance with a permit’s terms is deemed compliance with the act’s applicable requirements, as identified in the permit. Stated plainly, the shield means that a source holding a complete and accurate permit, and operating within its terms, cannot be pursued for violating a requirement the permit addressed. The shield is conditional rather than absolute: it protects only the requirements the permit actually includes, and it fails where the permit was issued on inaccurate information or omitted an applicable duty. The provision thus rewards careful permit writing and punishes sloppy applications, because the source’s protection extends exactly as far as its disclosure. The neighboring water statute took a different path worth noting: under the Clean Water Act section 404 permit program, the discharge permit has been the central regulatory instrument since 1972, while the air statute operated for two decades before acquiring a comparable document.

The permit program’s procedural details matter as much as its architecture. A source awaiting its initial permit operates under an application shield: a timely and complete application protects the source from enforcement for operating without a permit while the agency processes the paperwork, provided the source complies with applicable requirements in the interim. Once issued, the permit carries ongoing duties, including at least semiannual monitoring reports, prompt reports of deviations, and an annual compliance certification signed by a responsible official under penalty of perjury. The agency’s oversight role is structured around two clocks. After the state proposes a permit, the agency has forty five days to object; if it does not, any person may petition the Administrator to object within sixty days after the forty five day period expires, and the Administrator must grant or deny the petition within sixty days more. The petition provision, section 505(b)(2), has become a regular channel for advocacy organizations to challenge state issued permits, and the resulting petition orders form a body of agency precedent on what Title V requires. The program is funded by fees: section 502(b)(3)(A) required a presumptive fee, originally twenty five dollars per ton of regulated pollutant, adjusted for inflation, sufficient to cover the reasonable costs of the permit program. The fee provision made Title V the rare environmental program that pays for its own administration.

Mechanism Four: The Courthouse Rules

The act does not leave judicial review to the general default. Section 307, codified at 42 U.S.C. 7607, writes a specialized review regime that differs from ordinary administrative law in three ways that decide cases: where a challenge must be filed, how fast it must be filed, and what arguments survive to be heard. A litigant who masters the statute’s substance but misses these provisions will lose before the court reaches the merits.

The sixty-day clock and the only court that may hear the case

A petition for review of a nationally applicable regulation or other final agency action must be filed in the United States Court of Appeals for the District of Columbia Circuit within sixty days after the action’s publication in the Federal Register. A locally or regionally applicable action, such as the approval of a single state’s implementation plan, goes instead to the court of appeals for the appropriate regional circuit. Where the action’s reach is mixed, the Administrator may make a formal finding that it rests on a determination of nationwide scope or effect, which routes the petition to the District of Columbia Circuit as well. The venue provisions are jurisdictional rather than matters of convenience: a petition filed in the wrong circuit, or one day after the sixtieth day, is dismissed without regard to its merits. There is no forum shopping under this statute. The concentration of nationally applicable cases in a single court has made the District of Columbia Circuit the effective court of appeals for air law, with a docket and a body of precedent deeper than any other circuit’s in the field.

The Administrator’s nationwide scope finding deserves a closer look, because it decides the forum for the rules that matter most. For national emission limits, ambient standard revisions, and other measures that apply across the country, the agency routinely includes a finding that the action rests on a determination of nationwide scope or effect, which sends any challenge to the District of Columbia Circuit even where the action also contains region specific elements. The finding is itself reviewable, but courts have sustained it where the agency’s reasoning for national treatment is coherent. The practical result is that nearly every consequential Clean Air Act rulemaking is litigated in one court, before judges who have seen the statutory scheme dozens of times. State plan approvals, which apply to a single state, travel the other road to the regional circuits, where the question is usually whether the agency reasonably evaluated the state’s submission rather than whether the underlying national policy was lawful.

The sixty day period runs from publication, not from when a challenger learns of the action or feels its effects. Trade associations and public interest organizations therefore maintain standing watches on the Federal Register, and the filing decision is typically made within days of a major rule’s appearance. Once the circuit court decides, the losing side may seek certiorari in the Supreme Court, which is how the most consequential air cases have reached the justices. Massachusetts v. EPA traveled exactly this path: a petition for review of the agency’s denial of a rulemaking petition was filed in the District of Columbia Circuit under section 307, and the Supreme Court’s 2007 decision followed on writ of certiorari. The specialized venue gets the case to the right appellate court; only the Supreme Court’s discretionary review takes it further.

Two traps surround the sixty day clock, and both punish the unwary. The first is reconsideration. A party that asks the agency to reconsider a final measure does not pause the filing deadline; the petition for review must still be filed within sixty days, so experienced counsel file the petition and the reconsideration request in parallel and ask the court to hold the case while the agency responds. The second is guidance. The act’s specialized procedures apply to rules, and the agency issues a large volume of guidance documents, memoranda, and policy statements that interpret rules without going through rulemaking. Guidance does not bind like a rule, and a challenger that treats it as binding will find the court asking whether the document is final agency action at all. In Appalachian Power Co. v. EPA, 208 F.3d 1015 (D.C. Circuit 2000), the court held that an agency guidance document on monitoring requirements was a legislative rule in disguise and had to go through notice and comment. The decision is the standing warning for both sides: the agency cannot make binding law without the rulemaking machinery, and a challenger cannot attack a nonbinding document as though it were a rule.

The record is the entire case

Section 307(d) layers rulemaking procedures on top of the general notice and comment model, and those procedures exist to build the record on which the court will decide. The agency must maintain a public docket containing the data and methodology behind the proposal, publish a statement of basis and purpose explaining the measure, provide an opportunity for oral presentation where the issues warrant it, and respond to each significant comment, criticism, and new data point submitted during the comment period. The response to comments is not a courtesy. It is the document in which the agency demonstrates, issue by issue, that it confronted the evidence against its position and gave reasons for proceeding anyway.

