Arbitrary and Capricious Review Explained: Why Agencies Have to Give Reasons, and When the Reasons Aren’t Enough

In 1977, the Department of Transportation adopted a regulation requiring new cars to have passive restraints — protection that worked without the occupant doing anything. Manufacturers could comply using airbags or automatic seatbelts. The agency expected roughly 60% of new cars to have airbags and 40% to have automatic seatbelts. That’s not what happened. By 1981, manufacturers were planning to install automatic seatbelts in approximately 99% of new cars — and most of those belts could be disconnected and left that way permanently. The agency became concerned the rule wouldn’t produce the predicted safety benefits. In October 1981, before the requirements took effect, it rescinded the entire regulation. The government had wanted protection people didn’t have to remember to use. Manufacturers were giving them something they could easily turn off. The agency’s response was to walk the whole rule back. State Farm and other insurance companies challenged that decision. In 1983, the Supreme Court held the agency had acted arbitrarily and capriciously. The problem wasn’t that the agency had to require airbags or automatic seatbelts. It was that the agency hadn’t adequately explained why it abandoned the requirement. That’s today’s subject — the standard that polices the reasoning behind agency decisions, not just the evidence.

This is the fifth merits-side episode in Season 3’s judicial review arc, and the final one before the winter break. Last episode covered substantial evidence — whether an agency’s factual findings have enough support in the record. Even when the facts are adequately supported, an agency can still make an arbitrary and capricious decision. This episode covers the standard that asks not whether the agency had evidence, but what it did with the evidence. Congress put the standard in the APA but didn’t really explain what it means. Section 706 tells courts to set aside agency action that is “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” The rest has been developed by the Supreme Court over time — starting with Citizens to Preserve Overton Park v. Volpe, where the Court said the reviewing judge needed to determine whether the agency had considered the relevant factors and whether there had been a clear error of judgment, while emphasizing the judge couldn’t substitute their own judgment for the agency’s. That approach became known as “hard look review,” a phrase associated with Harold Leventhal of the D.C. Circuit. The agency is supposed to take a hard look at the problem. The court then takes a hard look at whether the agency did that. Which sounds neat until the court decides the agency didn’t look hard enough.

State Farm laid out four ways an agency action could be arbitrary and capricious. The agency could rely on factors Congress didn’t intend it to consider. It could entirely fail to consider an important aspect of the problem. It could offer an explanation that conflicted with the evidence before it. Or it could offer an explanation so implausible the court couldn’t attribute it to a genuine difference in judgment or agency expertise. Wrong factors, missing something important, contradicting the evidence, or giving an explanation that just doesn’t make sense. Notice what isn’t on that list — that the judge would have made a different decision. Disagreement with policy isn’t enough. The relationship with substantial evidence review is one of overlap, not identity. If the agency makes a factual finding that isn’t supported by substantial evidence, relying on the finding would also be arbitrary and capricious. The reverse isn’t necessarily true. An agency can have substantial evidence supporting its finding and still act arbitrarily and capriciously — the agency might have ignored an important part of the problem, failed to consider an obvious alternative, or offered an explanation that didn’t connect its factual findings to the decision it made. Substantial evidence addresses whether the facts have adequate support. Arbitrary and capricious review asks a broader question about whether the agency made a reasoned decision. The APA specifically requires substantial evidence review of factual findings in certain formal proceedings; other statutes like Social Security have their own substantial evidence requirements. Arbitrary and capricious review is the general requirement of reasoned agency action. The grounds aren’t mutually exclusive.

Back to the seatbelts. The agency’s concern — that automatic seatbelts wouldn’t work because people could disconnect them — was legitimate. But the regulation allowed compliance using automatic seatbelts or airbags. Nobody can disconnect an airbag by deciding not to buckle up. The agency’s explanation for abandoning automatic seatbelts didn’t explain why it was abandoning airbags. It went from requiring passive protection to requiring none without considering whether it should narrow the rule to just airbags. All nine justices agreed the agency had failed to consider that important alternative. The second part of the Court’s analysis was less clean. The agency had also argued it could no longer reliably predict a significant increase in restraint use because people could disconnect the belts. The Court pointed out an important difference — with a manual seatbelt, you have to do something to put it on. With an automatic seatbelt, you have to do something to take it off. The agency hadn’t adequately addressed whether that difference might produce higher usage, or explained why it rejected automatic belts that couldn’t readily be detached. Four justices declined to join that portion of the opinion. Now the Court was getting into a dispute about how drivers would actually behave. The distinction between identifying an omitted alternative and second-guessing the agency’s prediction became much less clear. But notice what the Court didn’t do. It didn’t order the agency to require airbags or decide which restraint system was best. It sent the decision back because the agency hadn’t adequately explained its choice. The agency could still reach the same result — but it would need to consider the relevant alternatives and provide a defensible explanation. A requirement to give reasons can change what an agency decides even if the court can’t dictate the ultimate policy.

