Opinion concurring in part and concurring in the judgment filed by Circuit Judge-ROGERS,
MILLETT, Circuit Judge:
This is the Case of the Incredible Shrinking Airline Seat. As many have no doubt noticed, aircraft seats and the spacing between them have been getting smaller and smaller, while American passengers have been growing in size. Paul Hudson and the Flyers Rights group became concerned that this sharp contraction in ’passenger seating space was endangering the safety, health, and comfort of airline passengers. So they petitioned the Federal Aviation Administration to promulgate rules governing size limitations for aircraft seats to ensure, among other things, that passengers can safely and quickly evacuate a plane in an emergency. The Administration denied the petition, asserting that seat spacing did not affect the safety or speed of passenger evacuations. [741]*741To support that conclusion, the Administration pointed to (at best) off-point studies and undisclosed tests using unknown parameters. That type of vaporous record will not do—the Administrative Procedure Act requires reasoned decisionmaking grounded in actual evidence. Accordingly, we grant the petition for review in part and remand to the Administration.
I
A
Congress has charged the Federal Aviation Administration with ensuring the safety and security of commercial airline passengers. See 49 U.S.C. §.§ 44701, 40101(d); see also Wallaesa v. Federal Aviation Admin., 824 F.3d 1071, 1079 (D.C. Cir. 2016). In fulfilling that role, the Administration has “‘plenary authority to [mjake and enforce safety regulations governing the design and operation of civil aircraft’ in order to ensure the ‘maximum possible safety.’ ” Bargmann v. Helms, 715 F.2d 638, 642 (D.C. Cir. 1983) (alteration in original) (quoting H.R. Rep, No. 2360, 85th Cong., 2d Sess. 2, 7 (1958)).
As relevant here, the Federal Aviation Act charges the Administration with “promotfing] safe flight of civil aircraft in air commerce by prescribing * * * minimum standards required in the interest of safety for * * * the design, material, construction, quality of work, and performance of aircraft,” as well as “regulations and minimum safety standards for other practices, methods, and procedure[s] * * * necessary for safety in air commerce[.]” 49 U.S.C. § 44701(a)(1), (5). When issuing such minimum safety standards and regulations, the Administration.must consider “the duty of an air carrier to provide service with the highest possible degree of safety in the public interest!)]” Id. § 44701(d)(1)(A). In addition, the Administration “shall consider the following matters, among others, as being in the public interest: ■ (1) assigning, maintaining, and enhancing safety and security as the highest priorities in air commerce!) and] (2) regulating air commerce in a way that best promotes, safety and fulfills national defense requirements.” Id. § 40101(d)(1), (2). The Administration thus has broad authority to . promulgate regulations “reasonably related to safety in flight.” Wallaesa, 824 F.3d at 1079 (internal quotation marks and citation omitted).
Members of the public may petition the Administration to promulgate, amend, or repeal regulations. See 49 U.S.C. § 106(f)(3)(A);' 14 C.F.R. § 11.61(a). Such a petition must include, among other things, the purpose of the proposed action, an “explanation of why [the] proposed action would be in the public interest,” and “[a]ny specific facts or circumstances that support” the proposed action. 14 C.F.R. § 11.71(a). Once it receives a petition, the Administration has six months to respond either “by dismissing such petition[], by informing the petitioner of an intention- to dismiss, or by issuing a notice of proposed rulemaking or advanced notice of proposed rulemaking.” 49 U.S.C. § 106(f)(3)(A); see 14 C.F.R. § 11.73(a), (e).