The proposal and the final measure must also stand in a particular relationship. Under the logical outgrowth doctrine, the final rule must be a logical outgrowth of the proposal: interested parties must have had fair notice of what the agency might do, so that the comment period was a genuine opportunity to respond. An agency that proposes one approach and finalizes a substantially different one, without an additional round of comment, invites vacatur on procedural grounds alone. The doctrine disciplines the common temptation to use the comment period as a suggestion box and then write a different rule. It also explains why agencies sometimes issue supplemental proposals when the comments persuade them toward a new alternative. The docket that supports all of this is specified by the statute with unusual care: section 307(d)(3) and (4) require the docket to contain the factual data on which the proposal rests, the methodology used to obtain and analyze it, and every major legal interpretation and policy consideration underlying the proposal. A thin docket is a vulnerable rule.

Judicial review then proceeds exclusively on that record. Section 307(d)(9) authorizes the court to reverse agency action found to be arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law; contrary to constitutional right; in excess of statutory jurisdiction or authority; or adopted without observance of procedure required by law. The court takes no new evidence, hears no witnesses, and conducts no de novo inquiry into the science. It reads what the agency had before it and asks whether the agency examined the relevant data, considered the important aspects of the problem, and drew a rational connection between the facts found and the choice made. A challenger who failed to put its technical evidence into the docket during the comment period will find that the court has nothing to read on its behalf.

The content of the arbitrary and capricious test, as applied to air rules, comes largely from State Farm itself. The Court’s formulation requires the agency to have examined the relevant data and articulated a satisfactory explanation for its action, including a rational connection between the facts found and the choice made. The formulation is deferential in tone and demanding in application: courts do not substitute their judgment for the agency’s on technical questions, but they do insist that the agency confronted the significant issues, answered the substantial criticisms, and explained departures from prior positions or from the evidence in its own record. The D.C. Circuit’s air docket is in large measure an extended application of that paragraph to successive generations of rules. A challenger rarely wins by proving the agency’s science wrong; it wins by showing that the agency failed to address a question its own record raised, which returns the dispute to the docket where the question should have been answered.

Arguments never made in comments are generally lost

The statute converts the comment period from an advisory exercise into the decisive litigation event. Section 307(d)(7)(B) provides that only an objection raised with reasonable specificity during the public comment period may be raised on judicial review. The provision contains a narrow safety valve: where a challenger shows it was impracticable to raise the objection during the comment period, or where the grounds arose afterward but within the review window and go to the central relevance of the measure’s outcome, the Administrator must convene a reconsideration proceeding with the same procedural rights the comment period would have afforded. Courts construe the valve narrowly. An argument that could have been made from information available during the comment period, but was not, is forfeited even if it would have won.

The safety valve’s two prongs cover distinct situations. Impracticability typically means the information was not available during the comment period: a study the agency placed in the docket after the comment period closed, or data generated by events that postdated the proposal. The new grounds prong typically means an intervening legal development, such as a court decision issued after the comment period that changed the legal landscape for the rule. Both prongs require, in addition, that the objection go to the central relevance of the measure’s outcome, which screens out peripheral complaints. And critically, the valve leads to reconsideration by the agency, not to a free pass in court. The challenger that invokes it must persuade the Administrator to convene a new proceeding, with the original rule remaining in force in the meantime.

The practical consequence is that comment drafting is merits briefing by another name. Regulated companies, states, and advocacy organizations retain counsel and technical experts during the proposal stage precisely because the final measure’s preamble will cite the comments it answered and the reviewing court will measure the agency’s reasoning against them. The well known pattern of air litigation, in which the District of Columbia Circuit decides challenges on grounds fully aired in the docket, follows directly from this design. Congress made a deliberate tradeoff: finality and expedition in exchange for a genuine opportunity to be heard before the decision, with forfeiture as the price of silence.

The deference regime changed in 2024

For four decades the most cited administrative law decision in the country was itself a Clean Air Act case. Chevron U.S.A. v. Natural Resources Defense Council, 467 U.S. 837 (1984), upheld the agency’s bubble policy for defining a stationary source under new source review and announced a two step approach: where a statute is ambiguous, courts ask whether the agency’s interpretation is reasonable, and if so they defer. Generations of air rules survived challenges because reviewing courts found the statutory text ambiguous and the agency’s reading reasonable.

In 2024 the Supreme Court overruled Chevron in Loper Bright Enterprises v. Raimondo, 600 U.S. 181 (2024). The Court held that the Administrative Procedure Act requires courts to exercise independent judgment in deciding whether an agency has acted within its statutory authority, and that courts may not defer to an agency’s interpretation of law simply because a statute is ambiguous. Agency views may still carry weight to the extent they persuade, under the older Skidmore principle that respects thoroughness and expertise, but persuasion is not deference. The Court added that holdings decided under Chevron remain decided; the overruling changes the measure for future cases rather than reopening past judgments. For Clean Air Act practice the effect is straightforward: challenges to agency readings of ambiguous provisions, of which the act has many, will be decided by judges applying ordinary tools of statutory construction to the agency’s best arguments, without the thumb on the scale that Chevron supplied. Briefs filed after the decision read differently, and so will the opinions that answer them.

What the overruling does not change is equally worth stating, and briefly. Loper Bright addressed the interpretation of statutes, not the review of technical judgments. An agency’s findings of fact, its weighing of scientific evidence, and its policy choices within a statutory delegation remain subject to the arbitrary and capricious measure described above rather than to de novo judicial redetermination. The Skidmore respect the Court preserved means that a thorough, longstanding agency construction, consistently applied and well explained, will still persuade even where it no longer commands deference. The practical effect for Clean Air Act litigation is a shift in where cases are fought: less on whether the agency’s reading of an ambiguous phrase was reasonable, more on what the phrase means as a matter of ordinary legal interpretation, with the agency’s technical record continuing to carry weight on the questions of fact and policy beneath the legal one.

Mechanism Five: One State’s Standards and the Rest of the Country’s Cars

The act generally assigns vehicle emission limits to the federal government, and it says so bluntly: section 209(a) preempts states from adopting or enforcing their own standards for new motor vehicles and engines. The reason is commercial rather than environmental. Automakers sell a national product, and fifty different emission regimes would fracture the market into fifty regulatory fiefdoms. Yet the statute contains an exception with a date on it, and that exception has made one state the effective co author of national vehicle policy.