The governing idea is reasoned decision making. An agency must examine the relevant information, consider important aspects of the problem, and explain the relationship between what it found and what it decided to do. Burlington Truck Lines (1962) described that last requirement — there must be “a rational connection between the facts found and the choice made.” Suppose an agency announces a new safety requirement, explaining that injuries have been steadily declining for 10 years, and then imposes a costly new requirement without explaining why it’s necessary. There might be a perfectly good reason — maybe injuries are declining but still far too common, maybe an existing safety measure is about to expire. The problem isn’t the regulation. The problem is that the agency hasn’t explained how its evidence supports the decision. The judge isn’t supposed to invent that explanation. The court needs to be able to see how the agency got from the evidence to the decision it made. The agency doesn’t have to explain every detail or anticipate every possible objection, and a court can uphold a decision when the agency’s reasoning is reasonably discernible. What the court can’t do is apply an entirely different justification the agency never adopted. The obligation to explain can improve decision making before anyone ever files a lawsuit. An agency preparing to eliminate a regulation whose lawyers ask “have we addressed the important alternatives raised by this decision in the record?” — and someone points out a middle option that preserves most of the benefits while eliminating much of the cost — may reject that option, but at least it has to confront it. Confronting it might change the decision.

An obvious concern: if the agency has already decided what it wants to do, can’t it just come up with a convincing explanation afterward? Department of Commerce v. New York (2019) addressed that directly. The Commerce Secretary had added a citizenship question to the 2020 census, stating the question was needed to help enforce the Voting Rights Act. Record evidence showed he’d begun pursuing the citizenship question before the Justice Department had requested it — and had encouraged the Justice Department to make that request. The stated explanation didn’t match how the decision actually came about. The Court concluded the explanation appeared “contrived.” It couldn’t accept an explanation that didn’t reflect the agency’s genuine rationale. The problem wasn’t that the government couldn’t ask for a citizenship question. The problem was that the explanation wasn’t genuine. Agencies can have political motivations or multiple reasons for acting, and courts generally aren’t authorized to investigate an official’s subjective motivations just because they think politics played a role. But when the record establishes that the stated explanation is contrived, that’s a different problem. The court isn’t really reviewing the agency’s reasons at that point — it’s reviewing a story created to defend a decision already made for different reasons.

Policy changes raise a different question. The agency in State Farm decided to require passive restraints and then decided not to. Should a reversal get more scrutiny than a first-time decision? FCC v. Fox Television (2009) answered partly no. The FCC had changed its approach to fleeting expletives on television — previously, isolated unscripted expletives generally wouldn’t trigger the same response as repeated or deliberate use, but the agency changed that policy. The Supreme Court rejected the idea that every policy change automatically receives a heightened standard of review. An agency doesn’t have to prove its new policy is better than the old policy. Elections can change regulatory priorities. But the agency does have to acknowledge it’s changing course and give good reasons — and some changes do require a more detailed explanation. Two situations in particular: when the new policy relies on factual findings that contradict those underlying the old policy, and when people have developed serious reliance interests based on the old policy. Fox itself found the FCC had cleared that bar. It had acknowledged the change and provided a reasonable explanation, and the agency prevailed on the arbitrary and capricious issue (constitutional questions remained for further proceedings).

The reliance-interest piece showed up in the DACA case, Regents, decided in 2020. That case came up earlier this season because the government argued ending DACA was an unreviewable enforcement decision and the Supreme Court disagreed. The deeper problem with the rescission was remarkably similar to State Farm. DACA had two components — forbearance from deportation and eligibility for benefits like work authorization. The administration’s explanation focused on the legal problems it saw with the benefits but never separately considered whether it could end those benefits while retaining forbearance. Just like the airbags, the government considered eliminating everything without considering whether it should preserve part of the program. The administration also failed to adequately consider the serious reliance interests of people who had organized their lives around DACA — hundreds of thousands of people who had pursued education, accepted jobs, started businesses, and made family decisions. That doesn’t mean reliance interests always prevent an agency from changing policy. They have to be confronted and weighed, not necessarily controlled for. The Court invalidated the rescission but left open the possibility of a different, lawfully explained decision. Later litigation raised a different question — whether the agency had statutory authority to establish the program in the first place — and in 2025, the Fifth Circuit concluded important parts of DACA exceeded that authority, though it treated the forbearance separately. An agency might have the power to end a program but fail to explain the decision adequately, or explain a program beautifully but lack the authority to create it. Different grounds for judicial review. Regents also involved an attempt by Secretary Nielsen to issue a later memorandum offering additional justifications — and the Supreme Court refused to treat those new reasons as the basis for the earlier rescission. That raises a Chenery question about whether a later explanation clarifies original reasoning or supplies an entirely new rationale. January’s episode picks that up.

