Page 2Petitions for Review of a Decision of the Federal Railroad Administration Agency No. FRA-2021-0032

No. 24-11300

Petitions for Review of a Decision of the Federal Railroad Administration Agency No. FRA-2021-0032

No. 24-11366

ASSOCIATION OF AMERICAN RAILROADS,

Page 3Petitions for Review of a Decision of the Federal Railroad Administration Agency No. FRA-2021-0032

No. 24-11367

Page 4Respondents.

Petitions for Review of a Decision of the Federal Railroad Administration Agency No. FRA-2021-0032

No. 24-11428

Petitions for Review of a Decision of the Federal Railroad Administration Agency No. FRA-2021-0032

Page 5No. 24-11444

Petitions for Review of a Decision of the Federal Railroad Administration Agency No. FRA-2021-0032

No. 24-11445

NEBRASKA CENTRAL RAILROAD COMPANY, Petitioner, versus

Page 6Petitions for Review of a Decision of the Federal Railroad Administration Agency No. FRA-2021-0032

No. 24-12003

Page 7Petitions for Review of a Decision of the Federal Railroad Administration Agency No. FRA-2021-0032

Before ROSENBAUM, BRANCH, and KIDD, Circuit Judges.

ROSENBAUM, Circuit Judge:

It’s hard to tell the United States’s story without dedicating a long chapter to the railroad. Back in 1869, the Transcontinental Railroad tied together a disparate country in a new way . Today , rail continues to play a critical role in transporting people and goods.

But like all forms of transportation, trains have posed some safety concerns from the beginning. The ' Angola Horror' of December 1867 drove a wave of safety reforms after some 50 people died in a train derailment and fire in New York.¹ And on occasion, unfortunately, train accidents continue to happen today . 2

Still, technology has helped. Yet as technology has evolved, railroads have moved towards using smaller and smaller train crews. So in recent years, the Federal Railroad AdministrationPage 8 ('FRA') has fielded concerns that some railroads' shift towards oneperson train crews could pose safety risks.

In 2024, the FRA issued the regulation that sparked this case. See 89 Fed. Reg. 25052 (Apr. 9, 2024). That regulation, the 'Crew Size Rule” or “Rule,” requires railroads to operate with at least two crewmembers on board unless using a one-person crew would be at least as safe.

Some parts of the rail industry dislike this new requirement. Six railroads and two industry trade organizations (“Petitioners” or 'Railroads') filed petitions for review of the FRA's rulemaking. We consolidated those petitions in this Court.

The Railroads assert that the FRA violated the Administrative Procedure Act by exceeding its statutory authority and acting arbitrarily and capriciously in several ways. One group of petitioners, representing the interests of larger railroads, raises five issues. A second group, representing smaller railroads, raises another two.

We address these concerns below. But because none of the Railroads’ arguments have merit, we deny their petitions.

The Secretary of Transportation holds broad power to issue regulations to promote railroad safety. The Federal Railroad Safety Act of 1970, as amended, authorizes that power. It provides, “The Secretary of Transportation, as necessary, shall prescribe Page 9regulations and issue orders for every area of railroad safety.” 49 U.S.C. § 20103(a). The Secretary delegated this authority to the Administrator of the FRA, an agency within the U.S. Department of Transportation. 49 C.F.R. § 1.89(a).

Under that delegation, the FRA has issued a wide range of safety regulations. These rules address, for example, track inspections, Id. § 213.233; qualification of locomotive engineers and train conductors, id. pts. 240, 242; locomotives' braking systems, id. § 229.13; random drug and alcohol testing of railroad employees, id. § 219.603; and track switches, id. § 213.135.

Technological changes have allowed railroads to shrink their average crew size over time. The number of crewmembers on each train has dropped from around five in the 1960s to two in the 1990s. See 87 Fed. Reg. 45564, 45567 col.1 (July 28, 2022).

Today, most railroad operations use at least a two-person crew, consisting of a locomotive engineer and a conductor. Id. at 45578 col.1. The locomotive engineer performs nearly all actions that control the movement of a train. Id. At the same time, the conductor leads the train crew as a whole, managing the train consist (the cars that make up the train), interacting with dispatchers and others outside the train cab, and dealing with unusual situations like mechanical problems. Id. at 45567 col.3–45568 col.1.

In recent years, though, more railroads have started operating with a one-person crew. See id. at 45578 col.2.

Page 10Before the Crew Size Rule, no regulation specifically regulated minimum crew size. Yet the seeds for that regulation were planted more than a decade earlier.

Regulatory attention to one-person crews spiked after a catastrophic accident in Quebec, Canada, in July 2013. A train with a one-person crew saw numerous crude-oil tank cars derail at the town of Lac-Mégantic. See 87 Fed. Reg. at 45568 col.3-45569 col.1. The ensuing explosions and fires caused 47 fatalities and required 2,000 people to evacuate the surrounding areas. See id.

The FRA wasted no time trying to prevent something similar from happening here. The next month, in August 2013, the FRA directed the Railroad Safety Advisory Committee to “review whether train crew staffing practices affect railroad safety.” See 81 Fed. Reg. 13918, 13936 col.2 (Mar. 15, 2016). But while that committee conducted its review, another major accident occurred, this time in Casselton, North Dakota. There, a multimember crew took “heroic actions” to prevent a crude-oil fire from spiraling out of control after a crude-oil train collided with a derailed 'grain train.” Id. at 13923 col.2-13924 col.3. The committee's working group did not reach consensus on recommendations by the April 1, 2014, deadline. Id. at 13939 col.1.

Two years later, in March 2016, the FRA started a rulemaking proceeding to regulate crew size. Id. at 13918 col.1. The notice of proposed regulation (“2016 Notice”) described a rule that would have generally required trains to operate with at least two crewmembers, with exceptions for “operations that FRA believes do not Page 11pose significant safety risks . . . by using fewer than two-person crews.” Id. The 2016 Notice contemplated exceptions for most railroads that had already used one-person crews since before 2015. Id. at 13964 col.2–13966 col.3. It also considered options for allowing railroads to start new one-person operations after submitting certain information to the FRA, with agency approval required in some instances. See id.

In 2019, the FRA withdrew the 2016 Notice. 84 Fed. Reg. 24735 (May 29, 2019) (“Withdrawal Order”). The agency explained its view that a crew-size rule was unwarranted because “FRA’s accident/incident safety data does not establish that one-person operations are less safe than multi-person train crews.” Id. at 24739 col.2. Besides declining to issue a crew-size regulation, the Withdrawal Order also purported to 'negatively preempt any state laws' requiring a minimum number of crewmembers. Id. at 24741 col.3.

In 2021, the U.S. Court of Appeals for the Ninth Circuit vacated the Withdrawal Order and remanded to the FRA. Transp. Div. of the Int’l Ass’n of Sheet Metal, Air, Rail, & Transp. Workers v. Fed. R.R. Admin. ( SMART-TD I ), 988 F.3d 1170 (9th Cir. 2021). The court reasoned that the Withdrawal Order was, in effect, a nationwide one-person crew rule. Id. at 1182. Among other flaws, the court held that a one-person crew-size mandate was not a “logical outgrowth” of the initial proposal. Id. at 1180-81. The court also questioned the Withdrawal Order’s reasoning. Id. at 1182. As the Ninth Circuit saw things, the FRA “fail[ed] to address the multiplePage 12 safety concerns” that the majority of commenters raised and offered “no safety or economic justification for preemption” of state minimum-crew-size rules. Id. at 1183-84.

In July 2022, the FRA published a notice of proposed rulemaking (“2022 Notice”) on a rule that would regulate crew size. 87 Fed. Reg. 45564 (July 28, 2022). During a 146-day comment period, the FRA received around 13,500 comments from commenters. Those commenters included labor organizations, individuals, railroads, trade organizations, members of Congress, and state and local elected officials. 89 Fed. Reg. at 25059 col.1–25075 col.2. More than 98.5 percent of the comments supported the proposed rule. See id. at 25059 col.2. The FRA also held a public hearing on the rule on December 14, 2022. See id.

In April 2024, the FRA finalized the Crew Size Rule in an order published in the Federal Register ('Final Rule'). See id. at 25052 col.1. The Crew Size Rule by default requires trains to operate with a minimum of two crewmembers onboard. 49 C.F.R. § 218.123(b). But the Rule has several exemptions.³ It also opens pathways, either by right or upon approval, for covered trains to Page 13begin one-person operations. How a railroad moves towards oneperson operations varies based on the size of the railroad and whether it hauls hazardous materials.⁴

The first, “notice” option is available to smaller, Class II and Class III freight railroads.⁵ These railroads may continue 'legacy' one-person operations established at least two years before the Crew Size Rule went into effect, including continuing to transport hazardous materials with a one-person crew. Id. § 218.129(a). They simply must provide written notice and information to the FRA and adopt safety-related operating rules covering three topics: accident response, radio or wireless communication, and installation of 'alerters' on the train's controlling locomotive.⁶ See id. § 218.129(b)–(c).

Class II and Class III railroads without legacy status may also exercise this same “notice” option to use a one-person crew—butPage 14 only if that operation doesn't transport hazardous materials. See id. § 218.129(a)(2).

The second, 'special approval' option is the pathway for other railroads to use a one-person crew. So Class I freight railroads, and any railroad seeking to start transporting hazardous materials, must obtain the FRA’s approval to use a one-person crew. These railroads must wait for sign-off from the FRA before they may start using a one-person crew. See id. § 218.131(a)(1).

The FRA approval process is more involved. Besides everything required in the “notice” option, these railroads must submit a detailed risk assessment to the FRA. See id. §§ 218.131(b)(15), 218.133. The agency then publishes the railroad’s petition and accepts public comments for 60 days. Id. § 218.135(b)-(c). The FRA will grant a railroad’s special-approval petition if the agency concludes that the railroad’s proposal is “as safe or safer than” using a two-person crew. See id. § 218.135(d)(2). Under the agency's timetable, it will normally grant or deny a petition within 120 days of receiving it. Id. § 218.135(d)(2). A railroad that receives special approval must also submit annual reports compiling safety and operational data from its one-person crews. Id. § 218.137.

The final Crew Size Rule differs somewhat from the rule that the 2022 Notice proposed. Four substantial changes are relevant here. The initial proposal categorically prohibited using oneperson crews to transport hazardous materials, but the final version permits this practice based on two-year “legacy” status or special approval. See 89 Fed. Reg. at 25082 col.3-25083 col.1. The final Page 15version of the Rule also allows Class II and Class III railroads that aren’t transporting hazardous materials to begin a new one-person operation through the 'notice' path, without special approval. Id. at 25073 col.3–25074 col.1. For railroads that need special approval, the FRA changed the standard railroads must meet from 'consistent with railroad safety” to “as safe or safer” than a two-person crew, explaining that this provided a clearer baseline. Id. at 25056 col.3. Last, the FRA provided an extended, two-year grace period before “legacy” one-person operations would need to comply with the Rule’s alerter requirement. See id. at 25074 col.3-25075 col.2.

The Final Rule states the FRA’s expectation that “the national, uniform standard provided in this rule [will] preempt State laws governing crew size.' Id. at 25061 col.3.

The Crew Size Rule went into effect on June 10, 2024.

Six railroads and two trade groups filed petitions for review of the Final Rule in the U.S. Courts of Appeals for the Fifth, Seventh, Eighth, Eleventh, and District of Columbia Circuits. The U.S. Judicial Panel on Multidistrict Litigation randomly selected the Eleventh Circuit as the court in which to consolidate the petitions for review, under 28 U.S.C. § 2112(a)(3).