The commercial logic behind preemption is worth pausing over, because it explains why the exception matters so much. A vehicle is designed years before it is sold, certified once, and distributed through a national dealer network. Separate state emission regimes would require separate certifications, separate calibrations, and in the limit separate models, multiplying the cost of compliance and balkanizing the used vehicle market as cars migrate across state lines. Congress accepted that logic and federalized the field, then carved out the single state whose regulatory program predated the federal one. The carve out preserved California’s program without sacrificing the national market’s integrity, at least so long as only one state used it. Section 177 changed the arithmetic by letting other states join California’s program, which is why a provision drafted as a grandfather clause became the hinge of national vehicle policy.

The waiver and the 1966 cutoff

Section 209(b), codified at 42 U.S.C. 7543, authorizes the Administrator to waive preemption for any state that adopted emission limits for new motor vehicles before March 30, 1966. Only California had done so, which makes the provision a waiver for one state wearing the clothes of a general rule. The statute does not give the agency free discretion over the waiver. It provides that the Administrator shall grant the waiver unless she finds one of three things: that the state’s determination that its limits will be at least as protective as the federal limits is arbitrary and capricious; that the state does not need its own limits to meet compelling and extraordinary conditions; or that the state’s limits and enforcement procedures are inconsistent with section 202(a), the provision directing federal limits. The presumption runs in favor of the waiver, and the findings required to defeat it are demanding by design.

The waiver’s history is a record of grants far outnumbering denials. Across the decades the agency granted dozens of waivers covering successive generations of California’s vehicle program, from early hydrocarbon and nitrogen oxide limits through the low emission vehicle programs. Denials have been rare enough to be memorable. The structure explains why: Congress wrote the exception for a state whose air problems were visibly extraordinary, the Los Angeles basin’s smog was already infamous when the cutoff date was chosen, and it placed the burden on the agency to justify refusing rather than on the state to justify asking.

The three findings that defeat a waiver each have a developed meaning. The protectiveness determination asks whether the state’s limits, taken as a whole, will be at least as protective of public health and welfare as the federal limits; the comparison is aggregate rather than pollutant by pollutant, which gives the state room to design a differently shaped program. Compelling and extraordinary conditions refers to the geography, meteorology, and vehicle density that made California’s smog problem distinctive: a basin topography that traps pollution, sunlight that cooks it into ozone, and a car fleet larger than most countries’. Consistency with section 202(a) turns largely on lead time and technological feasibility: the state must give manufacturers enough time to comply, and the limits must be within the reach of available technology. The zero emission vehicle program, first adopted in 1990 and repeatedly waived, illustrates all three findings in operation. Its sales mandates survived protectiveness review because the aggregate program was at least as protective as the federal alternative, its justification rested on the state’s extraordinary air quality challenge, and its phased percentages gave manufacturers a defined compliance path.

The exception’s history explains its shape. California created the country’s first motor vehicle pollution control board in 1960 and adopted the first tailpipe limits, responding to the Los Angeles smog that had become a national symbol of air pollution. When Congress first set federal vehicle limits in the mid 1960s and then preempted state limits in the Air Quality Act of 1967, it preserved the program that already existed rather than destroying it. The March 30, 1966 cutoff date marks the line Congress drew between the established state program and the federal program to come. The waiver provision has never been a general delegation to states; it is a grandfather clause for the one state that was already in the field, and every waiver decision since has been an application of that original judgment to a new generation of California’s limits.

Why other states may adopt but not invent

Section 177, codified at 42 U.S.C. 7507, extends California’s limits to any other state with an approved nonattainment plan. Such a state may adopt limits identical to California’s, for which a waiver has been granted, provided it gives at least two model years of lead time. The critical word is identical. An opt in state may not write its own variant, blend California and federal requirements, or set a third limit. The statute thus contemplates exactly two vehicle emission regimes in the United States: the federal program and the California program, with other states choosing between them.

Adopting California’s limits is a state law act with federal consequences. The opt in state typically adopts the California program by regulation, sometimes by statute, and must do so early enough to give manufacturers the two model years of lead time the provision requires. The identicality requirement is enforced strictly: a state that alters the California test procedures, adds its own averaging provisions, or phases in the limits on a different schedule has not adopted identical standards and loses the protection of section 177, exposing its program to preemption. The provision thus produces genuine uniformity within each of the two markets, which is what makes manufacturer compliance planning possible. The northeastern and mid Atlantic states that formed the core of the opt in bloc adopted the program through coordinated rulemakings, and their collective sales volume is what turned California’s limits into the effective national ceiling described above.

The choice has not been theoretical. More than a dozen states adopted California’s limits in the years the provision has operated, and together with California they account for a large share of national new vehicle sales. Automakers consequently face a commercial reality the statute’s drafters anticipated: building to the stricter of the two programs is cheaper than maintaining separate California and federal fleets, so the California limits function as the effective national ceiling for much of the market. The greenhouse gas chapter of this story shows the mechanism at full strength. The Massachusetts v. EPA decision held in 2007 that greenhouse gases are air pollutants under the act, which led to the agency’s endangerment finding in December 2009. California sought a waiver for its own greenhouse gas vehicle limits, denied in late 2007 and granted in July 2009, and the agency then issued joint federal greenhouse gas and fuel economy limits for light duty vehicles in May 2010. The sequence ran through every mechanism this article describes: a Supreme Court interpretation, a waiver proceeding, and a rulemaking built on the resulting record.

The waiver’s national consequences produced a negotiated settlement worth noting. In 2009, as the agency moved toward federal greenhouse gas limits and California held a granted waiver for its own, the manufacturers faced the prospect of two binding programs with different test procedures and compliance arithmetic. The resolution was a coordinated national program: California agreed to deem compliance with the federal 2012 through 2016 model year limits as compliance with its own standards for those years, and the agency issued a single joint rule with the federal fuel economy program. The episode demonstrated the waiver’s bargaining power in its purest form. The leverage did not come from any threat the statute’s text contains; it came from the commercial arithmetic of two markets, which made a negotiated single program preferable to parallel compliance for every party at the table.

Why is California’s waiver the lever for national vehicle policy?

Because section 177 lets other states adopt California’s standards verbatim, automakers face two national markets rather than fifty separate ones. A waiver granted for Sacramento therefore sets the effective ceiling for much of the country’s fleet, since manufacturers build to the stricter of the two programs. The leverage is structural, not political.