The standard is still being applied to stop major agency actions. Judge Patti Saris’s December 2025 decision in New York v. Trump (D. Mass.) involved the administration’s direction to federal agencies to pause certain wind project approvals while reviewing federal wind policies. Several states and a clean energy organization challenged the implementation. The administrative record contained two documents — the president’s memorandum and Interior’s written order implementing it. Two documents for a decision affecting federal wind permitting across the country. That isn’t automatically too few, but it made clear a lot was missing. The agency conceded that the president’s direction was the sole factor it considered. The court concluded the agency hadn’t reasonably considered the relevant issues or adequately explained the decision, hadn’t properly addressed the reliance interests of states affected by the pause, and had separate legal problems with indefinitely suspending the permitting process. The agencies couldn’t simply say the president told them to do it. The president can set policy priorities, but when an agency takes reviewable action under authority delegated to it, the agency must comply with the legal requirements governing that action. The court vacated the order. The administration appealed in February 2026 and voluntarily dismissed the appeal that June, leaving the district court judgment in place. The agency isn’t necessarily prohibited from imposing a similar pause — it could be able to reach the same result after considering the relevant factors and giving an adequate explanation, assuming it has the statutory authority and a sufficient factual and legal basis. A court setting aside an inadequately explained decision doesn’t necessarily prohibit the agency from adopting that policy. But having to consider what it skipped might change the agency’s mind.

The deeper tension in the doctrine is how demanding courts should be. Fox formally rejected a heightened standard for every policy reversal but recognized that serious reliance interests can require fuller explanation — which protects people who made decisions based on the old policy, but can also make it harder for an elected administration to change direction. Justice Rehnquist raised the democratic point in State Farm — an election producing a change in administration could legitimately cause an agency to reassess its policies. Political priorities don’t eliminate statutory obligations, but the question is how closely judges should examine whether the agency has met them. Two 2024-25 decisions show the question is unresolved. In Ohio v. EPA (2024), the EPA had adopted a rule addressing air pollution crossing state lines. Challengers raised an important question about how the agency’s approach would work after several states were excluded from the program. The Supreme Court concluded the challengers were likely to succeed on their arbitrary and capricious claim because the agency hadn’t adequately addressed that issue, and stayed the rule’s enforcement against the applicants while litigation continued. That wasn’t a final merits decision, but the Court was willing to scrutinize the agency’s reasoning closely enough to grant a stay. In Seven County Infrastructure Coalition v. Eagle County (2025), on the scope of an environmental impact statement under NEPA, Justice Kavanaugh wrote for the Court and emphasized substantial deference to agency decisions, warning courts against micromanaging choices within a broad zone of reasonableness. One decision insists the agency explain more. The other tells judges not to demand so much. They involve different questions and aren’t necessarily inconsistent, and Seven County didn’t overrule State Farm — but together they illustrate the continuing question of how demanding judicial review should be.

When the Court says it isn’t substituting its judgment, that doesn’t mean judges aren’t exercising judgment at all. They’re deciding whether the agency has met the legal requirements for reasoned decision-making. That’s an exercise of judgment too. All nine justices agreed in State Farm that the agency had failed to consider keeping an airbag requirement. Four wouldn’t join the Court’s examination of the agency’s prediction about automatic seatbelts. The first was a missing alternative. The second got much closer to deciding whether the agency’s prediction made sense. Both were presented as problems with the agency’s reasoning, but the justices disagreed about how far the court should go in scrutinizing it. That’s the tension at the heart of hard look review. Courts must ensure that agencies make reasoned decisions without taking over the job of making policy. An agency can have facts that are adequately supported by evidence and still make an arbitrary and capricious decision — because factual support is only part of what reasoned decision-making requires. An agency also has to consider the important aspects of the problem, act within its legal authority, and explain how the record supports the choice it made. If the court decides the agency didn’t do that, it can send the decision back. The agency might still be able to reach the same result. The court is reviewing the agency’s decision making, not making the policy decision itself. January’s episode picks up with Chenery — when a court reviews an agency decision, does it have to evaluate the reasons the agency actually gave, or can it uphold the decision because there were other good reasons the agency could have given?