Petitioners organized themselves into two groups. The first group (“AAR Petitioners”) consists of five petitioners. The Association of American Railroads (“AAR”) is a “nonprofit trade association representing the nation’s major freight railroads, Amtrak, andPage 16 some smaller freight railroads and commuter authorities.' The group includes Union Pacific Railroad Company and BNSF Railway Company, which are Class I freight railroads. It also includes Florida East Coast Railway LLC and Indiana Rail Road Company, which are Class II freight railroads.

The second group ('Short Line Petitioners') consists of three petitioners. The American Short Line and Regional Railroad Association describes itself as a “non-profit trade association representing the interests of the nation’s Class II (regional) and approximately 600 Class III (short line) railroads.' This group includes Texas & Northern Railway Company and Nebraska Central Railroad Company, which are short line railroads.

Respondents are the FRA, the Administrator of the FRA, and the U.S. Department of Transportation.

We granted a motion to intervene as a respondent by the Transportation Division of the International Association of Sheet Metal, Air, Rail and Transportation Workers ('SMART-TD'). SMART-TD is a labor organization representing train employees who are employed by all Class I railroads in the United States and by numerous Class II and Class III railroads. SMART-TD asserts that it “is the largest railroad union in North America.” 7

Page 17We have jurisdiction to review the FRA’s final actions under its safety rulemaking authority. See 49 U.S.C. § 20114(c). When we do so, the Administrative Procedure Act (“APA”) requires us to set aside final agency actions that are “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law” or that are 'in excess of statutory jurisdiction [or] authority.'⁵ U.S.C. § 706(2)(A), (2)(C).

In arbitrary-and-capricious review, a “court may not substitute its own policy judgment for that of the agency.' FCC v. Prometheus Radio Project , 592 U.S. 414, 423 (2021). Instead, we simply ensure that “the agency has acted within a zone of reasonableness and, in particular, has reasonably considered the relevant issues and reasonably explained the decision.” Id. This standard is 'exceedingly deferential.' City of North Mia. v. Fed. Aviation Admin. , 47 F.4th 1257, 1266 (11th Cir. 2022) (quoting Miccosukee Tribe of Indians of Fla. v. United States , 566 F.3d 1257, 1264 (11th Cir. 2009)).

In contrast, whether an agency has exceeded its statutory authority is a legal question that we decide independently. See Ins. Mktg. Coal. Ltd. v. FCC , 127 F.4th 303, 311 (11th Cir. 2025).

Page 18AAR Petitioners challenge the Crew Size Rule on five grounds. First, AAR Petitioners contend that the FRA exceeded its statutory authority because the Rule is not “necessary” for safety. Second, they argue the FRA arbitrarily finalized the Rule on the existing record because the FRA declined to finalize a two-person crew-size rule in 2019 and ignored commenters urging it to gather more data before regulating. Third, AAR Petitioners assert that the FRA unreasonably failed to consider the “substantial” labor costs of the Rule. Fourth, they posit that the Rule conflicts with the Rail Safety Improvement Act's 'railroad safety risk reduction program.” Finally, AAR Petitioners urge us to conclude that the FRA exceeded its statutory authority by taking too long to finalize the Rule.

We address each argument in turn.

AAR Petitioners first argue that the Crew Size Rule falls outside the FRA’s statutory authority to issue safety rules “as necessary.” See 49 U.S.C. § 20103(a). As AAR Petitioners see things, the statute empowers the FRA to regulate if, and only if, its rule is “essential”—and therefore “necessary”—to safety. Under this view, having insufficient evidence that a requirement would improve safety would leave the FRA powerless to act.

We’re not convinced. Reading the FRA’s rulemaking provision as a whole reveals a run-of-the-mill general rulemaking authority. The “as necessary” phrase is a grant of discretion about when to issue safety rules—not a stringent evidentiary bar.

Page 19Instead, the proper inquiry asks whether the FRA reasonably exercised its broad rulemaking authority. AAR Petitioners forfeited this argument, and the agency satisfied that obligation in any event.

AAR Petitioners' 'as necessary' argument fundamentally misunderstands the FRA’s rulemaking authority. We explain how that works and why the phrase “as necessary” doesn’t set an evidentiary standard.

When assessing the scope of an agency’s statutory authority, we look first to the statute’s text. See Oceanair of Fla., Inc. v. U.S. Dep’t of Transp. , 876 F.2d 1560, 1565 (11th Cir. 1989) (“An administrative agency . . . is a creature of Congress and has no authority beyond that granted by Congress.”). Here, the FRA promulgated the Crew Size Rule under the Secretary of Transportation’s general safety rulemaking authority. See 89 Fed. Reg. at 25058 col.2 n.33 (invoking 49 U.S.C. § 20103(a) as its legal authority). That statute says, 'The Secretary of Transportation, as necessary, shall prescribe regulations and issue orders for every area of railroad safetyPage 20 supplementing laws and regulations in effect on October 16, 1970.” 49 U.S.C. § 20103(a).⁸

This type of general rulemaking authority is common. A recent analysis found that more than 10,000 sections of the Code of Federal Regulations contain regulations that agencies issued under these sorts of general rulemaking powers. See Tamar Katz et al., General Rulemaking Grants and the Federal Trade Commission , 125 COLUM. L. REV. 2215, 2255 (2025).

General rulemaking provisions are typically broadly worded. Almost unavoidably, though, they contain some language discussing a nexus between the rules they authorize and the statutory framework the agency implements. Often, Congress uses the word “necessary” to describe that connection. Lesko v. United States , 161 F.4th 1352, 1362 & n.11 (Fed. Cir. 2025) (en banc) (listing five statutes showing “Congress regularly delegates authority to agencies to prescribe regulations ‘necessary’ to implement a statutory scheme'); see also, e.g. , 38 U.S.C. § 501(a) (“The Secretary [of the Department of Veterans Affairs] has authority to prescribe all rules and regulations which are necessary or appropriate to carry out the laws administered by the Department . . . .'); 42 U.S.C. § 7601(a) (“The Administrator [of the Environmental Protection Agency] is authorized to prescribe such regulations as are necessary to carry out his functions under this chapter.”).

Page 21Our precedent explains how we view the phrase “as necessary” in an agency’s general rulemaking-authority grant. We’ve said that “[w]hen the ‘empowering provision’ of a statute directs the agency to regulate as necessary to carry out what Congress intends, ‘the validity of a regulation promulgated thereunder will be sustained so long as it is reasonably related to the purposes of the enabling legislation.’” United States v. Florida , 938 F.3d 1221, 1230 (11th Cir. 2019) (quoting Mourning v. Family Publ’ns Serv., Inc. , 411 U.S. 356, 369 (1973)); accord U.S. Pipe & Foundry Co. v. Webb , 595 F.2d 264, 272-73 (5th Cir. 1979).⁹

AAR Petitioners don’t claim that the Crew Size Rule falls short of this standard. Rather, they wish the standard away. But neither of the ways they contend the standard is “doubly wrong” holds water.

First, they suggest that we ignore the Mourning standard because Congress used the word “necessary” instead of “related to.” But that’s also true of the cases making up this line of precedent. See, e.g. , Webb , 595 F.2d at 273 (“necessary”); Mourning , 411 U.S. at 361-2 ('necessary or proper'); Thorpe v. Hous. Auth. of City of Durham , 393 U.S. 268, 277 (1969) ('necessary'). So that word choice can’t distinguish these precedents.

Page 22AAR Petitioners’ second swing misses, too. They argue that precedents about broad authorities to “effectuate the aim of the Act” are inapt because the FRA’s authority addresses a “single purpose” of furthering railroad safety. We don’t see why this distinction matters. But in any case, it makes no difference here. Congress enacted the FRA’s rulemaking authority as part of the Federal Railroad Safety Act of 1970. Pub. L. No. 91-458, § 202(a), 84 Stat. 971, 971 (1970) (codified, as amended, at 49 U.S.C. § 20103(a)). So effectuating the aim of the Act is the same as advancing the Act’s single purpose: “to promote safety in all areas of railroad operations.' Id. § 101, 84 Stat. at 971.

And when we apply the proper standard—whether the regulation is reasonably related to the purposes of the enabling legislation—the Crew Size Rule fits comfortably within the FRA’s authority to issue safety rules. In its Final Rule, the FRA gave several explanations why it concluded that the Crew Size Rule promotes safety. First, removing the second crewmember’s ability to apply hand brakes or troubleshoot problems would remove an important check against “human factor caused accidents.” Second, it ensures that all railroads “incorporate safety risk information before implementing one-person operations, thereby reducing the risk of needing to take corrective action retroactively.” Third, the FRA credited commenters who “provided first-hand accounts of close calls and lives saved by the action of two crewmembers working as a team” and emphasized that a single crewmember’s fatigue or medical problems pose a greater safety risk when no redundancy exists Page 23on board. In the face of these explanations, we cannot say that the agency’s Crew Size Rule is not even “rationally related” to safety.

Even if we don’t consider the Mourning line of precedent, AAR Petitioners’ argument fails as a simple matter of statutory interpretation.

We must read the words of a statute in their context. See Republic of Sudan v. Harrison , 587 U.S. 1, 11 (2019). And when doing so, we favor giving statutory text its “most natural reading.” Id. Here, the parenthetical phrase “as necessary” plays a clear role in the sentence, “The Secretary of Transportation, as necessary, shall prescribe regulations and issue orders for every area of railroad safety supplementing laws and regulations in effect on October 16, 1970.” 49 U.S.C. § 20103(a). Without it, the sentence would be a stark, seemingly unreasonable mandate. It would read, “The Secretary of Transportation shall prescribe regulations and issue orders for every area of railroad safety supplementing laws and regulations in effect on October 16, 1970.” Read literally, the Secretary would have to issue rules for every area of railroad safety that preexisting law didn’t already address. “As necessary” tempers that command. Put differently, the phrase increases the Secretary’s discretion about when to regulate and on what subjects.

We see nothing on the other side of the ledger. The enacted statute gives no indication that Congress intended “as necessary” to restrict discretion. And AAR Petitioners’ repeated proposals to Page 24examine the word “necessary” in isolation clash with the axiom that statutory text 'cannot be construed in a vacuum.' Home Depot U.S.A., Inc. v. Jackson , 587 U.S. 435, 441 (2019) (quoting Davis v. Mich. Dept. of Treasury , 489 U.S. 803, 809 (1989)).

Driving home the conclusion, the FRA identified on-point judicial decisions supporting our understanding of the statutory text. AAR Petitioners did not respond to, much less undermine, that persuasive authority.

In Sierra Club v. Jackson , for instance, the D.C. Circuit considered a similar statute addressed to the Environmental Protection Agency (“EPA”). The statute directed that the EPA Administrator “shall . . . take such measures . . . as necessary to prevent the construction or modification of a major emitting facility” that doesn’t satisfy certain clean-air rules. Sierra Club v. Jackson , 648 F.3d 848, 851 (D.C. Cir. 2011) (emphasis added). The court held that this phrase made the administrator’s decision of whether or not to act an unreviewable discretionary decision. Id. at 855-56. Noting that there is no guidance as to what action is “necessary,” the court concluded that “‘as necessary to prevent the construction or modification of a major emitting facility’ . . . leaves it to the Administrator’s discretion to determine what action is 'necessary.'' Id. at 856. AAR Petitioners offer no way to distinguish the case. Indeed, they failed even to respond to the FRA’s citation to it.

Other parts of Title 49 similarly disfavor interpreting “as necessary” to restrict safety regulation. The first section of the chapter containing the rulemaking authority states that the “purpose of this Page 25chapter is to promote safety in every area of railroad operations and reduce railroad-related accidents and incidents.' 49 U.S.C. § 20101. That provision doesn’t mention countervailing considerations like minimizing compliance costs, much less suggest only “essential” safety regulations could outweigh such concerns. And enacted declarations of policy “can be a particularly helpful guide to selecting between permissible readings of a statutory phrase.” Drummond v. S. Co. Servs., Inc. , 177 F.4th 1076, 1099 (11th Cir. 2026) (citing Georgia v. President of the U.S. , 46 F.4th 1283, 1300 (11th Cir. 2022) (opinion of Grant, J.)).