That structural leverage explains why waiver proceedings draw the heaviest participation of any action the agency takes under the mobile source titles. A waiver decision nominally concerns one state’s limits, but its commercial consequences run through every opt in state and, through manufacturer compliance planning, through the federal fleet as well. The docket for a major waiver routinely contains submissions from automakers, suppliers, dealers, labor organizations, other states, and foreign governments, each addressing the three statutory findings. The Administrator’s decision must then survive the same section 307 review described above, on the same sixty day clock, with the same forfeiture rule for unraised objections. The waiver is a small provision with a long shadow, and the shadow falls across the entire vehicle market.

The Comment-or-Forfeit Rule

Strip away the venue provisions, the measures of review, and the deference doctrines, and the most consequential procedural feature of Clean Air Act practice is a single sentence in section 307(d)(7)(B): objections not raised with reasonable specificity during the comment period generally cannot be raised in court. The provision makes the rulemaking docket, not the courtroom, the real venue for shaping an air rule. Everything else in the review regime is elaboration on that choice.

The design rewards a particular kind of participant. The winners in air litigation are rarely the parties with the best argument discovered after the final rule; they are the parties that built the best record before it. Industry groups that submit detailed technical comments, states that file modeling analyses, and organizations that identify legal vulnerabilities during the proposal stage arrive in court with objections the agency was forced to answer. Parties that stayed silent, or that commented in generalities, arrive with forfeiture. The asymmetry is intentional. Congress wanted rules that could be finalized and relied upon, and it was willing to purchase finality by closing the courthouse door to late arguments.

The rule shapes litigation strategy from the first day of a rulemaking. Sophisticated participants treat the proposal as a discovery device: they file detailed comments identifying every legal vulnerability they might later litigate, attach the technical analyses the agency would need to answer, and propose alternative approaches the agency must consider or explain away. Trade associations pool resources to fund the modeling and engineering studies that individual members cannot afford, precisely because a study submitted during the comment period becomes part of the record the court will read, while the same study commissioned after the final rule is inadmissible. The term of art is exhaustion, borrowed from the administrative law requirement that litigants exhaust their remedies before the agency, and the Clean Air Act’s version is stricter than the general default because the statute writes the forfeiture into its text rather than leaving it to judicial doctrine.

Can a challenger raise an argument in court that it never raised in comments?

Generally no. Section 307(d)(7)(B) bars objections not raised with reasonable specificity during the comment period, with a narrow reconsideration path when raising the point was impracticable or the grounds arose afterward. A lawyer who discovers the winning argument after the final rule has usually discovered it too late.

The narrowness of the exception deserves emphasis, because practitioners regularly overestimate it. Reconsideration requires showing impracticability or genuinely new grounds plus central relevance to the measure’s outcome, and courts have not treated the provision as a second comment period for arguments that better lawyering would have caught the first time. The working advice that follows is correspondingly blunt: treat the proposal’s publication as the opening of the case, staff the comment effort as the trial, and calendar the final rule as the verdict. The courtroom matters, but by the time the judges read the briefs, the dispositive work has usually been done.

The reasonable specificity standard is where close cases are fought. A comment that raises an issue in general terms, without identifying the provision at stake, the data supporting the objection, or the alternative the agency should have considered, may not preserve the point. Courts have held that vague or conclusory comments do not satisfy the statute, while detailed comments that fairly apprised the agency of the objection do. The lesson for the commenter is mechanical: name the issue, cite the record evidence, propose the alternative, and explain why the agency’s approach fails. For the agency, the lesson runs the other way: the response to comments must engage each fairly raised objection on its merits, because a reviewing court will measure the response against the comment and treat silence as concession.

Why a New Administration Cannot Simply Erase a Rule

A persistent misunderstanding holds that a change of administration brings a free hand to undo the prior administration’s air rules, as though each election reset the regulatory slate. The law is otherwise. Rescinding or substantially revising a final rule is itself agency action, and it must survive the same procedural and substantive demands as the original rulemaking: a proposal, a comment period, a record, and a reasoned explanation capable of withstanding section 307 review.

The governing precedent is Motor Vehicle Manufacturers Association v. State Farm, 463 U.S. 29 (1983), which reviewed an agency’s rescission of a passive restraint requirement. The Court held that an agency changing course must supply a reasoned analysis for the change beyond what would be required for a fresh policy on a blank slate. It must confront the factual findings underlying the prior rule, explain why they no longer support it or why other considerations outweigh them, and consider the reliance interests that regulated parties, states, and the public built around the existing requirement. A later decision, FCC v. Fox Television Stations, 556 U.S. 502 (2009), refined the point: the agency must acknowledge that it is changing position and show good reasons for the new policy, with a more detailed justification required where the new policy contradicts prior factual findings or upsets serious reliance.

The reliance interests the doctrine protects are concrete. A power company that installed controls to meet a mercury limit, a state that rewrote its implementation plan around a federal guideline, an automaker that tooled factories for a tailpipe limit: each ordered its affairs on the assumption that the published rule was the law, and each would bear stranded costs if the rule vanished without a reasoned replacement. State Farm requires the agency to weigh those costs openly rather than ignoring them. The opinion also rejected an asymmetry the government had urged, that rescission should face lighter review than promulgation. The Court held the opposite: a change in course is judged by the same measure, because the reasons that made the original rule lawful do not evaporate when a new administration prefers a different outcome. One boundary of the constraint deserves stating plainly. The doctrine binds the agency, not Congress. A statute can always be amended, and an amendment that rewrites the underlying delegation needs no reasoned explanation beyond the legislative process itself. The durability of air rules against executive reversal is a product of administrative law, not of legislative entrenchment.

The constraint exists for reasons the statute’s structure makes visible. Congress delegated rulemaking power to the agency as a continuing institution, not to the transient officeholder who happens to direct it. The section 307 machinery, the record, the response to comments, the sixty day review, assumes that agency action rests on reasons stated in public and testable in court. A reversal announced without engaging the prior record would convert reasoned decisionmaking into preference, and the courts that Congress assigned to police the boundary have not permitted the conversion. None of this makes air rules permanent. It makes them durable against anything less than the full procedural effort that created them: a new proposal, a new record, and reasons sufficient to persuade a court that the change, too, was reasoned.