Listen Now

Apple Podcasts | Listen on our site

What We Cover

  • Motor Vehicle Manufacturers Association v. State Farm (1983) and the automatic-seatbelt case that produced the modern framework for arbitrary and capricious review
  • Why manufacturers responded to the 1977 passive-restraint regulation with detachable seatbelts rather than airbags
  • The agency’s 1981 rescission of the entire regulation and the insurance companies’ challenge
  • APA § 706’s bare instruction to set aside agency action that is “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law”
  • Citizens to Preserve Overton Park v. Volpe and the “relevant factors / clear error of judgment” framework
  • Harold Leventhal’s “hard look” formulation and the double-barreled structure (agency takes a hard look at the problem; court takes a hard look at whether the agency did)
  • The four State Farm grounds: wrong factors, missing important aspects, contradicting evidence, implausible explanations
  • Why disagreement with policy isn’t on the list
  • How substantial evidence and arbitrary and capricious review overlap but ask different questions
  • Why an agency can have substantial factual support and still act arbitrarily and capriciously
  • Why State Farm’s airbag problem was the unanimous part of the opinion and the seatbelt-prediction problem was not
  • Burlington Truck Lines (1962) and the requirement of a “rational connection between the facts found and the choice made”
  • Why a court can uphold a decision when the agency’s reasoning is “reasonably discernible” but can’t invent new reasons for the agency
  • Department of Commerce v. New York (2019) and the pretextual-explanation problem in the census citizenship question case
  • Why courts don’t investigate political motivations generally — but can set aside explanations that are “contrived”
  • FCC v. Fox Television (2009) and the rejection of a heightened standard for every policy change
  • Two situations where policy changes do require a fuller explanation: contradicting factual premises of the old policy, and serious reliance interests
  • Regents (2020) and the DACA rescission: the forbearance/benefits problem as a direct parallel to the airbag/seatbelt problem in State Farm
  • How reliance interests work in arbitrary and capricious review — something to be confronted and weighed, not something that automatically controls
  • The 2025 Fifth Circuit ruling on DACA’s underlying statutory authority and how it distinguishes separate grounds for review
  • Why the Nielsen memorandum’s additional justifications didn’t count — the Chenery problem the next episode picks up
  • Judge Patti Saris’s December 2025 ruling in New York v. Trump on the wind permitting pause
  • Why “the president told us to do it” isn’t a sufficient explanation when the agency is exercising delegated authority
  • Ohio v. EPA (2024) and the willingness to grant a stay based on arbitrary and capricious analysis
  • Seven County Infrastructure Coalition v. Eagle County (2025) and Justice Kavanaugh’s emphasis on substantial deference under NEPA
  • Why these two 2024-25 decisions aren’t necessarily inconsistent but illustrate the continuing question of how demanding judicial review should be
  • Justice Rehnquist’s democratic-accountability point in State Farm about elections and changes in regulatory direction
  • Why “not substituting our judgment” doesn’t mean the court isn’t exercising judgment

Full Rough Transcript

Gwen: Welcome to Administrative Remedies, because you can't fix what you don't understand. Brought to you in part by the University of Tulsa College of Law's Mabee Legal Information Center. I'm Gwendolyn Savitz, an associate professor here at TU.

Marc: And I'm Marc Roark, a professor at the College of Law. We'll be breaking down complex doctrines with real-life analogies and examples to demystify the world of administrative law for everyone trying to understand how government actually works.

Gwen: Agencies are the main way the federal government gets things done. It's not through Congress for reasons we'll be addressing over the course of the series. Marc, do you remember automatic seatbelts?

Marc: It has been a long time since I've seen an automatic seatbelt. When you would get into a car and the car turned on, the shoulder belt would slide across your chest and then you had to affix the lap belt yourself.

Gwen: Yes. What did you think of them?

Marc: They were kind of annoying. To be honest, you'd open the door and the belt would move back along the track and then slide forward again when you closed it. They didn't feel particularly safe in and of themselves.

Gwen: But you didn't have to remember to put the belt on or at least the shoulder belt part of it. Although, like you said, you generally still needed to manually buckle the lap belt.

Marc: Right. Although you could still unbuckle the shoulder belt, too, if you wanted to.

Gwen: That last part is what got the federal government into trouble. Back in 1977, the Department of Transportation adopted a regulation that required new cars to have passive restraints, something that would protect occupants without requiring them to take any action. And manufacturers could comply using either airbags or automatic seatbelts. And at the time, the agency expected roughly 60% of new cars to have airbags and 40% to have automatic seatbelts.

Marc: But that's not what happened.

Gwen: Not even close. By 1981, manufacturers were planning to install automatic seatbelts in approximately 99% of new cars. And most of those belts could be disconnected and left that way permanently.

Marc: So the government wanted protection people didn't have to remember to use, and manufacturers were giving them something they could easily turn off.

Gwen: Yes. The agency became concerned that the rule wouldn't produce the safety benefits it had predicted. So in October 1981, before the requirements took effect, it rescinded the entire regulation.

Marc: Okay, that sounds reasonable. The government expected airbags, got detachable seatbelts instead, and decided the rule wasn't going to work as intended.

Gwen: That was the agency's position, but State Farm and other insurance companies challenged that decision. And in 1983, the Supreme Court held that the agency had acted arbitrarily and capriciously.

Marc: Okay, why was there a decision that they acted arbitrarily and capriciously? It sounds like the agency was just responding to new information.

Gwen: That's what makes the case interesting. The court wasn't saying the agency had to require airbags or automatic seatbelts. It was saying the agency hadn't adequately explained why it abandoned the requirement. That brings us to today's question. Last week, we talked about substantial evidence, about whether an agency's factual findings have enough support in the record. But even when the facts are adequately supported, an agency can still make an arbitrary and capricious decision.