B
On August 26, 2015, Paul Hudson and the non-profit organization Flyers Rights Education Fund of which he is president (collectively, “Flyers- Rights”) petitioned the Administration to promulgate rules ■governing the minimum requirements for seat sizes and spacing on commercial passenger airlines. In its petition, Flyers Rights provided evidence that commercial airline seat and spacing dimensions have steadily decreased in size over the last several decades. The petition noted that economy-class “seat pitch”—the distance [742]*742between a point on one seat and the same point on the seat directly in front of it— has decreased from an average of 35 inches to 31 inches, and in some airplanes has fallen as low as 28 inches. Evidence in the petition further indicated that average seat width has narrowed from approximately 18.5 inches in the early-2000s to 17 inches in the early- to mid-2010s. The petition also noted that, since the 1960s, the average American flyer had grown steadily larger in both height and girth. Flyers Rights expressed concern that the decrease in seat size, coupled with the increase in passenger size, imperiled passengers’ health and safety by slowing emergency egress and by causing deep vein thrombosis (a potentially fatal condition involving blood clots in the legs), as well as “soreness, stiffness, [and] other joint and muscle problems.” Pet. for Rule-making 6.1
Accordingly, Flyers Rights asked the Administration to: promulgate regulations that would (i) “set[] maintenance standards and limit[] the extent of seat size changes [on commercial airlines] in order to ensure consumer safety, health, and comfort”; (ii) “plac[e] a moratorium on any further reductions in seat size, width, pitch, padding, and aisle width until a final rule is issued”; and (iii) “[a]ppoint an advisory committee or task force to assist and advise the [Administration] in proposing seat and passenger space rules and standards[.]” Pet. for Rulemaking 3.
On February 1, 2016, the Administration denied Flyers Rights’ petition for rulemak-ing. The Administration explained that, in addressing petitions for rulemaking, it weighs: “(1) [t]he immediacy of the safety or security concerns * * raise[d], (2) [t]he priority of other issues the [Administration] must deal with, and (3) [t]he resources we have available to address these issues.” Denial of Pet. for Rulemaking 1; see also 14 C.F.R. § 11.73(a). The Administration then concluded that Flyers Rights’ concerns did not warrant action because the issues raised “relate[d] to passenger health and comfort, and d[id] not raise an immediate safety or security concern.” Denial of Pet. for Rulemaking 2. The Administration reasoned that it already “require[s] full-scale evacuation demonstrations and analysis that set the limit for the maximum number of passengers for any given airplane model,” including for aircraft with “interior configurations that are more critical (less seat pitch and higher number of passengers) than most configurations operated by the airlines,” and that emergency egress tests “have been successfully conducted at 28- and 29-inch pitch[.]” Id. The Administration added that “[s]eat pitch alone does not determine the amount of space available between seats * * * [because] modern, thinner seats at lower seat pitch provide more space than older seats did at higher pitch.” Id. The Administration further noted that the medical concerns identified in the petition exist “irrespective of the seat pitch[.]” Id. With respect to Flyers Rights’ concerns about deep vein thrombosis, the Administration concluded that the condition was “rare”; it can occur with “any long-duration seated activity”; and its risks [743]*743are “the same for economy-class and business-class.” Id.
The Administration’s denial of the petition for rulemaking did not cite any studies or tests to corroborate its representations. Nor did it challenge Flyers Rights’ characterization of seat dimension decreases or passenger size increases.
Flyers Rights sent a follow-up letter to the Administration’s Director of the Aircraft Certification Service asking the Administration to “formally cite the studyfies) [it] * * * reified] on” in denying the petition. J.A. 173. In response, the Administration identified a series of its own reports on airplane emergency egress and links to medical websites that discussed deep vein thrombosis. The studies cited in the letter did not address the impact of smaller seat dimensions or increased passenger size on the ability of passengers to expeditiously leave their seats and reach the emergency exits.
Dissatisfied with the Administration’s unsubstantiated representations about matters of passenger health and safety, Flyers Rights timely petitioned this court for review.
II
We review the Administration’s actions to determine whether they were “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” Safe Extensions, Inc. v. Federal Aviation Admin., 509 F.3d 593, 604 (D.C. Cir. 2007) (quoting 5 U.S.C. § 706(2)(A)). Under that standard, we will reverse “only if the agency’s decision is not supported by substantial evidence, or the agency has made a clear error in judgment.” Id. (citation omitted). Upon review, we may “affirm, amend, modify, or set aside any part of the order and may order the * * * Adminis-tratfion] to conduct further proceedings.” 49 U.S.C. § 46110(c).