The Verdict: Procedure Is Where the Statute Becomes Real

The series thesis holds that the interesting distance in American legislation lies between what a statute authorizes and what is actually done, and no article in the series illustrates the distance better than this one. The Clean Air Act authorizes the EPA to protect public health with an adequate margin of safety, a sentence of breathtaking ambition. What is actually done with that sentence is decided in dockets and permits and courtrooms: in the technology survey behind a toxics floor, in the modeling that justifies a prevention of significant deterioration permit, in the offsets a new factory must buy before breaking ground in dirty air, in the comments that preserve or forfeit the arguments a court will later hear, and in the waiver proceeding that lets one state’s limits set the terms for much of the national vehicle fleet.

The machinery’s virtues are also its costs, and an honest verdict names both. The same procedures that produce reasoned, judicially tested rules also produce delay: major rulemakings routinely take years from proposal to final action, and the comment response and litigation that follow add more. The same forfeiture rule that buys finality raises the price of participation: effective engagement requires technical expertise and legal counsel during the comment period, which advantages organized interests over diffuse ones. The same venue concentration that builds expertise in one court distances review from the regions where the rules bite. None of these costs is an argument against the machinery, which Congress chose deliberately and which has produced the emissions record the companion articles in this series document. They are the terms of the bargain, and the practitioner who understands them as terms rather than as flaws will navigate the system better than one who mistakes either the virtues or the costs for the whole story.

The practitioner’s version of that thesis fits in a working method. Read the record before reading the rule, because the preamble’s account of its own basis is where vulnerabilities show. Comment with specificity while the docket is open, because the comment period is the trial and the courtroom is the appeal. Calendar sixty days from Federal Register publication for every action that matters, and file in the circuit the statute names rather than the one convenience suggests. Treat the Title V permit as the binding document it was designed to be, and mind the shield’s conditions. Track the waiver docket if vehicles are the client’s business, because the two market structure makes Sacramento’s proceedings nationally consequential. And expect every reversal to arrive as its own rulemaking, with its own record and its own reasons, because the law gives no administration a shortcut around the machinery.

Written from the vantage of 2011, this account describes a system midcourse. The developments dated in these pages, the 2022 and 2024 decisions and the waiver grants and the national program they enabled, arrived after the frame year and were folded in where they changed the analysis. What did not change across those years is the structure: Congress writes the authorizations, the agency fills them through rulemaking, states and sources comply through plans and permits, and courts police the route. The gap between the statute’s sentences and the air’s condition is where all the interesting work happens, and the work belongs to whoever learns the machinery. That is the practitioner’s verdict in the only sense that counts: the law rewards showing up early, writing specifically, and calendaring the clock.

None of this diminishes the statute’s ambition. It locates the ambition’s price. A law that governed by pronouncement alone could be reversed by pronouncement; a law that governs through records, permits, dockets, and reasoned explanations can only be moved by the same instruments. The machinery is slow, technical, and unforgiving of inattention, which is why it rewards the participants who learn its routes. The gap between authorization and action is filled by procedure, and procedure, for all its tedium, is where the air gets cleaned.

The Rule-to-Facility Pipeline

Step	Actor	Legal instrument	Review route	Deadline
Statute	Congress	Authorizing provision, such as CAA section 111 or section 109	None; legislative act	Statutory deadline, if the statute sets one
Proposed rule	EPA, Office of Air and Radiation	Notice of proposed rulemaking in the Federal Register	Public comment; OIRA review under Executive Order 12866	Comment period, typically 60 to 90 days
Final rule	EPA	Final rule with statement of basis and purpose	Petition for review in the D.C. Circuit	Effective date, usually 30 to 60 days after publication
State plan	State air agency	State implementation plan	EPA approval, conditional approval, or disapproval under 110(k)	12 months for EPA action; sanctions at 18 months and FIP at 24 months on disapproval
Permit	State permitting authority	Title V operating permit; preconstruction permit	EPA objection within 45 days; citizen petition under 505(b)(2)	Five-year Title V terms; preconstruction permit before building
Inspection	State inspectors; EPA	Stack tests; continuous emission monitoring; records review	Compliance reports and monitoring data review	Ongoing; semiannual monitoring reports
Enforcement	State; EPA; citizens	Orders and civil or criminal actions under sections 113 and 304	Federal courts	60-day notice required before citizen suits

Studying the Machinery

Start with the map, not the cases. The article describes two machines and one overlap. The first machine is the rulemaking chain: Congress authorizes, EPA proposes, the public comments, the agency answers, the final rule issues, and a court tests the record. The second is the plan approval machinery: the national rule becomes a state implementation plan, the plan becomes a permit, the permit becomes a set of monitored duties, and enforcement backs the whole chain. The overlap is the facility, where section 111 performance standards, new source review permits, section 112 toxics limits, and the Title V operating permit converge on the same stack. Hold that picture before drilling into any single provision, because exam questions and practice problems both test whether you can place a new fact in the right machine.

Learn the statute’s skeleton by section number, because the numbers are the handles practitioners use. Section 307(d) is the rulemaking procedure: docket, oral hearing opportunity, written answer to major criticisms, record-defined review. Section 307(b) is the courthouse door: the D.C. Circuit for nationally applicable actions, regional circuits for local ones, sixty days from Federal Register publication. Section 307(d)(7)(B) is the forfeiture rule: objections not raised with reasonable specificity in comments are generally lost. Section 110 is the state’s planning duty; section 110(c) is EPA’s federal plan backstop; section 179 is the sanctions clock. Section 111 is new source performance standards; section 112 is the toxics program rebuilt in 1990 around maximum achievable control technology; Title V is the operating permit that gathers everything. Sections 113 and 304 are enforcement, agency and citizen. Section 209 is the California waiver; section 177 is the opt-in for other states. A reader who can recite what each of these does, in one sentence each, has the article’s core cold.