Marc: Because it's not just about whether the agency has evidence, it's about what it does with that evidence.

Gwen: Yeah. So how would a court decide whether the agency acted reasonably without substituting the judge's own policy preferences?

Marc: Especially when the agency's decision sounds pretty reasonable.

Gwen: That's why we're starting with Motor Vehicle Manufacturers Association v. State Farm.

Marc: Okay. So where does this arbitrary and capricious standard come from? Is that something Congress defined?

Gwen: Congress put the standard in the Administrative Procedure Act, but it didn't really explain what it means. Section 706 tells courts to set aside agency action that is arbitrary, capricious, an abuse of discretion or otherwise not in accordance with law.

Marc: That's it? Congress didn't say what makes a decision arbitrary or capricious?

Gwen: Not specifically. This has been developed by the Supreme Court over time. And we've already talked about Citizens to Preserve Overton Park v. Volpe. That's the case that involved the proposed highway through the park in Memphis. And the court said the reviewing judge needed to determine whether the agency had considered the relevant factors and whether there had been a clear error of judgment. But it also emphasized that the reviewing judge couldn't substitute their own judgment for the agency's.

Marc: Okay, look carefully at the agency's judgment, but don't substitute your own.

Gwen: Yeah, this approach would become known as the hard look review. And that phrase was associated with Harold Leventhal of the D.C. Circuit. He wrote about examining whether an agency had genuinely engaged with the important problems before it. But there's an interesting distinction in that phrase.

Marc: Okay, what is that?

Gwen: The agency is supposed to take a hard look at the problem, and the court then takes a hard look at whether the agency did that.

Marc: Which sounds neat until the court decides that the agency didn't look hard enough.

Gwen: Yes, that's where State Farm comes in. The Supreme Court identified four ways the agency action could be arbitrary and capricious. The agency could rely on factors Congress didn't intend to consider. It could entirely fail to consider an important aspect of the problem. It could offer an explanation that conflicted with the evidence before it. Or it could offer an explanation so implausible that the court can't attribute it to a genuine difference in judgment or agency expertise.

Marc: So wrong factors, missing something important, contradicting the evidence, or giving an explanation that just doesn't make sense.

Gwen: Yes. Notice what's not on that list.

Marc: That the judge would have made a different decision?

Gwen: Yes. Disagreement with policy isn't enough.

Marc: How does this relate to substantial evidence, which we discussed last week?

Gwen: They overlap considerably. If the agency makes a factual finding that isn't supported by substantial evidence, relying on the finding would also be arbitrary and capricious. But the reverse isn't necessarily true. An agency can have substantial evidence supporting its finding and still act arbitrarily and capriciously.

Marc: Because something else could be wrong with the decision.

Gwen: Yes. The agency might have ignored an important part of the problem, failed to consider an obvious alternative, or offered some sort of explanation that didn't connect its factual findings to the decision it made.

Marc: Substantial evidence addresses whether the facts have adequate support. Arbitrary and capricious review asks a broader question about whether the agency made a reasoned decision.

Gwen: Yes, and there's an important distinction about when the standards apply. The APA specifically requires substantial evidence review of factual findings in certain formal proceedings. Other statutes like Social Security have their own substantial evidence requirements, but arbitrary and capricious review is the general requirement of reasoned agency action. Those grounds for judicial review aren't mutually exclusive, and State Farm is a really good illustration of why having evidence isn't necessarily enough.

Marc: OK, back to the seatbelts.

Gwen: Yes. The agency's concern was that automatic seatbelts wouldn't work as anticipated because people could disconnect them. And that was a legitimate concern. But remember, the regulation allowed manufacturers to comply using either automatic seatbelts or airbags. But the agency decided to rescind the entire regulation.

Marc: And nobody can disconnect an airbag by deciding not to buckle up.

Gwen: Yes. The agency's explanation for abandoning automatic seatbelts didn't explain why it was abandoning airbags. The agency never adequately considered retaining the requirement for airbags alone.

Marc: It went from requiring passive protection to requiring none without considering whether it should narrow the rule to just airbags.

Gwen: Yes. This doesn't mean that agencies in general have to consider every imaginable alternative, but here airbags were already part of the regulation and the agency's reason for eliminating the requirement didn't address them. And the Supreme Court unanimously agreed that the agency had failed to consider that important alternative.

Marc: And that seems pretty straightforward. The agency's explanation didn't address the whole decision.

Gwen: It's the clearest part of the case. But the court also questioned the agency's reasoning about automatic seatbelts. The agency argued that because people could disconnect the belts, it could no longer reliably predict a significant increase in restraint use. And the court said that explanation overlooked an important difference between automatic and manual seatbelts.