Congress has continued to stress its overriding focus on safety. A later statute instructed the FRA to “consider the assignment and maintenance of safety as the highest priority, recognizing the clear intent, encouragement, and dedication of Congress to the furtherance of the highest degree of safety in railroad transportation.” 49 U.S.C. § 103(c). And the same statute requires the Federal Railroad Administrator to have “professional experience in railroad safety, hazardous materials safety, or other transportation safety.” 49 U.S.C. § 103(d). That requirement further reflects what one of our sister circuits called the FRA’s “broad mandate to assign, maintain, and enhance safety.” Transp. Div. of Int’l Ass’n of Sheet Metal, Air, Rail & Transp. Workers v. Fed. R.R. Admin. ( SMART-TD III ), 40 F.4th 646, 656 (D.C. Cir. 2022) (quotation marks and brackets omitted). Together, these statutes provide guidance about how the FRA reasonably exercises its discretion.

Page 26Reading “as necessary” to confer discretion also comports with precedent. Binding authority from the old Fifth Circuit held that a statute stating that a district judge “‘may order additional names . . . as necessary’ clearly contemplates allowing the District Judge discretion to determine when and if such addition is necessary.” United States v. Rodriguez , 588 F.2d 1003, 1006 n.10 (5th Cir. 1979). Other courts share the same view. A broad consensus of our sister circuits has recognized “as necessary” or “necessary” as affording discretion.¹⁰

AAR Petitioners contend that two Supreme Court decisions favor their contrary reading. They invoke the plurality opinion in the Supreme Court’s 1980 “ Benzene case,' Indus. Union Dep’t, AFLCIO v. Am. Petroleum Inst. , 448 U.S. 607 (1980), and the Court’s decision in Alabama Association of Realtors v. Department of Health and Human Services , 594 U.S. 758 (2021) (per curiam) (emergency docket). Neither case helps them.

In Benzene , the Court considered an Occupational Safety and Health Administration (“OSHA”) rule limiting the amount of benzene, a carcinogen, in nearly every U.S. workplace. OSHA issued Page 27the rule under a statute delegating authority to the Secretary of Labor to issue standards “reasonably necessary or appropriate to provide safe or healthful employment.” 29 U.S.C. § 652(8). A fourJustice plurality reasoned that the statute required the Secretary to make a “threshold finding” that a workplace poses “a significant risk of harm' before issuing a regulation. Benzene , 448 U.S. at 642– 44.

Benzene is readily distinguishable. Its reasoning relies largely on factors absent here. To support the Benzene Court’s “threshold finding” requirement, the Court identified two important points: 'the combination of the government's concession [in that case] that a cost-benefit analysis was required and the explicit provisions of the OSH Act ‘requiring the elimination of the most serious hazards first.'' Nat’l Mining Ass’n v. United Steel Workers , 985 F.3d 1309, 1317 (11th Cir. 2021) (quoting Benzene , 448 U.S. at 644). We have no similar concession here that the FRA’s rulemaking authority imposes a cost-benefit-analysis requirement, and unlike the OSH Act, the FRA's governing statute does not prioritize specific types of regulations. The Benzene Court also “emphasized the fact that the OSH Act had a pervasive impact on workplaces all across American industry.' Id. at 1318. In contrast, the FRA’s authority is limited to a single context for which Congress has repeatedly stressed the preeminent importance of safety. See, e.g. , 49 U.S.C. §§ 103(c), 20101.

Even if we adopted a Benzene -style 'threshold finding' requirement, it wouldn’t show that the FRA exceeded its authority.

Page 28The agency in Benzene made no finding that its safety standard was “reasonably necessary or appropriate.” Benzene , 448 U.S. at 638. Rather, the agency asserted that the statute compelled it to tightly restrict carcinogens, even without determining what level of exposure would be safe. See id. at 613. Here, though, the FRA concluded the Crew Size Rule was “necessary” for safety. 89 Fed. Reg. at 25053 cols.1-3. Whether an agency has adequate evidentiary support for such a finding is a distinct question to assess under a deferential standard of review, allowing agencies to “risk[] error on the side of overprotection rather than underprotection.” Benzene , 448 U.S. at 656.

Alabama Association of Realtors is even less favorable to AAR Petitioners’ cause. That decision held that realtors and rental property managers had a substantial likelihood of success in showing that the Centers for Disease Control and Prevention’s COVID-19 eviction moratorium exceeded its statutory authority. Ala. Ass’n of Realtors , 594 U.S. at 763.

The Court's reasoning supports our analysis here. The Court reviewed a list of seven specific examples of measures the Surgeon General could implement, such as “fumigation” and “disinfection.” Id. at 761 (quoting 42 U.S.C. § 264(a)). It considered those examples to inform the reach of the broader rulemaking authority, which allowed the Surgeon General to issue “regulations as in his judgment are necessary to prevent the introduction, transmission, or spread of communicable diseases.' Id. (quoting 42 U.S.C. § 264(a)). The Court explained that the seven examplesPage 29 'directly relate to preventing the interstate spread of disease by identifying, isolating, and destroying the disease itself.” Id. at 763. Meanwhile, the CDC’s eviction moratorium “relate[d] to interstate infection far more indirectly.' Id.

In other words, the Court assessed how closely the CDC’s eviction moratorium related to the Act's purpose of addressing communicable diseases. In doing so, it considered the statute’s specific examples as guidance for how tight that relationship needed to be. This precedent fits neatly with Mourning 's inquiry of whether a regulation is “reasonably related to the purposes of the enabling legislation.” 411 U.S. at 369 (quoting Thorpe , 393 U.S. at 280–81). And Alabama Association of Realtors gives no support for reading “as necessary” as an evidentiary straitjacket.

Because the Crew Size Rule falls within the FRA’s statutory authority, the proper question is whether the Rule is a reasonable exercise of the agency’s authority. The APA’s deferential standard requires an agency’s regulatory action to be “reasonable and reasonably explained.” Prometheus Radio , 592 U.S. at 423; see Biden v. Texas , 597 U.S. 785, 806–07 (2022) (applying same standard for an agency's 'exercise of discretion within [its] statutory framework').

AAR Petitioners forfeited any general challenge that the Crew Size Rule is an unreasonable exercise of the FRA’s regulatory discretion. Of course, they raise some narrower arguments under the arbitrary-and-capricious standard-we'll address them soon.

Page 30Yet AAR Petitioners’ opening brief does not contend that it was unreasonable to conclude that the Crew Size Rule is a warranted exercise of the FRA’s power to issue safety rules “as necessary.” That's a curious omission.¹¹ Although AAR Petitioners tack on this issue in their reply brief, that’s too late to properly present the issue for review. See LaCourse v. PAE Worldwide Inc. , 980 F.3d 1350, 1360 (11th Cir. 2020). So AAR Petitioners forfeited the argument.

But even if we could excuse AAR Petitioners’ forfeiture, we would conclude that the FRA exercised its rulemaking discretion reasonably.

On reply, AAR Petitioners suggest it was unreasonable for the FRA to regulate because it has “no data suggesting that oneperson crews are less safe than two-person crews.” We see two points AAR Petitioners might be trying to assert. But neither is persuasive.

To the extent they imply that only robust quantitative data can justify regulation, AAR Petitioners misunderstand the APA. “Administrative law doesn’t constrain agencies to act only when Page 31they have gold-standard quantitative data.” Citadel Sec. LLC v. SEC , 178 F.4th 621, 636 (11th Cir. 2026). It’s well-established that an agency “need not—indeed cannot—base its every action upon empirical data.” Id. (quoting Nasdaq Stock Mkt. LLC v. SEC , 38 F.4th 1126, 1142 (D.C. Cir. 2022)); cf. Prometheus Radio , 592 U.S. at 427 (“The APA imposes no general obligation on agencies to conduct or commission their own empirical or statistical studies.”).

If AAR Petitioners instead used “data” as a synonym for “evidence,” their argument fares no better. The FRA competently explained its reasoning that the Crew Size Rule would promote railroad safety and identified evidence supporting that reasoning. The agency also explained the Rule’s measured approach in an area undergoing technological shifts. Rather than blanket bans, the Rule permits any freight train to begin using one-person crews so long as doing so would be as safe as the status quo. The FRA pointed to at least four reasonable grounds that justify the Rule.

First, the FRA stressed the need for caution in rolling out a major operational shift like switching to one-person crews, given two safety-related trends. One is the increasing average length of Class I freight trains, rising to nearly three miles long in some instances. See 89 Fed. Reg. at 25053 col.2. The FRA noted that this trend adds “complexity and safety challenges” for operating those trains. Id. The other is a sharp rise in the rate of human-factorcaused accidents—a more-than-40 percent increase from 2013 to 2022, which accelerated dramatically from 2021 to 2022. Id. at 25053 col.3. The FRA concluded that departures from the statusPage 32 quo should happen carefully, particularly because only a second crewmember can effectively perform certain safety functions, such as securing a train with a hand brake. See id .

Second, the FRA invoked its analysis of safety research on the cognitive and collaborative demands on crewmembers. Id. at 25083 cols.2-3. The agency focused on five FRA-sponsored research reports and one Transportation Research Board report synthesizing several other studies. See 87 Fed. Reg. at 45572 col.145573 col.1. Based on its assessment, the FRA identified safety concerns about operating with a one-person crew that weighed in favor of the measured approach in the Final Rule. See 89 Fed. Reg. at 25083 cols.2–3.

Third, the FRA concluded that uncertainty around the introduction of a new technology, Positive Train Control ('PTC'), 12 justified its precautionary approach. The agency noted the unplanned and planned outages of PTC systems, along with “initialization failures, cut outs, and malfunctions,' with the technology. Id. at 25978 col.2. It also emphasized that the mandate for railroads to fully implement PTC technology only recently came into effect, leading to a continued learning curve. Id.

Page 33Fourth, the FRA relied on a broad set of public comments and hearing testimony supporting the importance of a second crewmember in railroad safety. The agency viewed these comments as corroborating the concerns it drew from safety research. Id. at 25067 col.2. For example, one former locomotive engineer testified about an incident when he tried to use the train’s horn and bell to warn a young child on the train track. Id. at 25062 col.3. Those tools failed. The former engineer credited the conductor on his train with saving the child’s life by running out onto the nose of the train's engine to visually warn the child. Id.

Other commenters discussed the importance of a second crewmember as a backstop to human error, the value of jointly troubleshooting unexpected problems, the psychological challenges of operating as a one-person crew, and the role a second crewmember plays in assisting their fellow crew in medical emergencies. Id. at 25062 col.1, 25063 col.1, 25063 col.3, 25065 col.3.

Still others emphasized the impact of multiperson crews on local emergency responders. Multiple commenters discussed how a second crewmember can help open railroad crossings blocked by a stopped train, with one discussing an ambulance that would have been blocked at the crossing for far longer without a second crewmember. Id. at 25064 col.2-25065 col.1. A union representing firefighters also noted its concern that a one-person crew poses a higher risk of failing to relay critical information about a train's cargo to on-scene first responders and to perform functions like separating train cars. Id. at 25063 col.2.

Page 34Overall, the FRA received numerous “first-hand accounts of close calls and lives saved by the action of two crewmembers working as a team.” Id. at 25064 col.2. The agency explained that this public input validated its determination that second crewmembers “play an active role in maintaining the safe operation of the train and safeguarding their fellow employees and the public.” Id. at 25067 col.2. Despite AAR Petitioners’ attempts to dismiss them as “anecdotes,” public comments can provide a sound evidentiary basis for administrative action. See Citadel Sec. , 178 F.4th at 636. Indeed, the Ninth Circuit’s rebuke of the 2019 Withdrawal Order emphasized that “comments supporting a two-person crew rule and . . . proffered anecdotal evidence” are evidence that the FRA ought to consider. SMART-TD I , 988 F.3d at 1183.