Master the distinctions examiners love to test. First, 307(d) versus the APA: know the three additions (docket, oral presentation opportunity, written response to significant comments) and why Congress added them (the scale of the rules, and distrust after the early-1970s litigation). Second, BACT versus LAER: BACT weighs cost and applies in clean-air areas under PSD; LAER ignores cost and applies in dirty-air areas under nonattainment review, paired with offsets at greater than one-to-one ratios. Third, the sanctions order: eighteen months brings the two-to-one offset requirement, and the highway funding cutoff arrives six months after that, at the twenty-four-month mark, together with the federal plan duty. The common error is collapsing both sanctions into the eighteen-month clock; the statute staggers them. Fourth, guidance versus rule: guidance explains, rules bind, and a document that imposes duties without notice and comment is a legislative rule in disguise that courts set aside. Fifth, Chevron versus Loper Bright: from 1984 to 2024 courts deferred to reasonable agency readings of ambiguous provisions; after 2024 courts exercise independent judgment on statutory meaning while still reviewing technical judgments under the arbitrary and capricious standard. Sixth, vacatur versus remand without vacatur: courts sometimes leave a defective rule in place while the agency repairs it, which is why an unlawful rule can keep operating.

Work the timeline discipline until it is reflex. Sixty days to petition for review, measured from Federal Register publication, not from learning of the rule. Twelve months for EPA to act on a complete state plan submission. Eighteen months from disapproval to the offset sanction; twenty-four months to the highway sanction and the federal plan. Forty-five days for EPA to object to a proposed Title V permit, then sixty days for citizens to petition. Sixty days’ notice before a citizen suit. Eight-year review cycles for section 111 standards and section 112 toxics limits. These clocks are the article’s quantitative spine, and questions about them reward the student who memorized the numbers rather than the one who understands only the concepts.

Read the cases for what each one stands for, not for their facts. Portland Cement (1973) is the docket-disclosure ancestor of 307(d). New Jersey v. EPA (2008) is the delisting-procedure vacatur that restarted the mercury rulemaking. Michigan v. EPA (2015) is cost considered at the threshold, remand without vacatur. Massachusetts v. EPA (2007) is greenhouse gases as air pollutants. Chevron (1984) is the bubble policy and the two-step deference that governed for forty years; Loper Bright (2024) is its end. State Farm (1983) is the reasoned-analysis requirement that makes rule reversal as hard as rule creation. West Virginia v. EPA (2022) is the major questions limit on transformative agency action. New York v. EPA (2006) is the equipment-replacement vacatur policing the modification trigger. Appalachian Power (2000) is guidance treated as a legislative rule. One sentence each, and the courthouse section of the article becomes a gallery of principles rather than a blur of names.

Notice the pattern that unites the article: every mechanism converts an open-ended grant of authority into a dated, reviewable act. The proposal must answer comments; the plan must be approved or disapproved within twelve months; the permit must contain monitoring sufficient to assure compliance; the waiver must survive the three statutory findings. Procedure is the statute’s way of making power legible. That is the series thesis in this article’s vocabulary, and it is the answer to the broadest kind of exam question: why does the Clean Air Act work this way? Because Congress learned that authorization without procedure is a wish, and it wrote the procedure into the law.

For structured practice, work through a legislation study notebook that drills the clocks, the findings, and the forum for each kind of challenge section by section, available at a legislation study notebook. Then set the Act’s cooperative federalism and judicial review design inside the larger system of separated powers with a U.S. government civics study guide, which tests the constitutional context these mechanisms operate within.

Frequently Asked Questions

How does EPA write Clean Air Act rules?

EPA writes Clean Air Act rules through notice-and-comment rulemaking under the Administrative Procedure Act, 5 U.S.C. 553, with additional requirements imposed by section 307(d) of the Act, 42 U.S.C. 7607(d). EPA publishes a proposed rule in the Federal Register, holds an opportunity for oral presentation of views, and must respond to each significant comment in the preamble of the final rule. Before proposal, EPA typically collects emissions data, runs air quality and cost modeling, and submits economically significant rules to the Office of Information and Regulatory Affairs for review under Executive Order 12866. The final rule is published in the Federal Register with a statement of basis and purpose and an effective date. Challenges to landmark rules, from the 1997 ozone and particulate standards to the 2023 Good Neighbor Plan, routinely turn on whether EPA followed these procedures and answered the record.

What is a new source performance standard under the Clean Air Act?

New source performance standards are national emission limits for new, modified, and reconstructed stationary sources in categories EPA has listed under section 111(b), 42 U.S.C. 7411(b). EPA must first list categories that in its judgment cause or contribute significantly to air pollution that may endanger public health or welfare, then set standards reflecting the best system of emission reduction that has been adequately demonstrated, taking cost, energy requirements, and non-air health and environmental impacts into account. The standards apply directly to new sources nationwide; existing sources are addressed separately under section 111(d). A source that begins construction or modification after the proposal date must comply by the final standard’s effective date, and violations are enforceable under section 113. NSPS for categories from cement kilns (1971) to new gas turbines (2024) show the provision’s reach.

What is the Clean Air Act California waiver?

The California waiver is the exception in section 209(b), 42 U.S.C. 7543(b), that lets California adopt its own motor vehicle emission standards stricter than federal ones, but only after EPA grants a waiver. EPA must grant the waiver unless it finds California’s determination is arbitrary and capricious, California does not need the standards to meet compelling and extraordinary conditions, or the standards are inconsistent with the lead-time and technology-feasibility requirements of section 202(a). California has used waivers since the 1960s to lead on catalytic converters, low-emission vehicles, and zero-emission vehicle mandates. The practical stakes are national because other states may copy California standards under section 177. In 2025 EPA rescinded several California waivers, including for the Advanced Clean Cars II program, and California and allied states sued to restore them.

Can other states adopt California Clean Air Act standards?

Yes. Section 177, 42 U.S.C. 7507, permits states that have nonattainment plan provisions to adopt and enforce California’s motor vehicle emission standards in place of the federal standards, provided the state standards are identical to California’s and California has received a section 209(b) waiver for them. Adoption is all-or-nothing: a state may not blend California and federal standards or write its own third set, because section 209(a) otherwise preempts state vehicle standards entirely. As of 2024 more than a dozen states, including New York, Colorado, and Washington, had adopted California’s standards. The practical consequence is a two-market regulatory landscape for automakers. When EPA rescinded California waivers in 2025, opt-in states faced legal uncertainty over their section 177 programs, and several joined litigation defending them.