Marc: With a manual seatbelt, you have to do something to put it on. With an automatic seatbelt, you have to do something to take it off.

Gwen: Yeah, the agency hadn't adequately addressed whether the difference might produce higher usage. And the court said the agency hadn't adequately explained why it rejected automatic belts that couldn't readily be detached. And this part of the opinion divided the court. Four justices declined to join the portion scrutinizing the agency's automatic seatbelt reasoning.

Marc: Because now the court was getting into a dispute about how drivers would actually behave.

Gwen: Yeah, the distinction between identifying an omitted alternative and second-guessing the agency's prediction became much less clear. But notice what the court didn't do. It didn't order the agency to require airbags or decide which restraint system was best. It sent the decision back because the agency hadn't adequately explained its choice.

Marc: So the agency could still reach the same result?

Gwen: Potentially, yes, but it would need to consider the relevant alternatives and provide a defensible explanation. And that matters. A requirement to give reasons can change what an agency decides, even if the court can't dictate the ultimate policy.

Marc: So is that the real requirement then? Explain yourself? Is that basically what the court is asking agencies to do?

Gwen: It's saying explain yourself, but not just in the sense of putting words on paper. The governing idea is reasoned decision making. An agency must examine the relevant information, consider important aspects of the problem, and explain the relationship between what it found and what it decided to do. And that last requirement was described in Burlington Truck Lines, a 1962 case, that there must be a rational connection between the facts found and the choice made.

Marc: Okay, give me another example.

Gwen: Let's suppose an agency announces a new safety requirement. It explains that injuries have been steadily declining for 10 years, and then it announces a costly new requirement without explaining why that requirement is necessary.

Marc: There might be a perfectly good reason for the new requirement. Maybe injuries are declining, but still far too common.

Gwen: Or maybe an existing safety measure is about to expire. The problem isn't necessarily the regulation. The problem is that the agency hasn't explained how its evidence supports the decision.

Marc: And the judge isn't supposed to invent that explanation for it.

Gwen: Right. The court needs to be able to see how the agency got from the evidence to the decision it made. But this doesn't mean the agency has to explain every detail or anticipate every possible objection. But there does have to be a rational connection between the facts and the choice.

Marc: So if the connection isn't there, the court can send the decision back to the agency.

Gwen: Right. The agency has to do the reasoning. The court can't do it for the agency. But it doesn't have to be perfectly written. A court can uphold a decision when the agency's reasoning is reasonably discernible. What the court can't do is apply an entirely different justification that the agency never adopted.

Marc: But there's something useful about making the agency explain itself before it acts.

Gwen: Sure, there is. Imagine an agency preparing to eliminate a regulation. Its lawyers ask, have we addressed the important alternatives raised by this decision in the record? And someone points out a middle option that would preserve most of the benefits while eliminating much of the cost. The agency might reject that option, but at least it has to confront it.

Marc: And confronting it might change the decision.

Gwen: Yes. The obligation to explain can improve decision making before anyone ever files a lawsuit.

Marc: But suppose the agency has already decided what it wants to do. Can't it just come up with a convincing explanation afterward?

Gwen: That is a real concern, and there's a Supreme Court case directly addressing it. We talked about it a couple of weeks ago. Department of Commerce v. New York, this was decided in 2019, where the court was reviewing the decision to add the citizenship question to the 2020 census. The Commerce Secretary had said that the question was needed to help enforce the Voting Rights Act, but there was record evidence showing that he'd begun pursuing the citizenship question before the Justice Department had requested it and had encouraged the Justice Department to make that request to begin with.

Marc: The agency's explanation just didn't match how the decision actually came about.

Gwen: Yeah. The court concluded that the stated explanation appeared contrived. It couldn't accept an explanation that didn't reflect the agency's genuine rationale.

Marc: The problem wasn't necessarily that the government couldn't ask for a citizenship question. It was that the explanation wasn't genuine.

Gwen: Yes, but that doesn't mean an agency can't have political motivations or multiple reasons for acting. And courts generally aren't authorized to investigate an official's subjective motivations just because they think politics played a role. But when the record establishes that the stated explanation is contrived, that's a different problem.

Marc: Because when the court isn't really reviewing the agency's reasons, it's reviewing a story created to defend the decision.

Gwen: Yes, and that's the point. The explanation has to reflect genuine reasoning, not merely offer a plausible defense of a decision already made for different reasons. And that distinction becomes especially important when an agency changes its policies.

Marc: Which is what happened in State Farm. The agency decided to require passive restraints and then decided not to. Should a reversal get more scrutiny than an agency making a decision for the first time?

Gwen: The Supreme Court looked at this in FCC v. Fox Television in 2009. The FCC had changed its approach to fleeting expletives on TV. Previously, an isolated, unscripted expletive generally wouldn't trigger the same response as repeated or deliberate use. But the agency changed that policy. The Supreme Court rejected the idea that every policy change automatically receives a heightened standard of review. An agency doesn't have to prove its new policy is better than the old policy.