Because Flyers Rights challenges the Administration’s decision not to engage in rulemaking—the Administration’s inaction—our review is “extremely limited.” WildEarth Guardians v. EPA, 751 F.3d 649, 651 (D.C. Cir. 2014) (citation omitted); see Massachusetts v. EPA, 549 U.S. 497, 527, 127 S.Ct. 1438, 167 L.Ed.2d 248 (2007) (“narrow” review of agency decision not to act). That is because an agency has “broad discretion to choose how best to marshal its limited resources and personnel to carry out its delegated responsibilities.” Massachusetts, 549 U.S. at 527, 127 S.Ct. 1438; see also Defenders of Wildlife v. Gutierrez, 532 F.3d 913, 919 (D.C. Cir. 2008) (“[A]n agency’s refusal to institute rulemaking proceedings is at the high end of the range of levels of deference we give to agency action under our ‘arbitrary and capricious’ review.”) (internal quotation marks and citation omitted).
In reviewing such decisions, we ask “whether the agency employed reasoned decisionmaking in rejecting the petition,” Defenders of Wildlife, 532 F.3d at 919, and we will overturn the agency’s decision “only for compelling cause, such as plain error of law or a fundamental change in the factual premises previously considered by the agency,” WildEarth Guardians, 751 F.3d at 653 (internal quotation marks and citation omitted). Our review turns, more specifically, on whether the agency “adequately explained the facts and policy concerns it relied on and [whether] * * * those facts have some basis in the record.” Id. (alterations in original; citation omitted).
Ill
Flyers Rights challenges two aspects of the Administration’s denial of its petition for rulemaking: (1) its conclusion that cur[744]*744rent seat pitch and width, as well as passenger size, do not negatively impact emergency egress, and (2) its .¡denial of authority to consider matters related to passenger health and comfort. We agree with Flyers Rights that the Administration failed to provide a plausible evidentiary basis for concluding that decreased seat sizes combined with increased passenger sizes have no effect on emergency egress. But we disagree with Flyers Rights’ challenge to the Administration’s declination to regulate matters of physical comfort and routine health.
Emergency egress
Flyers Rights’ petition for rule-making reasonably identified a safety concern arising from the commercial airlines’ documented pattern of placing ever larger passengers in ever smaller seats with still less space between them. The petition explained why such seating constrictions could make it more difficult for passengers to quickly leave their seats and escape an aircraft in the event of an emergency. Specifically, the petition asserted that, in an emergency, decreased seat spacing would increase panic, delay access to the center aisle, and impede the escape of injured passengers. The petition also included .multiple comments from airline passengers expressing safety concerns. One commenter stated that current seat spacing made it “necessary to climb onto [her] seat to get out.” J.A.-167, Another commenter asserted that, given current seat spacing, “[i]n an emergency, there is no way we would have been able to get to an exit row in less than three or four minutest.]” J.A. 169.
The Administration has a broad mandate to protect and promote passenger safety. Ensuring that all passengers can rapidly evacuate an airplane is of central importance to that safety,mission. See 14 C.F.R. § 25.803(c) (requiring that aircraft with a capacity of more than forty-four passengers be capable of evacuation within ninety seconds, and that actual egress demonstrations be undertaken to ensure compliance with Administration regulations). The Administration does not dispute that. Accordingly, when the Administration responds to a petition for rulemaking that exposes a plausible life- and-death safety concern, the Administration must reasonably address that risk in its response.
The Administration failed that task here. In asserting that decreasing seat size and pitch had no effect on emergency egress, the Administration pointed to certain studies and demonstration tests. But the cited studies say nothing about and do not appear to control for seat pitch, width, or any other seat dimension. Nor do they address or control for how increased passenger size interacts with the current seat dimensions to affect emergency egress. Studies cannot corroborate or demonstrate something that they never mention or even indirectly address.