What happens when EPA rejects a Clean Air Act state implementation plan?

When EPA disapproves a SIP under section 110, 42 U.S.C. 7410, two legal consequences follow. First, EPA incurs a mandatory duty to promulgate a federal implementation plan under section 110(c) within two years of the disapproval, covering the deficient portion of the plan. Second, the sanctions clock under section 179, 42 U.S.C. 7509, begins: if the state does not correct the deficiency within 18 months, mandatory sanctions take effect, starting with 2-to-1 emission offsets for new major sources, followed by potential restrictions on federal highway funding. A partial disapproval limits both the FIP and the sanctions to the disapproved portion, so states often rush to cure the deficiency before the 18-month mark. For example, after EPA disapproved many states’ 2015 ozone good-neighbor SIPs in February 2023, it issued the Good Neighbor Plan FIP in June 2023.

How long does a Clean Air Act rulemaking take?

Timelines vary widely. Routine technical corrections can move from proposal to final rule within a year, but major national standards commonly take three to five years, and some have taken far longer. Section 307(d) adds steps beyond the ordinary APA process: an opportunity for oral hearings, an expanded public docket, and a duty to answer every significant comment. Economically significant rules undergo Office of Information and Regulatory Affairs review under Executive Order 12866, which typically adds several months. Court orders and consent decrees often set the real schedule: the air toxics standards required by the 1990 Amendments ran years behind their statutory deadlines, and EPA missed the deadline for the 2015 ozone NAAQS review. Once a statutory deadline passes, section 304(a)(2) lets citizens sue to compel EPA to act, which is why missed deadlines routinely produce deadline lawsuits.

What is a Clean Air Act FIP and when is one imposed?

A federal implementation plan is EPA’s substitute for a missing or inadequate state plan, authorized by section 110(c), 42 U.S.C. 7410(c). EPA must impose a FIP when a state fails to submit a required SIP by the deadline, when EPA disapproves a SIP in whole or in part, or when EPA finds a state is not implementing its approved SIP. The agency has two years from the triggering finding to promulgate the FIP, which can contain emission limits, compliance schedules, and monitoring requirements. The FIP remains in force until EPA approves a state plan that corrects the deficiency, at which point the approved SIP replaces it. The most prominent example is the Good Neighbor Plan FIP of June 2023, which set emission budgets for power plants and industrial sources in states whose interstate-transport SIPs EPA had disapproved; the Supreme Court stayed its enforcement in June 2024.

How are Clean Air Act violations enforced?

Enforcement runs through three channels. Federal action: EPA may issue administrative compliance orders and assess administrative penalties, or refer cases to the Department of Justice for civil judicial enforcement under section 113, 42 U.S.C. 7413, seeking injunctive relief and civil penalties of up to statutory daily amounts adjusted for inflation. Knowing violations can be prosecuted criminally, with enhanced penalties for knowing endangerment. State action: states with delegated programs enforce their own SIPs, permits, and regulations, while EPA retains authority to step in when a state fails to act, a practice known as overfiling. Citizen action: section 304, 42 U.S.C. 7604, allows any person to sue violators of emission standards or EPA orders after giving 60 days’ notice to EPA, the state, and the violator. Settlements frequently take the form of consent decrees with compliance schedules.

Where and when must a Clean Air Act rule be challenged in court?

Section 307(b), 42 U.S.C. 7607(b), sets the venue and deadline. Petitions for review of nationally applicable rules must be filed in the United States Court of Appeals for the District of Columbia Circuit within 60 days after the rule’s publication in the Federal Register; actions that are locally or regionally applicable go to the court of appeals for the appropriate circuit. The 60-day deadline is strict and jurisdictional: courts dismiss late petitions even when the challenger had no actual notice of the rule. A challenger may also seek administrative reconsideration under section 307(d)(7)(B) when it was impracticable to raise an objection during the comment period, but reconsideration does not extend the filing window except as the statute narrowly allows. The practical consequence is that regulated parties must monitor the Federal Register daily, because missing the window forfeits judicial review entirely.

What is the comment-or-forfeit rule in Clean Air Act litigation?

Section 307(d)(7)(B), 42 U.S.C. 7607(d)(7)(B), provides that only an objection raised with reasonable specificity during the public comment period may be raised on judicial review, unless the objection arose after the comment period closed and it was impracticable to raise it earlier, or EPA’s grounds for the final rule changed so materially that earlier comment was impracticable. Courts enforce this strictly: arguments first offered in a petition for review or at oral argument are forfeited. The practical lesson is that stakeholders must file detailed, well-documented comments before the deadline, identifying flaws in EPA’s data, modeling, or legal theory. Comment periods for major rules commonly run 60 days, sometimes 90. This rule shapes rulemaking strategy: trade associations and state attorneys general treat the comment docket as the first phase of litigation, filing thorough objections up front rather than saving them for court.

What is a Title V operating permit under the Clean Air Act?

Title V, 42 U.S.C. 7661 through 7661f, requires major stationary sources to obtain operating permits that consolidate all applicable Clean Air Act requirements into a single enforceable document. Title V creates no new emission limits; it compiles existing ones from new source performance standards, hazardous air pollutant standards, SIP provisions, and preconstruction permits. State permitting authorities issue the permits under EPA-approved programs and must charge fees sufficient to fund the permitting program. Every proposed Title V permit undergoes EPA review, and EPA must object within 45 days if the permit does not comply with the Act; citizens may petition EPA to object and may seek judicial review of EPA’s failure to object. The practical consequence is that Title V permits are the public’s window into a facility’s compliance obligations, and permit challenges are a major venue for enforcement.

How do PSD and nonattainment new source review differ?

Prevention of Significant Deterioration, Part C of Title I, applies in attainment areas: new major sources and major modifications must install best available control technology, demonstrate that the project will not exceed air quality increments, and perform air quality analyses. Nonattainment new source review, Part D of Title I, applies where an area violates a NAAQS: the source must install the lowest achievable emission rate, obtain emission offsets from existing sources at greater than one-to-one ratios, and certify compliance of all its facilities in the state. The lowest achievable emission rate is stricter than best available control technology because cost cannot justify weaker control, and offsets make nonattainment permitting more expensive. The practical consequence is that siting decisions turn on area designation, and a tightening of a NAAQS can convert a PSD project into a costlier nonattainment review project.

