FERC's backdated refunds, and Roundup preemption

For a decade the Federal Energy Regulatory Commission kept getting the same grid-rate dispute wrong, and the D.C. Circuit kept reversing it. When the agency finally fixed the rate in 2024, it backdated the correction eight years and ordered refunds far longer than the statute's fifteen-month cap appears to allow. On June 5, 2026, the court held that lawful, because a narrow exception lets FERC reach back to cure its own error once a court has reversed it.[1] The rest of the opinion is a clinic in threshold rules: one challenger loses on standing, the other wins on standing but loses on law-of-the-case.[1]

The setup

A few terms are worth defining first. A petition for review is how a party asks a federal appeals court to check an agency's order. To vacate is to wipe an order out; to remand is to send the matter back to the agency to redo. Relief is prospective if it changes rates going forward, retroactive if it reaches money already collected.

MISO is the nonprofit operator that runs the interstate grid from Louisiana to Manitoba. The transmission-owning utilities earn a regulated Return (return on equity) on their grid investments, baked into customer rates. In 2013 and again in 2015, customer groups complained that Return was too high. What followed was a ten-year saga: FERC set a new Return, got reversed by this same court in 2017 and again in 2022 for methodological errors, and on remand in 2024 tried once more.[1] The trouble is when the fix applied. FERC made its corrected Return effective as of September 28, 2016, the date of its first flawed opinion, and ordered refunds from then through its 2024 order: roughly eight years of retroactive refunds.[1]

The question

The Federal Power Act sets a tightly constrained scheme. Section 206 lets FERC fix an unjust rate, but its ratemaking is prospective — the new rate is "to be thereafter observed."[2] Section 206(b) carves a narrow exception: FERC may order refunds, but only for a fifteen-month window tied to the complaint; refunds outside that window are, in this court's prior words, forbidden.[2] Against that sits Section 309, a broad grant of remedial authority — which, the court is careful to say, cannot be used to supersede specific statutory strictures like the fifteen-month cap. So how does an eight-year refund survive a fifteen-month ceiling?

The reasoning, walked

Start with the standard of review, because the court splits it in two. On whether FERC exceeded its statutory authority, the court applies independent judgment to find the best reading of the statute, citing Loper Bright (the 2024 decision that ended reflexive deference to agencies on questions of law).[3] On FERC's technical methodology, it applies the deferential arbitrary-and-capricious standard. Same case, two intensities of scrutiny, sorted by the kind of question.

The load-bearing move comes next. The court concedes the transmission owners are generally right: Section 206 caps refunds at fifteen months, and Section 309 can't override that.[2] But it identifies a specific, precedent-backed exception. When FERC is remedying its own errors after being reversed in court, it may backdate the effective date of relief to cure the very defect the court identified — notwithstanding the general bar on retroactive ratemaking.[1] The justification is notice and symmetry: because a FERC rate is always subject to court review, regulated parties are on notice that a judicial reversal can force retroactive change, and without this corrective power a court's reversal would be toothless. Here the 2022 decision had vacated all four prior opinions, so FERC could backdate to the first flawed one.[1] Claim denied.

Then the case pivots from merits to thresholds — the questions a court must answer before it is even allowed to rule. The owners' second argument is that FERC let customers file serial complaints to stretch the window. The court never reaches the merits, because the owners lack Article III standing — the constitutional requirement of a real, personal, redressable injury.[1] Their first injury (the cost of fighting the second complaint) fails redressability: they asked only for a rule barring future serial complaints, which refunds no costs already spent. Their second injury (being dragged through serial complaints someday) fails imminence: a future harm must be "certainly impending," and mere possibility is not enough.[3] Claim dismissed — a different disposition from "denied," because the court is declining jurisdiction, not rejecting the argument.

The customers' own cross-petition (Louisiana's utility commission) inverts the standing story. It clearly has standing, because its customers pay the disputed rates, even though it established this cleanly only on reply; the court forgives the timing because the injury is "patently obvious and irrefutable." But its substantive claims die on law-of-the-case: an issue decided earlier in the same litigation stays decided. The 2022 panel had already resolved which Return governs and blessed FERC's statistical inputs, so those cannot be relitigated.[1] A final evidentiary argument fails because FERC may refuse new evidence offered too late, at rehearing. Petitions denied.

What turns on it

For the grid, the ruling strengthens FERC's hand. When a court reverses the Commission, FERC can reach back to the origin of its error and settle up across the entire span of the flawed orders — the fifteen-month cap notwithstanding.[1] For anyone litigating against an agency, the case is a compact reminder that how you lose matters: losing on standing (dismissed) differs from losing on the merits (denied), and threshold doctrines quietly dispose of strong-sounding arguments before the merits get air.

A Missouri jury found Monsanto liable for failing to warn that Roundup causes cancer, and awarded the plaintiff about $1.25 million. On June 25, 2026, the Supreme Court wiped the verdict out, 7–2. A federal pesticide statute forbids states from demanding label warnings "in addition to or different from" the ones the EPA approved, and the EPA had specifically weighed a cancer warning for glyphosate and declined to require one.[1] Justice Jackson, in dissent, argued the majority mistook a state duty that runs parallel to federal law for one that conflicts with it.[1]

The setup

The alien vocabulary here is preemption: under the Constitution's Supremacy Clause, valid federal law can displace conflicting state law. Express preemption is when a statute says so in its own text. Certiorari is the Supreme Court's discretionary decision to hear a case; to reverse and remand is to overturn the decision below and send it back.