Marc: Which makes sense. Elections can change regulatory priorities.

Gwen: Yes, but the agency does have to acknowledge it's changing course and give good reasons. And it's possible that a change would require a more detailed explanation.

Marc: OK, when are those instances?

Gwen: So there's two important situations. The first is when the new policy relies on factual findings that contradict those underlying the old policy. And the second is when people have developed some sort of serious reliance interest based on the old policy.

Marc: Okay, so what happened in Fox itself? Did the FCC win?

Gwen: It did. The Supreme Court concluded that the FCC had acknowledged the change and provided a reasonable explanation. And the constitutional questions remained for further proceedings, but the agency prevailed on the arbitrary and capricious issue.

Marc: So Fox isn't just a case telling agencies how to change policy. It's an example of where the agency actually did enough.

Gwen: Yes. So that brings us to the other side of the question. What happens when an agency changes a policy that people have relied on? Remember Regents, the DACA case we discussed earlier this season?

Marc: Yeah, we talked about it because the government argued that ending DACA was an unreviewable enforcement decision and the Supreme Court disagreed.

Gwen: Yes, but we didn't get into why the court ultimately set aside the rescission. And that problem was remarkably similar to the one in State Farm. DACA had two components. There was forbearance from deportation and eligibility for benefits like work authorization. The administration's explanation for ending DACA focused on the legal problems it saw with the benefits, but it never separately considered whether it could end those benefits while retaining forbearance.

Marc: Okay, just like the airbags, the government considered eliminating everything without considering whether it should preserve part of the program.

Gwen: Yes, and that wasn't the only problem. The administration also failed to adequately consider the serious reliance interests of people who'd organized their lives around DACA.

Marc: Okay, hundreds of thousands of people had organized their lives around DACA. People pursued education, accepted jobs, started businesses, and made decisions about their families. And the administration had not adequately addressed those reliance interests in deciding to end the program.

Gwen: But that doesn't mean that reliance interests always prevent an agency from changing policy. There's something that the agency has to confront and weigh, not necessarily something that controls the outcome. The court invalidated the attempt to rescind DACA, but it left open the possibility of a different, lawfully explained decision.

Marc: OK, did that settle DACA's future?

Gwen: No. Later litigation raised a different question, whether the agency had statutory authority to establish the program in the first place. And in 2025, the Fifth Circuit concluded that important parts of DACA exceeded the authority, but it did treat the forbearance separately.

Marc: So an agency might have the power to end a program, but fail to explain the decision adequately, and it might explain a program beautifully, but lack the authority to create it.

Gwen: Yes, those are two different grounds for judicial review.

Marc: But there's another issue in Regents in there. The administration tried to give more reasons for ending DACA after the original decision.

Gwen: It did. Secretary Nielsen issued a later memorandum offering additional justifications, and the Supreme Court refused to treat those new reasons as the basis for the earlier rescission. That raises a Chenery question about whether a later explanation clarifies an agency's original reasoning as opposed to when it supplies an entirely new rationale. But we'll hit that in the next episode.

Marc: OK, we've talked about the Supreme Court setting aside major agency decisions. Give me a more recent example of the standard actually stopping an agency.

Gwen: All right. We can look at Judge Patti Saris's December 2025 decision in New York v. Trump. This was the federal district court in Massachusetts. The administration had directed federal agencies to pause certain wind project approvals while reviewing federal wind policies. Several states and a clean energy organization challenged the agency's implementation of that pause.

Marc: And what was in the administrative record?

Gwen: There were two documents, the president's memorandum and Interior's written order implementing it.

Marc: Two documents for a decision affecting federal wind permitting across the country. Is that automatically too few?

Gwen: No, but it made clear that a lot was missing. The agency conceded that the president's direction was the sole factor they considered. And the court concluded that it had not reasonably considered the relevant issues or adequately explained the decision. It also found that the agency hadn't properly addressed the reliance interests of states affected by the pause. And it identified separate legal problems with indefinitely suspending the permitting process.

Marc: The agencies couldn't simply say the president told us to do it.

Gwen: Right. The president can set policy priorities, but when an agency takes reviewable action under authority delegated to it, the agency must comply with the legal requirements governing that action. So the court vacated the order. The administration appealed in February 2026 and then voluntarily dismissed the appeal that June, leaving the district court judgment in place.

Marc: OK, what happens next then? Is the agency prohibited from ever imposing a similar pause?

Gwen: Not necessarily. It could be able to reach the same result after considering the relevant factors and giving an adequate explanation, assuming that it has the statutory authority and a sufficient factual and legal basis. A court setting aside an inadequately explained decision doesn't necessarily prohibit the agency from adopting that policy.

Marc: But having to consider what it skipped might change the agency's mind.