What is the difference between EPA guidance and a binding Clean Air Act rule?

Guidance documents, policy memos, and question-and-answer documents explain EPA’s interpretation of the statute, but they do not create legally binding obligations. Only statutes, rules adopted through notice-and-comment rulemaking, permits, and orders impose enforceable duties. Courts have vacated agency actions that imposed binding requirements through guidance while skipping rulemaking procedures, and regulated parties cannot be penalized for violating a guidance document alone. That said, guidance shows how EPA intends to exercise its discretion, so companies and states generally conform to it. The practical risk arises when policy changes arrive by memo rather than by rule, as several 2025 Clean Air Act policy shifts did: challengers argue the agency bypassed the procedures section 307(d) requires, and a court may treat the memo as an invalid legislative rule.

What is the administrative record in a Clean Air Act rulemaking and why does it matter?

Section 307(d), 42 U.S.C. 7607(d), requires EPA to maintain a docket, the administrative record, containing the proposed and final rules, all written comments, hearing transcripts, EPA’s responses to significant comments, and the data and studies on which EPA relied. On judicial review, courts decide the case on this record under the arbitrary-and-capricious standard, and EPA may not defend a rule with post hoc rationalizations offered only in litigation, a principle from Motor Vehicle Manufacturers Association v. State Farm (1983). The practical consequence is that gaps in the record, such as unexplained modeling choices or unanswered significant comments, are among the most common grounds for vacatur. For commenters, this means building the record: submitting technical data and critiques during the comment period so they become part of what the court reviews.

How do citizen suits work under the Clean Air Act?

Section 304, 42 U.S.C. 7604, authorizes any person to sue in three situations: against any person violating an emission standard, limitation, or EPA order; against the EPA Administrator for failure to perform a nondiscretionary duty, such as missing a statutory deadline; or against anyone constructing a major emitting facility without a required permit. The plaintiff must give 60 days’ notice to EPA, the state, and the alleged violator before filing; if EPA or the state diligently prosecutes the violation, the citizen suit is barred. Courts may award injunctive relief and civil penalties, which are payable to the United States Treasury rather than the plaintiff, plus attorney fees to prevailing parties. The practical consequence is that citizen suits are a major enforcement engine, and the 60-day notice period often prompts settlements before any complaint is filed.

What is section 177 of the Clean Air Act and which states can use it?

Section 177, 42 U.S.C. 7507, lets states that have nonattainment plan provisions adopt California’s motor vehicle emission standards instead of the federal standards, provided the standards are identical to California’s and California holds a section 209(b) waiver for them. A state may not modify the standards or blend them with federal ones; adoption is all-or-nothing, and any state may use the provision regardless of region as long as it has nonattainment plan provisions. As of 2024 more than a dozen states had adopted California’s low-emission vehicle or zero-emission vehicle standards. The practical consequence is that automakers must plan for two regulatory markets. When EPA rescinded California waivers in 2025, opt-in states faced uncertainty over whether their section 177 adoptions remained valid, and several joined litigation defending them.

How can EPA reverse or amend a final Clean Air Act rule?

EPA must use notice-and-comment rulemaking to reverse or materially amend a final rule; it cannot rescind a rule by memo or announcement. Under Federal Communications Commission v. Fox (2009), the agency must acknowledge the change, explain why the prior policy is being abandoned, and address serious reliance interests engendered by the old rule. A rescission receives the same arbitrary-and-capricious review as the original rule, so EPA must build a fresh record with data and analysis. The practical consequence is that reversals are slow and litigation-prone. The power-plant sequence illustrates it: the 2015 Clean Power Plan was stayed by the Supreme Court in 2016 and replaced by the 2019 Affordable Clean Energy rule, which the D.C. Circuit vacated in 2021, and EPA’s 2024 power-plant rules faced proposed repeal in 2025. Each reversal required a full new rulemaking.

What sanctions can EPA impose on a state that fails its Clean Air Act obligations?

Section 179, 42 U.S.C. 7509, starts a sanctions clock when EPA finds that a state failed to submit a required SIP, disapproves a SIP, or finds that a state failed to implement its SIP. Eighteen months after the finding, mandatory sanctions apply unless the deficiency is corrected: new or modified major sources must obtain emission offsets at a 2-to-1 ratio. If the deficiency persists, EPA must impose highway sanctions, withholding federal highway funds except for safety, transit, and certain other projects. Sanctions stop when EPA approves a SIP that corrects the deficiency. The practical consequence is that the clock gives states a strong incentive to fix deficiencies within 18 months, and the highway sanction threat has historically driven SIP submissions more effectively than any other enforcement tool.

What is the Clean Air Act “good neighbor” provision?

Section 110(a)(2)(D), 42 U.S.C. 7410(a)(2)(D), requires each state’s SIP to prohibit emissions that significantly contribute to nonattainment of a NAAQS in another state or interfere with another state’s maintenance of a NAAQS. When EPA disapproves good-neighbor SIPs, it must issue a FIP addressing the interstate transport; the June 2023 Good Neighbor Plan FIP imposed emission budgets on power plants and industrial sources across 23 states for the 2015 ozone NAAQS. Upwind states and industry challenged the plan, and the Supreme Court stayed its enforcement in June 2024. The practical consequence is that downwind states depend on upwind-state SIPs and EPA FIPs for attainment, making good-neighbor obligations one of the most heavily litigated areas of Clean Air Act implementation.

What information can EPA demand from sources under section 114?

Section 114(a), 42 U.S.C. 7414(a), authorizes EPA to require owners and operators of emission sources to keep records, install and use monitoring equipment, sample emissions, and submit reports and other information the Administrator reasonably requires. EPA routinely issues section 114 information collection requests to gather data for rulemakings, such as the 2022 request to oil and gas operators supporting methane rules. Failure to comply can trigger enforcement under section 113. Section 114(c) protects trade secrets from public disclosure, but emission data itself is generally not treated as confidential. The practical consequence is that a section 114 letter often signals an upcoming rulemaking or enforcement action, and recipients should treat compliance as mandatory rather than voluntary.