On the other hand, AAR Petitioners fail to identify “countervailing evidence” that a two-person crew worsens safety or affirmatively showing that crew size has no effect on safety. Prometheus Radio , 592 U.S. at 415.

We are satisfied that the FRA “reasonably considered the relevant issues and reasonably explained [its] decision” that the Crew Size Rule is a warranted exercise of its safety rulemaking authority. Id. at 423. We defer to an agency's judgments on matters that implicate its area of expertise, especially when making predictive judgments. Sunshine State Bank v. Fed. Deposit Ins. Corp. , 783 F.2d 1580, 1582 (11th Cir. 1986) (“The courts have uniformly recognized that certain types of agency judgments are not susceptible to strict 'proof' because they involve the exercise of discretion, technical Page 35expertise and informed prediction about the likely course of future events.”). After all, it’s the agency’s role—not ours—to consider the evidence and apply policymaking discretion. See Prometheus Radio , 592 U.S. at 423.

Exceeding our role under the APA would be particularly inappropriate in this case. “Where a statute is precautionary in nature . . . we will not demand rigorous step-by-step proof of cause and effect.' Miami-Dade County v. EPA , 529 F.3d 1049, 1064-65 (11th Cir. 2008) (quoting Ethyl Corp. v. EPA , 541 F.2d 1, 28 (D.C. Cir. 1976)). “[A]gencies can . . . adopt prophylactic rules to prevent potential problems before they arise. An agency need not suffer the flood before building the levee.” Stilwell v. Off. of Thrift Supervision , 569 F.3d 514, 519 (D.C. Cir. 2009) (Kavanaugh, J.).¹³

Demanding statistical certainty would require the FRA to sit and wait for a substantial number of accidents to occur with oneperson crews. Major railroad accidents are rare, but when they happen, they can be catastrophic. Congress recognized this fact when it directed the FRA to “consider the assignment and maintenance of safety as the highest priority.” 49 U.S.C. § 103(c). We willPage 36 not undermine Congress's command by judicially tying the agency’s hands.

Next, AAR Petitioners propose two narrow ways that it was arbitrary and capricious for the FRA to regulate crew sizes based on the existing rulemaking record. First, they contend that the FRA reversed its prior “position” that the evidence in 2019 didn’t warrant issuing a two-person crew-size rule. Second, they point to the National Transportation Safety Board’s (“NTSB”) 2016 recommendation that the FRA amend its accident/incident report form.¹⁴ AAR Petitioners contend that the FRA failed to respond to comments that urged it to heed this recommendation and wait to collect data before regulating crew size.

On both fronts, the FRA satisfied the requirement of reasoned decisionmaking.

AAR Petitioners argue that the Final Rule reversed the FRA’s prior policy without a reasoned explanation. In their view, the agency had to satisfy a heightened version of APA change-in- Page 37position doctrine because the Final Rule relies on factual findings that contradict those underlying the 2019 Withdrawal Order. Specially, AAR Petitioners assert that five statements in the 2022 Notice conflict with findings made in the 2019 Withdrawal Order.

This contention fails. AAR Petitioners can’t show that the FRA reversed a prior “policy,” as is necessary for change-in-position doctrine to apply at all, much less that a heightened standard applies 15 And in any case, the FRA provided reasoned explanations of why it chose to regulate crew size in 2024 even though it chose not to in 2019.

Change-in-position doctrine is well established. Generally, courts don't apply heightened scrutiny to an agency's reasoning when it rescinds or changes a prior policy. See Fox , 556 U.S. at514Page 3816. Still, the APA always requires an agency to “provide reasoned explanation for its action.” Id. at 515. And that explanation can look different when an agency is changing course.

We apply the change-in-position doctrine in two steps. The first step asks “whether an agency changed existing policy.” Food & Drug Admin. v. Wages & White Lion Invs., LLC , 604 U.S. 542, 569 (2025).

If it did, at the second step, we consider the agency’s reasoning. See id. When an agency changes course from a past policy, a reasoned explanation will ordinarily (1) demand that the agency “display awareness that it is changing position,' and (2) require it to 'show that there are good reasons for the new policy.' Fox , 556 U.S. at 515. But the agency has no need to convince a court “that the reasons for the new policy are better than the reasons for the old one.” Id. So the agency typically won't need to give 'a more detailed justification than what would suffice for a new policy created on a blank slate.' Id.

Sometimes, though, a heightened standard applies at this second step. An agency must provide a “more detailed justification . . . when, for example, its new policy rests upon factual findings that contradict those which underlay its prior policy; or when its prior policy has engendered serious reliance interests that must be taken into account.' Id. In those cases, the agency needs to give a “reasoned explanation . . . for disregarding facts and circumstances that underlay or were engendered by the prior policy.' Id. at 516.

Page 39AAR Petitioners fail at the first step. They contend that the 2019 Withdrawal Order represented the FRA’s “prior policy that a crew-size rule is not necessary for railroad safety.” But that vacated order isn’t cognizable as an “existing policy.” White Lion , 604 U.S. at 569.

As we’ve noted, the Ninth Circuit vacated the 2019 Withdrawal Order. SMART-TD I , 988 F.3d at 1179. We don't see how a judicially invalidated action leaves anything that constitutes agency policy. At least in the context of judicial orders, a vacated decision has no legal effect. After vacatur, a decision is “officially gone,” “void,” and has no “legal effect whatever.” United States v. Sigma Int’l, Inc. , 300 F.3d 1278, 1280 (11th Cir. 2002) (per curiam) (“None of the statements made . . . has any remaining force and cannot be considered to express the view of this Court.”) Other courts have persuasively written that we should treat vacated administrative orders the same way. See, e.g. , Action on Smoking & Health v. C.A.B. , 713 F.2d 795, 797 (D.C. Cir. 1983) (noting that, “by vacating or rescinding the rescissions [to the agency regulations] proposed by ER-1245, the judgment of this court had the effect of reinstating the rules previously in force').¹⁶

Page 40In the absence of the prior order, we have little more to consider than the agency’s earlier “failures to act” on crew size. Fox , 556 U.S. at 514. And departing from past agency inaction doesn't trigger the change-in-position doctrine. Id. at 514–16. AAR Petitioners cannot avoid this black-letter law.

As the Supreme Court noted last year, courts almost exclusively apply the change-in-position doctrine to positions expressed in “formal” settings. White Lion , 604 U.S. at 570 n.5. So reasoning that’s untethered to any valid agency action, formal or informal, surely cannot trigger the doctrine. In short, “we do not think that anything which can accurately be described as a change of official agency position has occurred here.” Smiley v. Citibank (S.D.), N.A. , 517 U.S. 735, 742 (1996).¹⁷

For these reasons, change-in-position doctrine does not apply here. So the FRA didn't have to provide a 'more detailed Page 41justification” than ordinarily required by that doctrine, as AAR Petitioners propose. Fox , 556 U.S. at 515.

But even if we disregarded the vacatur of the Withdrawal Order and assumed the FRA had changed its position, the FRA reasonably explained why it chose to issue a crew-size rule now. The Final Rule stated two independent reasons why the FRA promulgated a crew-size regulation now, despite choosing not to regulate in 2019. Either presents an independent, reasonable basis for changing course.

Once again, the FRA needs to show only that it “engaged in reasoned decisionmaking.” United Steel Workers , 985 F.3d at 1321. This “highly deferential” standard “presumes the validity of agency action.' Id. (quoting United Steel v. Mine Safety & Health Admin. , 925 F.3d 1279, 1283 (D.C. Cir. 2019)). The FRA’s explanation easily surpasses this bar.

As for the first reason, the FRA explained that it reassessed prior safety research after the Ninth Circuit vacated its 2019 Withdrawal Order. 89 Fed. Reg. at 25083 col.2. The Ninth Circuit concluded that the FRA's reasoning in the Withdrawal Order was faulty because it disregarded safety concerns that the FRAsponsored research we’ve discussed highlighted. SMART-TD I , 988 F.3d at 1183. It’s certainly not unreasonable to revisit an analysisPage 42 that a federal court of appeals held was arbitrary and capricious. Indeed, the opposite is true.

The FRA's other reason was the collection of trends we mentioned before: longer trains, increased accident rates, and uncertainties around new PTC technology. 89 Fed. Reg. at 25083 col.3. It was reasonable for the FRA to consider the sufficiency of safety rules after noting worrying trends in accident rates and substantial operational changes that bear on safety, such as train length. That is doubly true when we consider the primacy of safety to the FRA’s mandate. See 49 U.S.C. § 103(c).

AAR Petitioners make one last attempt to find a hook for requiring the FRA to provide robust empirical data. This time, they look to agencies’ obligation to “consider and respond to significant comments' in the rulemaking record. Perez v. Mortg. Bankers Ass’n , 575 U.S. 92, 96 (2015). AAR Petitioners’ argument centers on the 2016 NTSB recommendation for the FRA to update its accident report. Several commenters cited that recommendation while urging the FRA to wait and gather empirical data before considering a crew-size rule.

Before we address AAR Petitioners’ argument, a bit of background is useful. Unlike this time, the NTSB filed a comment in the crew-size rulemaking proceeding the FRA initiated in 2016. The NTSB was finalizing an investigation into a 2015 Amtrak Page 43derailment in Pennsylvania. Its comment asked the FRA to consider the NTSB’s forthcoming views on crew size following that investigation. See 84 Fed. Reg. at 24737 cols.1-2 n.9.

Later, the NTSB made two recommendations. It recommended that the FRA modify its “Accident/Incident Report” form to “include the number of crewmembers in the controlling cab of the train' during an accident. See Nat'l Transp. Safety Bd., Safety Recommendation: R-16-32 through -36, R-14-74 (Reiterated) (2016), https://www.ntsb.gov/safety/safety-recs/recletters/R-16-032036.pdf [https://perma.cc/6ALG-QD3K]. And it urged the agency to use the resulting data “to evaluate the safety adequacy of current crew size regulations.' Id.

AAR Petitioners frame those recommendations as telling the FRA not to regulate until after it updated the form and considered the resulting data. And they posit that three commenters’ references to the recommendation obligated the FRA to “explain why it rejected the NTSB’s recommendation.”

We disagree for two reasons. First, AAR Petitioners overread the NTSB’s recommendation. The NTSB never urged the FRA to refrain from regulating crew size without first collecting more accident data. See id. Sure, the NTSB recommended that the FRA use the resulting data to assess “the safety adequacy of current crew size regulations.” Id. But as AAR Petitioners stress repeatedly, the federal government hadn’t regulated crew size before the Crew Size Rule. The natural reading of the second recommendation is simply to use the data to assess whatever rules, state orPage 44 federal, happen to be in place.¹⁸ So we don't see the FRA's Final Rules as conflicting with this recommendation.

Second, the commenters who invoked the NTSB recommendation did so as part of a generalized complaint that the evidence doesn’t justify a regulation. The FRA responded fully and reasonably to that concern, as we’ve already discussed.

More specifically, AAR Petitioners cite three comments in the rulemaking record. One is from a member of Congress, another from the Alliance for Innovation and Infrastructure (“AII”), and a final from AAR, itself. Each of the three makes a short, passing reference to the NTSB recommendation while attacking the evidentiary basis for a rule.

AAR Petitioners seem to believe that we must invalidate the Rule if the FRA didn't address the NTSB recommendation by name. They are mistaken. Review under the APA is not a game of “gotcha.” An agency does not need to name-check every commenter and each document their comments reference. Such a requirement would force agencies to distort the 'concise general statement” of a rule’s basis and purpose that the APA requires, 5Page 45 U.S.C. § 553(c), into an interminable laundry list. And it would cause courts to abdicate our deferential role.