The Administration, argues that the omission of information about seat dimensions from the tests means that seat dimensions are categorically unimportant to emergency egress. That makes no sense. Tests generally require a limited number of variables to be workable and verifiable. The omission of other variables says nothing about such variables’ relevance to what is being tested; it says only that they w¿re not recorded, measured, or altered for that particular test. Take, for example, a study on tooth decay that only recorded participants’ sugar consumption. The study’s silence on the question of brushing and flossing would surely not imply that brushing qnd flossing have no effect on the risk of getting a cavity.
The Administration’s rationale also blinks reality. As a matter of basic physics, [745]*745at some point seat and passenger dimensions would become so squeezed as to impede the ability of passengers to extricate themselves from their seats and get .over to an aisle. The question is not whether seat dimensions matter, but when.
Indeed, an Administration study that addressed passenger size in a slightly different context actually corroborates Flyers Rights’ point. The study considered, among other things, the ability of wider passengers to pass through the emergency exit row and door. Importantly, this test found that increased passenger width had the greatest effect on exit speed of all the variables tested. See J.A. 89 (chart indicating “[w]aist [s]ize” had the largest “Relative Magnitude of [e]ffect[ ]” of the ten variables tested). Yet nowhere did the Administration explain why passenger size would impede progress through the relatively wide emergency exit rows, yet have no impact on passenger movement through the far more cramped (seat-pitch-decreased) seating'rows.2
The Administration also overlooks that its studies are outdated. They were conducted in the 2000s when, according to the petition, seat dimensions were larger. Agency reasoning, however, must adapt as the critical facts change. See American Horse Prot. Ass’n v. Lyng, 812 F.2d 1, 5 (D.C. Cir. 1987) (“[A]' refusal to initiate'a rulemaking' naturally sets off a special alert when a petition has sought modification of a rule on the basis of a radical change in its factual premises.”).
The Administration points out that evacuation tests must be run with the maximum allowable passenger occupancy for any given aircraft model. See 14 C.F.R. § 25.807(g) (regulating the number of pas-
sengers allowed in each specified aircraft model to promote emergency egress); id. § 25.803(c) (tests must be run with- maximum allowable occupancy). The problem for the Administration is that maximum occupancy is not an adequate proxy for cabin-seat. or passenger dimensions, Because planes commonly include different seating classes like first class,, business class, and economy plus, limiting the number of seats in an aircraft does not limit seat pitch and width in all of the seats, and especially in the ordinary economy-class seats. That means that economy-seating pitch could decrease to levels that could impede emergency egress, while the pitch and width in.the first class and business class seats would not. .
Finally, the Administration stated in its decision that emergency evacuation tests have been successfully run with seat dimensions as small as those being used by commercial, airlines.. The problem is that not one of those tests is in the record. So they provide no evident support for the Administration’s conclusion.
The Administration says they were omitted because the tests are “proprietary.” Administration’s- Br. 13. Of course, an agency may decline to include confidential business information in the public administrative record in certain narrow situations, as long as it discloses as much information’ publicly as it can. See MD Pharm., Inc. v. Drug Enforcement Admin., 133 F.3d 8, 13 (D.C. Cir. 1998) (upholding an agency’s decision not to include confidential business information in' the public record of a licensing hearing); cf. Mead Data Central, Inc. v. United States Dep’t of the Air Force, 566 F.2d 242, 260 [746]*746(D.C. Cir. 1977) (Under the Freedom of Information Act, “[i]t has long been a rule in this Circuit that non-exempt portions of a document must be disclosed unless they are inextricably intertwined with exempt portions.”).
The problem here is that the Administration has given no reasoned explanation for withholding the tests in their entirety, and it has declined to file them under seal or in redacted form. Yet the Administration explicitly relied on those missing studies in reaching its decision to deny the petition for rulemaking. See Denial of Pet. for Rulemaking 2 (“Full scale evacuation tests on widely used airplanes have been successfully conducted at 28- and 29-inch pitch[.]”); J.A. 178. And the Administration asks the court to trust those studies in reviewing the Administration’s decision. See Oral Arg. Tr. 29-33; Administration’s Br. 11-13.