Monsanto (now part of Bayer) makes Roundup, a glyphosate weedkiller. John Durnell sued under Missouri law, claiming the label should have carried a cancer warning; a jury agreed and awarded roughly $1.25 million on that failure-to-warn theory, and the Missouri Court of Appeals affirmed.[1] The Supreme Court granted certiorari and reversed 7–2. Justice Kavanaugh wrote for the Court, joined by the Chief Justice and Justices Thomas, Alito, Sotomayor, Kagan, and Barrett; Justice Thomas concurred; Justice Jackson dissented, joined by Justice Gorsuch.[1][4]

The question

The federal pesticide statute (FIFRA) contains an express preemption clause: states may not impose labeling "requirements" that are "in addition to or different from" those required under FIFRA.[2] EPA registers each pesticide and approves its label, and, crucial here, EPA repeatedly evaluated glyphosate, concluded it is "not likely to cause cancer," and did not require a cancer warning.[1] The precise question is whether a state failure-to-warn verdict imposes a labeling requirement "in addition to or different from" the federal one, or merely enforces the same federal standard, in which case it survives.

The reasoning, walked

The majority builds the holding in three steps. It first treats the jury verdict as a "requirement" at all: a common-law duty enforced by damages pressures a manufacturer to change its label just as a regulation would.[1] It then asks whether that requirement is "in addition to or different from" the federal one, and finds that it plainly is. EPA had weighed the cancer question, judged glyphosate not likely carcinogenic, and approved a label without a warning, so a Missouri verdict demanding one commands what federal law declined to command. From there the clause does its work, because the text forbids exactly this. To anchor the move, the majority leans on Riegel v. Medtronic (2008), where FDA premarket approval of a device imposed federal "requirements" that preempted state tort claims under a similarly worded clause.[3] Registration, on this view, is the pesticide analogue of premarket approval.

The subtle work is what the majority does with Bates v. Dow Agrosciences (2005), the precedent that had let some FIFRA failure-to-warn claims through. Bates said a state claim survives if it is genuinely parallel — enforcing a duty equivalent to FIFRA's own.[3] The majority draws its distinction here: a claim is parallel only when it mirrors a federal requirement, and there is no federal requirement to warn of cancer to mirror, because EPA went the other way. So the claim isn't parallel; it's additive.[1]

That is precisely where Justice Jackson's dissent attacks. FIFRA independently forbids "misbranding," which includes a label with inadequate warnings.[2] So, the dissent argues, a state failure-to-warn duty enforces the very same standard FIFRA already imposes, adequate warnings, and is therefore parallel and permitted under Bates. The dissent's sharpest point is a distinction the majority blurs: EPA declining to require a cancer warning is not the same as EPA prohibiting one. If Monsanto could satisfy both duties at once, there is no genuine conflict — and, Jackson writes, the majority's reading leaves Durnell without a remedy.[1] Whether the two duties truly conflict, or merely coexist, is the hinge the whole case swings on, and the Court resolves it by treating EPA's specific consideration-and-rejection of a warning as decisive.

What turns on it

The decision resolves a split among lower courts and hands pesticide makers a strong preemption shield: where EPA has specifically examined a risk and approved a label without that warning, state failure-to-warn suits over that risk are largely foreclosed.[5] But the shield is bounded. By its own logic it protects only label-based claims about risks EPA actually considered; design-defect, manufacturing-defect, and deceptive-advertising claims, and claims about risks EPA never weighed, remain open.[4] The practical upshot is an incentive to get EPA to examine each hazard in detail during registration — because the depth of that specific review is what determines how much the defense is worth.

Assessment

This is analysis of legal reasoning, not legal advice.

The pair rhymes on a single theme: both cases are about how much deference an expert agency's determination commands against someone who wants to override it. But they sit on different axes. In FERC, the question is whether the agency may reach backward to fix its own court-identified mistake; the court says yes, and it reads as generosity toward agency flexibility. In Durnell, the question is whether the agency's determination displaces an independent state-law decisionmaker — a jury; the Court says yes, and it reads as agency primacy over the states. One expands an agency's power over its own past; the other expands its power over other people's present. A careless reader would collapse them.

On which opinion reasons better, FERC is the cleaner of the two, because its hardest move rests on a genuine, narrow, precedent-marked exception, and it is candid that the transmission owners are generally correct before explaining why they still lose. It is unanimous and it earns the unanimity. Durnell is more contestable, and Jackson's dissent lands a real hit: the holding depends on reading "EPA didn't require a warning" as functionally equivalent to "federal law forbids the warning," and that equivalence is asserted more than proved. The counter, that FIFRA's own misbranding rule makes a state warning duty parallel rather than additive, is not obviously wrong, which is why the case drew a cross-ideological 7–2 with two dissenters who don't usually vote together.

What the pair shows about appellate reasoning is that the outcome often turns not on the sympathetic facts but on a small hinge of characterization: is a jury verdict a "requirement," is a future harm "imminent," is a state duty "parallel" or "additive." A systems engineer will recognize the pattern: the whole system's behavior is set by a single load-bearing definition, and everything else is downstream of how that one term gets resolved.