In any event, the FRA acknowledged and responded to all three of the comments AAR Petitioners cite. See 89 Fed. Reg. at 25068 cols.1-2, 25080 col.1, 25082 col.2-25083 col.2. More importantly, the FRA responded to the substance of their argument that a sufficient evidentiary basis to regulate didn’t exist. See id. at 25080 col.1, 25083 col.2–25084 col.1. The agency explained, for example, that a body of research “raise[d] safety concerns regarding one-person train crews, such as the loss of a second crewmember to notice and correct errors.' Id. at 25083 col.2. It also noted “troubling industry trends” in the latest rail safety data that, coupled with new challenges from using longer freight trains and the “enormous and sometimes incalculable costs” rail accidents can impose, called for 'heightened caution and awareness.' Id. at 25083 col.3.

The FRA fully considered the course of action the commenters proposed, too: essentially, the “no action alternative” that the FRA evaluated as the “main alternative to the final rule.” See Joint Pet’rs’ App’x (“JPA”) Doc. 13201 (2024 Regulatory Impact Analysis) at 38. As the FRA saw things, that option would have allowed railroads to reduce crew size “in a manner that impose[s] additional risk to railroad employees, or the general public,' and leave the FRA without data giving “insight into the safety protocols beingPage 46 established.” Id . AAR Petitioners may still not be convinced. But that doesn't mean the agency acted arbitrarily and capriciously.¹⁹

AAR Petitioners’ third argument is that the FRA failed to consider a major category of costs, making the Final Rule arbitrary and capricious. In their view, Michigan v. EPA , 576 U.S. 743 (2015), compelled the FRA to analyze the Rule’s costs. AAR Petitioners assert that the agency refused to consider the “cost” of foregone savings for railroads who would have switched to a one-person crew but for the Crew Size Rule. These lost savings, they continue, are the “largest cost imposed by [the] rule.” AAR Petitioners contend that this alleged oversight makes the Final Rule arbitrary and capricious.

This argument rests on a misreading of precedent and mere speculation that railroads would be unable to move to one-person operations under the Crew Size Rule. So we reject it.

We start with AAR Petitioners’ overreading of Michigan v. EPA . That decision held that, within the EPA's statutory framework, its authority to issue “appropriate and necessary” regulations Page 47of power plants “require[d] at least some attention to cost.” Michigan , 576 U.S. at 752. But the Court noted the limits of its decision. It cautioned that it was not requiring the EPA to conduct a “formal cost-benefit analysis in which each advantage and disadvantage is assigned a monetary value.' Id. at 759. Instead, the Court left 'the Agency to decide . . . how to account for cost.' Id.

Michigan v. EPA is also distinguishable from this case in an important way. The Court drew the requirement to consider cost from Congress’s inclusion of the word “appropriate” in the EPA’s regulatory authority, finding it significant that Congress required the EPA to decide “whether regulation is ‘appropriate and necessary,'' not just 'whether regulation remains 'necessary.'' Id. at 757.

Here, the FRA’s rulemaking authority lacks the term “appropriate.” In fact, Congress removed that key word from the statute. Compare 49 U.S.C. § 20103(a) (stating that the Secretary of Transportation “as necessary, shall prescribe regulations and issue orders for every area of railroad safety'), with Federal Railroad Safety Act of 1970, Pub. L. No. 91-458, § 202(a), 84 Stat. 971, 971 ('shall . . . prescribe, as necessary, appropriate rules, regulations, orders, and standards for all areas of railroad safety' (emphasis added)). So Michigan v. EPA is not instructive here.

In sum, AAR Petitioners haven't shown that Michigan v. EPA applies, and its requirement is flexible even when it does. So the decision doesn’t help AAR Petitioners show that the FRA’s treatment of labor costs made the Final Rule arbitrary and capricious.

Page 48That brings us to AAR Petitioners’ repeated assertion that lost savings are the “largest” cost imposed by the Crew Size Rule. Of course, repetition is not factual support. And of AAR Petitioners’ four arguments that the FRA overlooked significant labor costs of the Rule, none has a basis in fact.

First, two AAR Petitioners point to their own comments echoing that lost labor savings will be the Rule’s largest cost. But both comments merely restate estimates of the costs of the rule proposed in the 2016 Notice. See Union Pacific Railroad Co., Comment Letter on Train Crew Size Safety Requirements, at 33-34 (Dec. 21, 2022), Docket No. FRA-2021-0032, https://www.regulations.gov/comment/FRA-2021-0032-13012 [https://perma.cc/A4WH-5U4M]; JPA Doc. 13056 at 46 (AAR). Not only are the underlying analyses apparently outside the administrative record before us, the regulation proposed in 2016 differed in critical ways from the Final Rule. 20 So those analyses are unhelpful here.

Second, AAR Petitioners point to a study that AAR submitted in 2022, raising concerns that automation may give the trucking industry a labor-cost advantage. But that study doesn’t estimate Page 49increased labor costs for railroads resulting from the Crew Size Rule. It simply assumes “a mandatory onboard train crew of two individuals.” JPA Doc. 13056-9, at 9 (Burton Study). The Crew Size Rule does not have that requirement. Plus, the Final Rule considers this study and discusses concerns that costs from the Rule would disadvantage railroads against the trucking industry. 89 Fed. Reg. at 28080 col.3–28081 col.3.

Third, the FRA’s cost estimates dropped from 2016 to 2024. The FRA's high-scenario cost estimate for the 2016 Notice was $27.7 million over 10 years, 21 but the agency estimated $6.6 million in costs for the 2024 Final Rule. See Fed. R.R. Admin., Train Crew Staffing: Notice of Proposed Rulemaking Regulatory Impact Analysis (March 15, 2016), https://www.federalregister.gov/documents/2016/03/15/2016-05553/train-crew-staffing [https://perma.cc/X42K-UAJQ] [hereinafter 2016 RIA ]; JPA Doc. 13201 (2024 Regulatory Impact Analysis) at 8. AAR Petitioners tell us that this 'dramatic 75-percent drop is largely' a result of the agency’s ignoring of labor costs this time around.

This claim is baseless. After accounting for the 2016 Notice’s proposed exceptions, the 2016 Regulatory Impact Analysis estimated railroads would face a net $2.01 million in costs over 10 yearsPage 50 to employ second crewmembers. See 2016 RIA at 56 tbl.13.²² That's less than 10 percent of the drop in cost estimates that AAR Petitioners highlight—far from the predominant driver that AAR Petitioners suggest it was.

Also, we note again that the final version of the Crew Size Rule is meaningfully different from the rule proposed in 2016. The earlier iteration contained a no-exceptions ban on one-person crews transporting certain hazardous material. See id. at 41. So railroads would have had to pay two-person crews for that work in perpetuity. But as promulgated, the Rule allows railroads to use one-person crews for hazardous materials if they can show that it’s safe to do so. In other words, the Final Rule allows railroads to achieve labor-cost savings that the 2016 proposal would have prevented. It’s no surprise if that difference shows up as a lower cost estimate.

That brings us to the fourth concern. Throughout their briefing, AAR Petitioners repeatedly imply that the special approval process is a sham. They suggest that many, or perhaps most, Class I railroads and railroads hauling hazardous materials will be prevented from moving to a one-person crew. AAR Petitioners assert that the special approval standard—“as safe or safer Page 51than a two-person operation”—is “virtually impossible to satisfy in practice.” But they don’t support that contention.

Woven throughout their briefing is the idea that the Crew Size Rule is meant to please a political constituency, not advance safety. AAR Petitioners suggest that, as a result, we should assume that the FRA cynically designed the special approval process so that the “FRA will always be able to come up with some way to deem the data inadequate or find some risk that in its view is not sufficiently mitigated” and reject the railroad’s application. We’re not convinced. Whatever stress the presumption of regularity may be under today, AAR Petitioners need more than speculation and innuendo to overcome it. See Nat’l Parks Conservation Ass’n v. U.S. Dep’t of the Interior , 835 F.3d 1377, 1385 (11th Cir. 2016).

For all these reasons, the FRA did not unreasonably ignore substantial foregone labor-cost savings.

As a fourth ground for their challenge, AAR Petitioners argue that the Crew Size Rule is contrary to law because it conflicts with the Rail Safety Improvement Act of 2008’s “risk reduction program.” But we see no conflict.

That statute requires certain categories of railroads to develop comprehensive risk-reduction programs for railroad safety. 49 U.S.C. § 20156(a)(1), (c). Those railroads, which include all Class I railroads, must create a risk-reduction program thatPage 52 systematically evaluates and manages safety risks, subject to the Secretary's review and approval. Id. § 20156(a).²³ The scope of this risk-reduction program is broad. The railroad must “identify and analyze the aspects of its railroad, including operating rules and practices, infrastructure, equipment, employee levels and schedules, safety culture, management structure, employee training, and other matters, including those not covered by railroad safety regulations or other Federal regulations, that impact railroad safety.” Id. § 20156(c).

Nothing in this statute poses an impediment to the Crew Size Rule. First, we will not lightly infer an intent by Congress to curtail a broad rulemaking authority, especially when the new statute does not expressly state the alleged exception or even refer to the rulemaking authority in question. Am. Hosp. Ass’n v. NLRB , 499 U.S. 606, 613 (1991).

Second, here, the text of the statute cuts in the opposite direction. It plainly contemplates overlapping coverage of railroads’ risk-reduction programs and federal safety regulations. The statute includes a caveat that the “other matters” addressed in a railroad’s risk-reduction plan should include subjects “not covered by railroad safety regulations or other Federal regulations” if they impact railroad safety. 49 U.S.C. § 20156(c). The natural implication of including that caveat is that a railroad “systematically evaluat[ing]Page 53 railroad safety risks” will touch on many topics already covered by federal safety regulations. Id. § 20156(a)(1).

Third, AAR Petitioners’ theory is textually untenable in another way. AAR Petitioners worry that the Crew Size Rule “eliminates th[e] pathway' of addressing crew-size safety concerns through a railroad’s risk-reduction program. But the statute requires risk-reduction programs to address a wide range of safety issues. Under AAR Petitioners’ logic, the Act also foreclosed regulation of “operating rules and practices, infrastructure, equipment, employee levels and schedules, safety culture, management structure, employee training, and other matters . . . that impact railroad safety.' Id. § 20156(c). That reading is implausible. Yet AAR Petitioners provide no way to prevent their theory from requiring this result. So for this reason, too, we cannot conclude that the Act’s risk-reduction program creates any conflict with the Crew Size Rule.

Finally, AAR Petitioners ask us to vacate the Crew Size Rule because the FRA took more than 12 months to finalize the regulation. They rely on Congress’s directive that the Secretary of Transportation “shall” prescribe procedural rules including a time limit to dispose of each proceeding “not . . . more than 12 months after the date it begins.” Id. § 20103(b); see 49 C.F.R. § 211.13 (implementing the statute).

Two panels of the D.C. Circuit have dealt capably with the same argument, based on the same authorities, in cases involving Page 54the same parties as this one. See Transp. Div. of Int’l Ass’n of Sheet Metal, Air, Rail & Transp. Workers v. Fed. R.R. Admin. ( SMART-TD II ), 10 F.4th 869, 873-74 (D.C. Cir. 2021); SMART-TD III , 40 F.4th at 665–66. The court rejected arguments that it should vacate FRA rules that the agency took more than a year to promulgate.²⁴

We think the D.C. Circuit got it right. The U.S. Supreme Court’s decision in Barnhart v. Peabody Coal Company controls: '[I]f a statute does not specify a consequence for noncompliance with statutory timing provisions, the federal courts will not in the ordinary course impose their own coercive sanction.” 537 U.S. 149, 159 (2003) (quoting United States v. James Daniel Good Real Prop. , 510 U.S. 43, 63 (1993)).