But that is not how judicial review works. We cannot affirm the sufficiency of what we cannot see. “[A]n agency decision based on ‘reliable data reposing in the [agency’s] files’ ” but hidden from judicial view “simply cannot withstand scrutiny.” United States Lines, Inc. v. Federal Maritime Comm’n, 584 F.2d 519, 535 (D.C. Cir. 1978).
Indeed, we have long held that, when “the data relied on by [an agency] in reaching its decision is not included in the administrative record and is not disclosed to the court[,]” we cannot “determine whether the final agency decision reflects the rational outcome of the agency’s consideration of all relevant factors[.]” United States Lines, 584 F.2d at 533 (footnote omitted). Whatever deference we generally accord to administrative agencies, “we will not defer to a declaration of fact that is ‘capable of exact proof but is unsupported by any evidence.” McDonnell Douglas Corp. v. United States Dep’t of the Air Force, 375 F.3d 1182, 1190 n.4 (D.C. Cir. 2004) (citation omitted).3
The problems with the Administration’s position do not stop there. Even with respect to its unseen tests, the agency cannot say whether those tests accounted for increased passenger size, which is a critical component of the egress problem raised by Flyers Rights’ petition. When questioned at oral argument, counsel for the Administration was unaware whether such tests take into account larger passengers. See Oral Arg. Tr. 29, 33-34.
To be sure, the record needed to support an agency’s decision not to engage in rulemaking can be sparser than that needed to support rulemaking. Normally, it “need only include the petition for rule-making, comments pro and con where deemed appropriate, and the agency’s explanation of its decision to reject the petition.” WWHT, Inc. v. FCC, 656 F.2d 807, 818 (D.C. Cir. 1981).
But this case is different because the Administration admits it relied materially on information it has not disclosed, and the Administration has pointed this court to that information as a basis for affir-mance. Having invited the court into its record, the Administration cannot hide the evidentiary ball. Cf. CTS Corp. v. EPA, 759 F.3d 52, 64 (D.C. Cir. 2014) (“It is black-letter administrative law that in an [Ad[747]*747ministrative Procedure Act] ease, a reviewing court should have before it neither more nor less information than did the agency when it made its decision”) (alteration in original; internal quotation marks and citation omitted). To do otherwise would reduce judicial review to a rubber stamp.4
In short, when an agency denies a petition for rulemaking, the record can be slim, but it cannot be vacuous. Especially so when, as here, the petition identifies an important issue that falls smack-dab within the agency’s regulatory ambit. While we do not require much of the agency at this juncture, we do require something. And information critically relied upon by the agency that no one can see does not count. We accordingly remand to the Administration to adequately address the petition and the emergency egress concerns it raises. If the petition for rulemaking is again denied, the Administration must provide appropriate record support for its decision.5
Flyers Rights asks the court to go further and order the Administration to institute rulemaking. That we will not do. “Our cases make clear * * * that such a remedy is appropriate only ‘in the rarest and most compelling of circumstances.’” American Horse Prot., 812 F.2d at 7 (quoting WWHT, 656 F.2d at 818). Rather, remand is the presumptive remedy when the agency record is insufficient “to permit [the court] to engage in meaningful review.” See id. (internal quotation marks and citation omitted). Because the Administration claims to have access to the information that' would fully justify its denial of the petition for rulemaking, an order to engage in rulemaking is unwarranted at this point.
Health and comfort concerns
Flyers Rights also objects to the Administration’s failure to address its concerns regarding passenger health and comfort. More specifically, Flyers Rights’ petition worried that cramped seat conditions cause deep vein thrombosis, “soreness, stiffness, [and] other joint and muscle problems[.]” Pet. for Rulemaking at 6. The Administration rejected such concerns partly on the ground -that they “relate to passenger health and comfort, and do not raise an immediate safety or security concern.” Denial of Pet. for Rulemaking 2.