Gwen: It might, and that's why the requirement isn't just paperwork. Sometimes a court's decision forces the agency to confront evidence or alternatives that change its policy. Sometimes the agency reconsiders and just reaches the same result.

Marc: And that's where remedies become important.

Gwen: Yes, and we'll go back to those later this season, too. But there's another issue we need to address first. Even if everyone agrees the agencies must give reasons, how demanding should courts be when they review the reasons?

Marc: Okay, when an agency creates a rule, it has to explain the decision. But once the rule exists, people might rely on it. So repealing it requires the agency to address those reliance interests too. This sort of seems to push in different directions if we look at different kinds of cases. So for example, in the DACA case, the reliance interest is about the economic opportunities that people would come to rely on. The State Farm case is going to save industry money. What is it about these two cases and the difference in reliance that end up in the same place? I mean, doesn't that make repeal harder than adoption?

Gwen: It can. Fox formally rejected a heightened standard for every policy reversal, but it also recognized that serious reliance interests can require a fuller explanation.

Marc: Which protects people who made decisions based on the old policy, but it can also make it harder for an elected administration to change direction.

Gwen: And both concerns are legitimate. Rehnquist raised the democratic point in State Farm. He argued that an election producing a change in administration could legitimately cause an agency to reassess its policies. But political priorities don't eliminate statutory obligations. The question is how closely judges should examine whether the agency has met them.

Marc: Is the Supreme Court moving toward greater deference?

Gwen: It's not that simple. Let's look at a couple of recent cases. Ohio v. EPA from 2024. The EPA had adopted a rule addressing air pollution crossing state lines. And challengers raised an important question about how the agency's approach would work after several states were excluded from the program. And the Supreme Court concluded that the challengers were likely to succeed on their arbitrary and capricious claim because the agency hadn't adequately addressed that issue. It stayed the rule's enforcement against the applicants while the litigation continued. That wasn't a final merits decision, but the court was willing to scrutinize the agency's reasoning closely enough to grant a stay. Now compare that with Seven County Infrastructure Coalition v. Eagle County decided in 2025. That case concerned the scope of an environmental impact statement under the National Environmental Policy Act. Kavanaugh wrote for the court and emphasized substantial deference to agency decisions, warning courts against micromanaging choices within a broad zone of reasonableness.

Marc: So one decision insists the agency explain more, while the other tells judges not to demand so much.

Gwen: Right. And they do involve different questions. They're not necessarily inconsistent. And Seven County didn't overrule State Farm, but together they illustrate the continuing question of how demanding judicial review should be.

Marc: There's a parallel here in eminent domain. In Kelo, the Supreme Court deferred to a city's development judgment and its comprehensive development plan mattered. But a plan isn't a magic word. If a taking is really just a pretext to benefit a favored private party, writing a detailed plan doesn't make it constitutional.

Gwen: And an agency's explanation works the same way. It can demonstrate genuine reasoning, but it can't substitute for it.

Marc: So when the court says it isn't substituting its judgment, we shouldn't imagine judges aren't exercising judgment at all.

Gwen: Exactly. They're deciding whether the agency has met the legal requirements for reasoned decision-making. Which brings us back to State Farm. All nine justices agreed that the agency had failed to consider keeping an airbag requirement. But four wouldn't join the court's examination of the agency's prediction about automatic seatbelts.

Marc: The first was a missing alternative. The second got much closer to deciding whether the agency's prediction made sense.

Gwen: Yes, both were presented as problems with the agency's reasoning. But the justices disagreed about how far the court should go in scrutinizing that reasoning. That's the tension at the heart of this hard look review. Courts must ensure that agencies make reasoned decisions without taking over the job of making policy.

Marc: Okay. An agency can have facts that are adequately supported by evidence and still make an arbitrary and capricious decision.

Gwen: Yes, because factual support is only part of what reasoned decision-making requires. An agency also has to consider the important aspects of the problem, act within its legal authority, and explain how the record supports the choice it made.

Marc: And if the court decides the agency didn't do that, it can send the decision back. But the agency might still be able to reach the same result.

Gwen: The court is reviewing the agency's decision making, not making the policy decision itself.

Marc: And that's where we're going to leave things for a little while.

Gwen: Yes, we are already in finals. So we're going to pause to deal with all of that. But we will be back in January with more administrative remedies.

Marc: And what are we starting with?

Gwen: A question we've touched on a few times so far. When a court reviews an agency decision, does it have to evaluate the reasons the agency actually gave? Or can it uphold the decision because there were other good reasons the agency could have given? That's the question at the heart of Chenery, and we will pick up there in January. So that does it for today's episode on administrative remedies. Thank you for joining us today. Please, if you enjoy this podcast and enjoy this episode, give us a like on Spotify, iTunes, or whatever platform you're listening on. And be sure to tune in next time where we'll continue to dive into the contours of administrative law, because remember, you can't fix what you don't understand.

Related Guides