That’s exactly the situation here. Like the D.C. Circuit, we don’t see anything in the “statutory language” or “relevant context' of the FRA's rulemaking authority that specifies ''consequences for missing a deadline’ []or otherwise ‘suggest[s] that vacatur should be the remedy for the agency’s failure to wrap up a regulatory action within twelve months.’” SMART-TD III , 40 F.4th at 665 (quoting SMART-TD II , 10 F.4th at 874). Pointing to a 'shall' in a statute or regulation is not enough to show “the grant of power Page 55can sensibly be read to expire when the job is supposed to be done.” Barnhart , 537 U.S. at 161.

Nor can AAR Petitioners distinguish Barnhart by noting that the FRA has the same statutory time limit in a procedural regulation. True, agencies are bound by their own regulations. Cf. United States ex rel. Accardi v. Shaughnessy , 347 U.S. 260, 267 (1954). But the question is what the consequence is when an agency fails to meet a regulatory deadline. And AAR Petitioners don’t explain why a regulatory time limit would put an expiration date on the FRA’s rulemaking authority.

In fact, the U.S. Supreme Court has applied Barnhart to time limits imposed in court-made procedural rules. See McIntosh v. United States , 601 U.S. 330, 339 (2024) (addressing Fed. R. Crim. P. 32(b)(2)(B)). And we don’t see the Court’s reasoning as suggesting we should treat agency -made procedural rules differently.²⁵ Just like in McIntosh , here, the FRA’s regulatory deadline “does not specify a consequence for noncompliance with its timing provisions.” Id. at 341 (citation omitted). So we won’t interpret the FRA’s regulation as a “rare” “jurisdictional deadline” that leaves an entity “completely powerless to take any relevant action.' Id. at 337.

We instead follow McIntosh ’s lead in interpreting the FRA’s deadline as a “time-related directive” that “functions ‘as a spur to Page 56prompt action, not as a bar to tardy completion of . . . business.’” Id. at 340 (quoting Barnhart , 537 U.S. at 172). A statute providing “a short period for rulemaking . . . indicates that the agency should move with dispatch.” SMART-TD II , 10 F.4th at 874.

If AAR Petitioners wished for a faster rulemaking proceeding, they could have sought the “ordinary remedy for tardiness”— “an order to ‘compel agency action unlawfully withheld or unreasonably delayed.’” Id. (quoting 5 U.S.C. § 706(1)). They did not seek one. They can’t now refashion Congress’s direction to work quickly into a message of 'better never than late.'²⁶

Short Line Petitioners challenge the Final Rule on two grounds of their own.²⁷ We address the flaws Short Line Petitioners identify in the Crew Size Rule’s legacy exception, then evaluate their complaints about the Rule’s alerter requirement.

Short Line Petitioners target three aspects of the Crew Size Rule’s exception for “legacy” one-person crew operations: the twoyear threshold for legacy status, alleged ambiguity about legacy status after a change in ownership, and the special-approval process for railroads that want to start carrying hazardous materials with a one-person crew.

We deal with the first two grounds together—Short Line Petitioners forfeited both—before turning to the third.

As for the first ground, Short Line Petitioners contend that the FRA didn’t explain why it set the threshold for legacy status at two years. But Short Line Petitioners didn’t identify any comment or testimony raising this issue to the agency during the rulemaking proceeding. So they’ve forfeited a challenge on this ground. See Advocs. for Highway & Auto Safety v. Fed. Motor Carrier Safety Admin. , 429 F.3d 1136, 1150 (D.C. Cir. 2005) (“[A] party will normally forfeit an opportunity to challenge an agency rulemaking on a ground that was not first presented to the agency for its initial consideration.”). Indeed, Short Line Petitioners all but concede on reply that they've forfeited this point.

With the second ground, Short Line Petitioners argue that the Crew Size Rule is “arbitrarily vague” about whether a change in ownership will end a one-person operation’s legacy status. The Rule confers legacy status on a “one-person train crew operation Page 58that has been established for at least two years” before the Rule’s effective date. 49 C.F.R. § 129(a)(1). It’s not clear enough, Short Line Petitioners urge, whether the FRA would consider a railroad the same “operation” after it’s acquired by a new holding company.

Here again, Short Line Petitioners fail to show that the FRA received comments or testimony raising this issue. And the relevant language was substantively identical in the 2022 Notice and the 2024 Final Rule. See 89 Fed. Reg. at 25110 col.1 (finalizing 49 U.S.C. § 218.129(a)(1)); 87 Fed. Reg. at 45618 col.2 (proposing equivalent language in the initial iteration of Section 218.131(a)). So they've forfeited this ground, as well.²⁸

The legacy exception is also arbitrary and capricious, Short Line Petitioners propose, because of how it handles operations transporting hazardous materials.

As we’ve mentioned, the Final Rule eliminated the 2022 Notice’s categorical ban on using one-person crews to transport hazardous materials. See 89 Fed. Reg. at 25082 col.3-25083 col.1. The final version of the Crew Size Rule allows any freight railroad to move towards one-person crews for hazardous materials. But Class II and III railroads face a substantial difference based on whether they started carrying hazardous materials with a one-person crew at least two years before the Rule went into effect.

Class II and III railroads with more than a two-year safety record may start using a one-person crew through the simpler “notice” option. But a railroad with less than two years transporting hazardous materials with a one-person crew—even if its one-person operation has transported non–hazardous materials for many years—must go through the more onerous special-approval path. Until they submit their petition, including the robust risk assessment and annual reports, and receive FRA approval, these railroads cannot haul hazardous materials with a one-person crew.

In effect, two types of legacy status exist for Class II and Class III freight railroads: 'standard legacy' and 'hazmat legacy.' A “standard legacy” operation could continue to haul non-hazardous materials but would need to use the special-approval process to Page 60start carrying hazardous materials.²⁹ In Short Line Petitioners' framing, requiring a “standard legacy” railroad to prepare a risk assessment and special-approval petition is an arbitrary burden. They assert that doing so “serves no safety purpose.”

But the FRA considered the relevant factors and reasonably explained its decision-all the APA requires. To start with, it's worth remembering how this disparity came about. In response to commenters, including the American Short Line and Regional Railroad Association (“Association”), the FRA removed the categorical prohibition on using one-person crews to haul hazardous materials. See 89 Fed. Reg. at 25070 cols.1-2, 25074 cols.1-2, 25082 cols.23. And while giving Short Line Petitioners much of what they sought, the FRA described its rationale for adopting a more careful approach with hazardous materials.

The Final Rule repeatedly notes the greater safety risks that hazardous materials pose. It explains that transporting hazardous materials 'creates the potential for a greater negative consequence' compared to other cargo by 'rais[ing] various specific Page 61safety hazards, such as the potential for an accidental or non-accidental release of a hazardous material.' Id. at 25055 cols.1-2. The Final Rule also relies on multiple federal regulators’ determination that the Rule’s defined set of hazardous materials “present[s] the greatest safety and security risks.” Id. at 25074 col.2. The FRA flagged that these hazards “would typically create additional tasks for a train crew” responding to a developing safety risk or attempting to mitigate an accident. Id. at 25055 col.1. It is rational for an agency to more carefully regulate the handling of materials that pose a greater safety risk.

The agency also set out the reasons it requires either a multiyear safety record or a risk assessment for one-person crews to transport hazardous materials.

As for the safety record, dating back to the 2022 Notice, the FRA expressed that at least two years of data would provide “an accident/incident safety record of a reasonable length on which [the] FRA could base any determination of the level of safety the operation provides.” 87 Fed. Reg. at 45595 col.2. In the Final Rule, the agency also explained its comfort with allowing “hazmat legacy” operations to continue because of the Rule’s imposition of new minimum safety requirements, such as mandating installation of alerters. 89 Fed. Reg. at 25074 col.1. And the FRA emphasized its commitment to “closely monitor this legacy exception and . . . scrutinize data or observations showing that the legacy operations may not be as safe as currently described.' Id.

Page 62The FRA also reasonably explained why it requires a risk assessment for railroads without a multiyear safety record. As the agency described it, the purpose of the risk assessment is 'evaluat[ing] risk in an objective manner” through a process to “systematically identify hazards, assess the degree of risk associated with those hazards, and . . . identify and implement measures to minimize or mitigate the risks to an acceptable level.' Id. at 25055 cols.1-2. Removing a second crewmember, the agency noted, could leave the remaining crewmember vulnerable in cases of accidents or derailments. Id. at 25053 col.3 n.5. And the FRA cannot assess the railroad’s proposed measures to compensate for those risks unless the agency can review an objective analysis. Id. at 25055 col.2.

We consider this decisionmaking to be reasonable. Short Line Petitioners quibble once more with the precise length of time for “hazmat legacy” status—why two years of data rather than oneand-a-half or three? But regulation often involves drawing this kind of line. See Mayo Found. for Med. Educ. & Rsch. v. United States , 562 U.S. 44, 59 (2011). Agencies have “‘wide discretion’ in making linedrawing decisions” like these. Nat’l Shooting Sports Found., Inc. v. Jones , 716 F.3d 200, 214 (D.C. Cir. 2013) (quoting WorldCom, Inc. v. FCC , 238 F.3d 449, 462 (D.C. Cir. 2001)). We will not disturb this sort of judgment call unless a petitioner shows the agency’s lines “are patently unreasonable, having no relationship to the underlying regulatory problem.' Id. at 215 (quoting ExxonMobil Gas Mktg. Co. v. FERC , 297 F.3d 1071, 1085 (D.C. Cir. 2002)). Short Line Petitioners have not carried this burden.

In their final argument, Short Line Petitioners take aim at the Crew Size Rule’s alerter requirement. Under the Rule, a train with a one-person crew must have a functioning alerter—the device that will activate warning alarms and eventually apply a “penalty brake” to stop the train if the engineer is non-responsive—in its controlling locomotive. 49 C.F.R. §§ 218.129(c)(3), 229.5.

Short Line Petitioners contend that the FRA ran afoul of the change-in-position doctrine in enacting this requirement. Before the Final Rule, they emphasize, the FRA required alerters only in locomotives that operate at speeds above 25 miles per hour. Id. § 229.140(a). But the Crew Size Rule’s alerter requirement applies to all freight trains with a one-person crew, even if they stay under 25 miles per hour. See id. § 218.129(c)(3).

As we discussed, to depart from a prior policy, the FRA must (1) “display awareness that it is changing position,' and (2) 'show that there are good reasons for the new policy.” Fox , 556 U.S. at 515. The agency satisfies that standard.

The Final Rule showed awareness that the Crew Size Rule would require railroads that aren’t covered by an existing alerter requirement to add alerters. See 89 Fed. Reg. at 25075 col.1. Indeed, the agency thoroughly addressed the American Short Line and Regional Railroad Association’s comments criticizing the addition of an alerter requirement for slower trains. See Id. at 25075 cols.1–2 & n.194.

Page 64Short Line Petitioners seek to portray this response as the FRA’s denial that it was changing its policy. They focus on the Final Rule's statement that the FRA was 'not issuing conflicting statements” relative to its 2012 order finalizing the earlier alerter rule. Id. at 25075 col.1.

But the agency was making an uncontroversial point. There is no inconsistency between stating that, all else equal, (a) slower trains pose a lower risk of injury and (b) trains without a second crewmember present a higher risk if there is no tool like an alerter to stop the train if the single crewmember becomes unresponsive. Besides, this argument mistakes our role under the APA. Short Line Petitioners ask us to search the Federal Register for a phrase that might look questionable in isolation when it comes to the FRA's determination that 'the agency is not issuing conflicting statements” about the alerter requirement. Id. But this is not a scavenger hunt. Rather, we conduct our deferential review with an eye towards the reasonableness of an agency’s action as a whole.