Flyers Rights argues that the Administration’s failure to consider matters of passenger health and comfort is a misinterpretation of its statutory authority, pointing to assorted statutory provisions that purportedly require consideration of “the availability of a variety of adequate, economic, efficient[,] and low-priced ser[748]*748vices” and the “develop[ment] and maintenance of] a sound regulatory system that is responsive to the needs of the public.” Flyers Rights’ Opening Br. 26 (emphasis omitted) (quoting 49 U.S.C. § 40101(a)(4), (7)). The problem for Flyers Rights is that the cited statutory provisions . apply only to the. Secretary of Transportation, not to the Administration. See 49 U.S.C. § 40101(a).
Flyers Rights also points out that “health” is a component of “safety”—a criterion the Administration without a doubt must consider under applicable statutory provisions. Flyers Rights’ Reply Br. 7-9; see Flyers Rights’ Opening Br. 26-27. See also 49 U.S.C. §§ 44701, 40101(d). We agree. We have held that the Administration’s statutory authority “embodies] a comprehensive scheme for the regulation of the safety aspect[s] of aviation[.]” Bargmann, 715 F.2d at 642 (internal quotation marks and citation omitted). That includes protecting passengers’ physical health in flight, even from harms that are not occasioned by the flight. Indeed, in Bargmann, we rejected the Administration’s position that its authority was confined to addressing only those health issues that were “caused or induced by flight.” Id. at 640 (emphasis and internal quotation marks omitted). We held instead that the Administration has the authority to regulate first aid kits for treating conditions that occur during the flight, whether or not those conditions are caused by flight conditions or operations. Id. at 642; see also Wallaesa, 824 F.3d at 1080 (reaffirming the Administration’s power to regulate “care for ill passengers”).
So there is no question that the Administration has the statutory authority to address at least some passenger health issues. See Wallaesa, 824 F.3d at 1079-1080. (Administration may regulate medical equipment to ensure “ ‘the personal safety of the stricken passengers’ and crew”) (citation omitted); Bargmann, 715 F.2d at 642-643 (“Not only are inflight medical emergencies of immediate concern to.the personal safety of the stricken passengers, but they may also be of concern to the safety of others.”) (citation omitted); 14 C.F.R. § 121, App. A (Administration regulation requiring “automated external [heart] defibrillator[s]” on passenger aircraft).6
The problem for Flyers Rights is that, in this case, the Administration acknowledged its authority to protect the health of passengers, stating that it would “continue to monitor seat designs and effects on safety and' health.” J.A. 175 (emphasis added). The Administration thus did not decline to regulate the types of circulatory harms identified by Flyers Rights because [749]*749it thought it could not address such matters. Rather, the Administration decided that it should not address those issues at this time,* making the very type of regulatory-effort and resource-allocation judgments that fall squarely within the agency’s province.
Specifically, with respect to the risk of deep vein thrombosis, the Administration cited evidence showing that it rarely occurs and, regardless, is not caused by seat size or spacing. See Denial of Pet. for Rulemaking 2; J.A. 176 (citing a study noting that guidelines issued by the American College of Physicians indicate that deep vein thrombosis is “extremely rare” and that risk of deep vein thrombosis is not any higher in economy class than business class) (citation omitted). Thus, the Administration reasonably declined to initiate rulemaking to assess Flyers Rights’ concerns about deep vein thrombosis.
Flyers Rights also noted passenger problems with “soreness, stiffness, [and] other joint and muscle problems” in its petition for rulemaking. Pet. for Rulemak-ing 6. Given that those conditions are commonplace, temporary,- and non-life-threatening discomforts, Flyers Rights’ petition failed to demonstrate that the Administration erred in declining to undertake immediate rulemaking.7
# * # * ⅜
We grant Flyers Rights’ petition for review in part, and remand to the Administration for a properly reasoned disposition of the petition’s safety concerns about the adverse impact of decreased seat dimensions and increased passenger size on aircraft emergency egress. We otherwise deny the petition for review.
So ordered,