The FRA also provided good reasons for its new policy. The Final Rule explains the obvious safety benefits of equipping trains with an alerter if they are operated by a one-person crew. Without a backup like an alerter, “if a one-person train crew becomes incapacitated while the train is moving, the train would continue to operate down the track out of control without another crewmember on-board who could apply the emergency brake.” Id. at 25054 col.2.

Page 65The FRA thoroughly considered the cost to install alerters in locomotives for slower trains. See JPA Doc. 13201 (2024 Regulatory Impact Analysis) at 16–18. In response to cost concerns, the FRA extended the alerter requirement’s compliance deadline for many Class II and Class III trains in the final version of the rule. See 89 Fed. Reg. at 25074 col.3–25075 col.2.

Finally, the agency explained that alerters are “a necessary precaution to prevent the potential for catastrophic harm due to an uncontrolled train movement.' Id. at 25075 col.1. Requiring them would therefore “reduce the likelihood of at least one type of foreseeable accident that is more likely to occur with a one-person train crew,” by stopping trains with a physically unresponsive locomotive engineer. Id. at 25052 col.3. These explanations reflect that the agency engaged in reasoned decisionmaking.

For the reasons we have discussed, we DENY the petitions for review of the Final Rule.

Page 66BRANCH, Circuit Judge, Dissenting:

In 2024, the Federal Railroad Administration (“FRA”) issued a rule requiring trains to operate with a minimum of two crew members on board, subject to exceptions, claiming that such a requirement would enhance the safety of railroad operations. But FRA had considered a similar requirement just a few years prior in 2019 and concluded that research and other evidence did not support a minimum-crew-size rule. Now, FRA claims that, on reconsideration, it was wrong the first time around. But it relies on substantially the same data available to it in 2019 and offers no new research or circumstances that reasonably connect railroad safety and train crew sizes. Plus, its cost-benefit analysis failed to account for the labor costs railroads would incur to comply with the rule. For these reasons, I agree with the railroads bringing this challenge that the rule is arbitrary and capricious and should be vacated. I therefore respectfully dissent.

The Federal Rail Safety Act of 1970 provides, in relevant part, that “[t]he Secretary of Transportation, as necessary, shall prescribe regulations and issue orders for every area of railroad safety supplementing laws and regulations in effect on October 16, 1970.” 49 U.S.C. § 20103(b). “In prescribing regulations and issuing orders under this section, the Secretary shall consider existing relevant safety information and standards.' Id. § 20103(c). The Secretary ofPage 67 Transportation has delegated his rulemaking authority to the Administrator of FRA, see 49 C.F.R. § 1.89(a), and Congress has charged FRA with 'carry[ing] out all railroad safety laws of the United States,” 49 U.S.C. § 103(b). “In carrying out its duties, [FRA] shall consider the assignment and maintenance of safety as the highest priority, recognizing the clear intent, encouragement, and dedication of Congress to the furtherance of the highest degree of safety in railroad transportation.' Id. § 103(c).

In furtherance of this mandate, in 2024, FRA issued a Final Rule mandating that railroads operate with a minimum of two crew members, subject to exceptions. Train Crew Safety Requirements , 89 Fed. Reg. 25052 (Apr. 9, 2024) (“2024 Final Rule”). While this 2024 Final Rule was the first time FRA had promulgated such a requirement, the relevant history of the Rule goes back to 2016.

In 2016, FRA issued a Notice of Proposed Rulemaking (NPRM) proposing “a minimum requirement of two crewmembers . . . for all railroad operations, with exceptions proposed for those operations that FRA believe[d] d[id] not pose significant safety risks to railroad employees, the general public, and the environment by using fewer than two-person crews.” Train Crew Staffing , 81 Fed. Reg. 13918, 13918 (Mar. 15, 2016) ('2016 NPRM'). The rule also sought to “establish minimum requirements for the roles and responsibilities of the second train crewmember on a moving train.' Id.

Page 68In proposing the rule in 2016, FRA acknowledged that it lacked 'reliable or conclusive statistical data to suggest whether one-person crew operations [were] generally safer or less safe than multiple-person crew operations.' Id. at 13919. Indeed, the agency had not 'collect[ed] sufficient data related to the size of a train crew' to make a conclusion about whether crew size affected safety, and available “accident reports and investigations [did not] generally address the size of a crew' as it related to safety. Id.

Instead, FRA relied primarily on two derailments—one in Quebec and the other in North Dakota—and six research reports to justify the 2016 NPRM. Id. at 13921–30. FRA determined that the Quebec derailment supported imposing a two-person-crew requirement even though the Canadian agency tasked with investigating the derailment “found that it could not be concluded that a one-person crew contributed to the accident.” Id. at 13922. FRA also determined that the North Dakota derailment supported its proposed rule even though that derailment involved a two-train collision in which one train was manned by two crew members and the other by three. Id. at 13923-24. FRA concluded that because the multi-person crew on one of the trains in North Dakota was able to help with the post-accident response, and a one-person crew could not have helped as much, a two-person-crew requirement was warranted. See id.

FRA claimed that the six research reports also justified its proposed rule, even though none of them stated with certainty or data that two-person crews were safer than one-person crews.

Page 69Instead, the reports asserted generally that two-person crews were preferable because (1) train engineers and conductors “operate as a joint cognitive system,” id. at 13925; (2) a passenger train conductor needs a lot of “knowledge, skills, abilities, and other characteristics” to do his job well, id. at 13926; (3) railroad employees experience fatigue on the job, id. ; (4) conductors might experience more “cognitive demands” with the “introduction of new train control technologies,” id. ; (5) there was uncertainty regarding operations due to new technology and conductors' responsibility to adapt thereto, id. at 13928-29; and (6) put simply, teamwork was important, id. at 13929-30.

FRA conducted a regulatory impact analysis, which it issued with the 2016 NPRM. In it, FRA conducted a thorough cost-benefit analysis to determine the costs of complying with the proposed rule. It concluded that such costs could be as high as $27.7 million over 10 years. See 2016 NPRM, 81 Fed. Reg. at 13919. Among the various costs FRA accounted for, it included costs “result[ing] from additional crew members on the trains that are currently operating with a one-person crew and from the delay or prevention of instituting new one-person crew operations in the future” (emphasis added).

With that background, the 2016 NPRM asked for public comment. However, in 2019 FRA withdrew the NPRM, “find[ing] that no regulation of train crew staffing [was] necessary or appropriate at [that] time.” Train Crew Staffing , 84 Fed. Reg. 24735, 24737 (May 29, 2019) ('2019 Withdrawal Order'). According to FRA,Page 70 “despite studying this issue in-depth and performing extensive outreach to industry stakeholders and the general public, FRA’s statement in the [2016 NPRM] that it ‘cannot provide reliable or conclusive statistical data to suggest whether one-person crew operations are generally safer or less safe than multiple-person crew operations' still holds true today.' Id.

The Withdrawal Order explained why the purported evidence it relied on in the 2016 NPRM did not support a two-person requirement. First, the Quebec and North Dakota derailments: FRA pointed out that any connections between the derailment and the one-person crews were “tangential at best and [did] not provide a sufficient basis for FRA regulation of train crew staffing requirements.' Id. at 24738. The Quebec derailment occurred because of the crew member’s failure to follow rules, not because he did not have another crew member assisting him. Id. And as for the North Dakota derailment, FRA concluded that the same type of post-accident response could have been conducted if the relevant train had one crew member and implemented a “well-planned, post-accident protocol.' Id. The 2019 Withdrawal Order reiterated its statement from the 2016 NPRM that “FRA [did] not have information that suggest[ed] that there ha[d] been any previous accidents involving one-person crew operations that could have been avoided by adding a second crewmember.' Id. at 24738.

Addressing the six research reports, FRA pointed out that “[w]hile the[] reports identif[ied] safety issues that railroads should consider when evaluating any reduction in the number of trainPage 71 crewmembers or a shift in responsibilities among those crewmembers, the reports [did] not indicate that one-person crew operations [were] less safe and therefore [did] not form a sufficient basis for a final rule on crew staffing.” Id. at 24740. And research into the issue that FRA had commissioned in 2013 was ultimately “unable to reach consensus on any recommendation or identify conclusive, statistical data to suggest whether there is a safety benefit or detriment from [two-person crews].' Id. at 24736.

Finally, FRA concluded that a two-person-crew rule would “unnecessarily impede the future of rail innovation and automation.” Id. at 24740 (capitalization omitted). Automation, FRA explained, would reduce human-caused accidents. Id. For all these reasons, FRA refused to “depart[] from [its] long-standing regulatory approach of not endorsing any particular staffing arrangement.” Id. The Order also stated that 'FRA intend[ed] to negatively preempt any state laws concerning' crew sizes on trains. Id. at 24741.

In 2021, in response to a challenge to the 2019 Withdrawal Order, the Ninth Circuit vacated the Order, finding that the Order's apparent attempt to preempt any state efforts to impose a two-person-crew rule was unlawful. See Transp. Div. of the Int'l Ass’n of Sheet, Metal, Air, Rail, & Transp. Workers v. Fed. R.R. Admin. , 988 F.3d 1170 (9th Cir. 2021). The Ninth Circuit determined that the Order’s statement about preemption of state laws attempted toPage 72 “authorize nationwide one-person train crews and to bar any contrary state regulations.' Id. at 1182.

The court held that while “[t]here [was] no doubt that the FRA could withdraw the [2016 NPRM]” as it related to requiring two crew members on trains, the “lack of data [did] not support the promulgation” of what was effectively a “one-person train crew rule and the preemption of state safety laws.” Id. at 1178, 1182. The court therefore vacated the 2019 Withdrawal Order’s preemption mandate as arbitrary and capricious. See id. at 1182-84. In so holding, the Ninth Circuit pointed to the lack of “reliable or conclusive statistical data to suggest whether one-person crew operations are generally safer or less safe than multiple-person crew operations.' Id. at 1182 (quotations omitted). According to the court, that lack of evidence—as well as a failure to address potential safety concerns involving one-person crews—made the “promulgation of a one-person train crew rule” arbitrary and capricious, under the Administrative Procedure Act (“APA”). Id. at 1183. But, notably, while the Ninth Circuit held that there was not enough evidence to mandate a one-person crew rule and preempt contrary state laws, it did not hold or suggest that a two-person-crew rule was warranted. See id. at 1182-84.

In July 2022, FRA issued a new NPRM, again proposing “[a] minimum requirement of two crewmembers . . . for all railroad operations, with exceptions proposed for those operations that do not pose significant safety risks to railroad employees, the public, Page 73or the environment.” See Train Crew Size Safety Requirements , 87 Fed. Reg. 45564, 45564 (July 28, 2022) (“2022 NPRM”). A 146-day public comment period followed, during which FRA received approximately 13,576 comments. 2024 Final Rule, 89 Fed. Reg. at 25059. In 2024, FRA issued the final rule we now review. See id. at 25052.

“The final rule establishes minimum crew safety size standards for all trains, including a risk assessment requirement to evaluate hazards and ensure risk mitigation for those railroads looking to initiate one-person train crew operations in the most complex environments nationwide.” Id. at 25052. According to FRA, the rule 'will reduce the likelihood of future accidents proactively.' Id.

The Final Rule provides different restrictions based on whether a railroad is classified as Class I, Class II, or Class III. According to FRA, Class I railroads are generally “the largest railroads operating in the United States”; Class II railroads are “regional railroads,” and Class III railroads are “shortline” railroads.

Specifically, the Final Rule includes the following provisions. As a baseline, all covered trains “shall be assigned a minimum of two crewmembers.' 49 C.F.R. § 218.123(b). However, some smaller railroads are automatically excepted from the two-person rule: If a Class II or Class III railroad has a “one-person train crew freight operation that has been established for at least two years before June 10, 2024,” that railroad “may continue to operate with a one-person train crew, including continuing to transport hazardous materials' identified elsewhere in FRA's regulations. Id.

Page 74§ 218.129(a)(1). A Class II or Class III railroad may also switch to a one-person crew for the first time if it does not carry hazardous materials. Id. § 218.129(a)(2); see 2024 Final Rule, 89 Fed. Reg. at 25056 (stating that Class II or Class III railroads may 'initiat[e] a new, non-legacy, one-person train crew freight operation not transporting hazardous materials” without petitioning FRA for special approval).

However, the Final Rule lays out significant compliance requirements for railroads seeking to proceed under these exceptions: they must “provide FRA with written notice” of their intent to continue operating with a one-person crew, 49 C.F.R. § 218.129(a)(1)(i)(A), (b); '[a]dopt and comply with an operating rule' that provides for mitigation measures and communication protocols “that provide a level of safety that is as safe or safer than a two-person train crew operation,” id. § 218.129(c)(1), (2); see also id. § 218.129(a)(1)(i)(B); and eventually equip the operating locomotive “with a functioning alerter” operating as the regulations elsewhere require, id. § 218.129(c)(3); see also id. § 218.129(a)(1)(ii).

Any railroads that do not qualify for an automatic exception (such as Class I railroads, and Class II or III railroads that wish to start carrying hazardous materials with a one-person crew for the first time) also face burdensome requirements if they decide to request special approval to operate with a one-person crew. See id. § 218.131. A railroad seeking special approval must submit a petition 'contain[ing] sufficient information for FRA to determine whether approving the operation described in the petition is as safePage 75 or safer than a two-person minimum train crew operation.” Id. § 218.131(b). The petition must include information on 15 specific topics related to the railroad’s operations, including “[t]he location of the operation, with as much specificity as can be provided'; “[t]he class(es) of track to be operated over [and] the method of operation”; “[t]he maximum authorized speed of the operation”; “[t]he approximate average number of miles and hours a person is projected to operate as a train crewmember in a one-person train crew operation”; “[t]he maximum number of cars and tonnage proposed for the operation”; and “[w]hether the railroad is seeking approval to transport hazardous materials.” Id. § 218.131(b)(1)-(15). The petitions must also include a risk assessment that identifies hazards associated with moving to a one-person crew, calculates and recalculates the risks identified for each hazard, and “[p]rovide[s] a statement with supporting evidence that the oneperson train crew operation with a fully implemented mitigation plan is as safe or safer than a two-person minimum train crew operation.” Id. § 218.133(a)(3). 'FRA may request any additional information, beyond what is provided in the petition, that it deems necessary.' Id. § 218.131(c).

'If FRA finds that the petition complies with the requirements of § 218.131 and that approving the petition is as safe or safer than a two-person minimum train crew operation, FRA will grant the petition, normally within 120 days of its receipt.” Id. § 218.135(d)(2). Otherwise, FRA will deny the petition. Id. § 218.135(d)(3). 'FRA may attach special conditions to the approval of the petition.' Id. § 218.135(d)(2).

Page 76The Final Rule purported to offer justifications for the new two-person requirement. But FRA simply pointed to general data related to railroad safety, explaining that 'the latest annual rail safety data reflects some troubling trends that point toward a need for heightened caution and awareness in railroad safety and operational planning.” 2024 Final Rule, 89 Fed. Reg. at 25053. It observed that '[t]he rate for all human factor caused accidents increased” by over 40% between 2013 and 2022, and by 13.6% from 2021 to 2022, but it did not offer any connection between these accidents and the crew sizes of the trains involved. See id.

Otherwise, FRA relied on individual anecdotes “of close calls and lives saved by the action of two crewmembers working as a team.” Id. at 25064. Commenters pointed out that 'the second crewmember can mitigate the risk of a sick or tired crewmember,” id. at 25063, and that 'two crewmembers keep each other alert and on task,” id. at 25064. Other testimony suggested that new technologies “introduced new complexities and levels of attention capture not seen” before, which “emphasized the need for a conductor on board due to the added level of distraction' on one-person crews. Id. at 25062. New technologies can also 'make[] the locomotive cab feel crowded,” and they require engineers to “direct significant attention to computer screens,” so FRA believed that imposing a two-person-crew requirement could mitigate these risks. Id.

But FRA did not point to data or statistical evidence that a general two-person-crew rule was necessary for railroad safety, and Page 77its regulatory impact analysis noted that it had “little insight into any . . . data on the safety impacts” of state regulations requiring two-person crews. It acknowledged that FRA “qualitatively discussed the benefits [of the rule] because it does not have sufficient data to monetize those benefits.' Id. at 25052.

Without concrete data to support the rule, FRA stated broadly that the rule would “reduce the likelihood of future accidents proactively” and had “the potential to reduce the likelihood of at least one type of foreseeable accident”: accidents caused by a crew member on a one-person crew that becomes “physically unresponsive.” Id. The agency stated that its conclusions about the likelihood of reducing future accidents were valid “even if the type of accident foreseen has not yet occurred.” Id. Nowhere in the Final Rule did FRA point to a single railroad accident or derailment caused by or attributed to operating with a one-person, rather than a two-person, crew. Instead, it relied on the allegedly commonsense notion that two are better than one.

FRA acknowledged that the 2024 Final Rule was a change in position from the 2019 Withdrawal Order, but it explained that it had “reassessed its prior positions for two independent reasons.” Id. 25083. First, in 2019 it had 'de-emphasized safety concerns raised by FRA-sponsored research”— i.e. , the six research reports from the 2016 NPRM—“on the cognitive and collaborat[ive] demands of crewmembers and by commenters on the 2016 NPRM.” Id. 'The final rule is justified based on FRA's reevaluation of those safety concerns and the threat they pose to public safety.' Id.

Page 78Second, FRA claimed it “considered information not analyzed in the 2019 withdrawal, such as technological trends and operational changes on Class I freight railroads since 2019.” Id. FRA maintained that (1) the industry is “utilizing longer trains than the historical norm”; (2) “the rate for all human factor caused accidents has increased in recent years, notably after the 2019 withdrawal”; and (3) there was “uncertainty related to new operating technologies [that] can effect train safety.” Id. Again, though, FRA did not explain how the longer trains, human factor caused accidents, or 'new operating technologies' were connected to any concrete safety risk posed by one-person crews. See id. Rather, it simply said that for those reasons, it “needed to change its position from the 2019 withdrawal.' Id.

As part of FRA’s regulatory impact analysis, it assessed the potential costs that the Rule would impose on the railroad industry. This time around, FRA estimated that the 10-year costs of the Final Rule would be $6.6 million, calculated primarily from the costs associated with maintaining an exception or applying for special approval. Id. at 25057. FRA did not account for increased labor costs; according to the agency, “[t]he requirements of this rule do not affect railroads operating trains with two-person crews”— i.e. , all Class I railroads.¹ '[T]herefore, they will not incur any costs.'²

Several railroads filed petitions challenging the 2024 Final Rule. The petitioners include Class I railroads Union Pacific Railroad Company and BNSF Railway Company, Class II railroads Florida East Coast Railway LLC and Indiana Rail Road Company, and Class III railroads Nebraska Central Railroad Company and Texas & Northern Railway Company, as well as two trade associations, the Association of American Railroads and the American Short Line and Regional Railroad Association, which represent the interests of various railroads.

We allowed the International Association of Sheet Metal, Air, Rail, and Transportation Workers—“an unincorporated association acting as a labor union” for railroad employees—to intervene as a respondent.

The petitioners challenge the 2024 Final Rule on several grounds, but two provide independent bases for concluding that the Rule is arbitrary and capricious. First, FRA failed to reasonably explain its change in position from the 2019 Withdrawal Order to the 2024 Final Rule. And second, FRA performed a cost-benefit Page 80analysis in developing the 2024 Final Rule but did not account for labor costs associated with maintaining two-person crews. I will consider each in turn.

The APA requires courts to 'hold unlawful and set aside agency action, findings and conclusions found to be . . . arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with the law.” 5 U.S.C. § 706(2)(A). “The scope of review under the ‘arbitrary and capricious’ standard is narrow and a court is not to substitute its judgment for that of the agency.” Motor Vehicle Mfrs. Ass’n v. State Farm Mut. Auto. Ins. Co. , 463 U.S. 29, 43 (1983). But the agency “must examine the relevant data and articulate a satisfactory explanation for its action including a rational connection between the facts found and the choice made.' Id. (quotation omitted). The agency’s explanation must be “sufficient to enable us to conclude that [its action] was the product of reasoned decisionmaking.' Id. at 52.

An agency decision may be arbitrary and capricious if the agency changes positions without adequately explaining its reasoning. The 'change-in-position doctrine' asks (1) 'whether an agency changed existing policy” and (2) whether the agency “display[ed] awareness that it is changing position and offer[ed] good reasons for the new policy.' FDA v. Wages & White Lion Invs., LLC , 604 U.S. 542, 569–70 (2025) (emphasis in original) (quotations omitted).

Page 81While “the agency must show that there are good reasons for the new policy,” “it need not demonstrate . . . that the reasons for the new policy are better than the reasons for the old one.' FCC v. Fox Television Stations, Inc. , 556 U.S. 502, 515 (2009) (emphasis in original). However, when the agency’s “new policy rests upon factual findings that contradict those which underlay its prior policy,” it must provide “a more detailed justification.” Id. When such a change occurs, “a reasoned explanation is needed for disregarding facts and circumstances that underlay . . . the prior policy.” Id. at 516.

The Majority contends that the Final Rule was not a change in position and, in any event, FRA sufficiently explained its new position. I disagree on both points.

In the 2019 Withdrawal Order, FRA found that there was not “reliable or conclusive statistical data to suggest whether oneperson crew operations are generally safer or less safe than multiple-person crew operations.” 2019 Withdrawal Order, 84 Fed. Reg. at 24737. Five years later, FRA’s 2024 Final Rule imposed a twoperson-crew rule and stated that the “final rule [was] justified based on FRA’s reevaluation of . . . safety concerns [related to one-person crews] and the threat they pose to public safety.” 2024 Final Rule, 89 Fed. Reg. at 25083. In other words, FRA clearly changed its position between 2019 and 2024 on whether a two-person-crew requirement would enhance public safety. And FRA recognized as Page 82much; in the 2024 Final Rule, it observed that it had “reassessed its prior positions.' Id.

But according to the Majority, no such change in position took place because the Ninth Circuit vacated the 2019 Withdrawal Order. I disagree.

First, we need not decide what effect, if any, the Ninth Circuit’s ruling had on the 2019 Withdrawal Order, because FRA has affirmatively conceded that its 2024 Final Order was a change in position. Both in the 2024 Final Order and in its briefs before this Court, FRA acknowledged that the Final Order was a change in position from the earlier 2019 Withdrawal Order. In the Final Order, FRA observed that it had “reassessed its prior positions” and “determined that it needed to change its position from the 2019 withdrawal.” 2024 Final Order, 89 Fed. Reg. 25083. And in its brief before this Court, FRA asserts that it “recognized that its decision to promulgate a rule addressing crew size safety differed from its choice in 2019 to withdraw a prior, different, proposed crew size rule.' Because FRA's Final Rule and briefing have conceded its change in position, we need go no further in evaluating the first step of the change-in-position doctrine. See Ohio v. EPA , 603 U.S. 279, 298 (2024) (observing, in response to a novel argument raised by the dissent regarding arbitrary and capricious review, that the Supreme Court 'normally declines to entertain arguments forfeited by the parties” (alteration adopted) (quotations omitted)).

In any event, the Majority has not pointed to a single Supreme Court or court of appeals decision holding that the change- Page 83in-position doctrine does not apply in this context.³ The Majority assumes that when an agency order has been vacated, it is as if the order never existed. But there is simply no caselaw supporting such an assumption. The cases the Majority cites suggest only that Page 84the vacated order no longer has any legal effect. They do not imply that the order-with all its factual findings-completely dissolves.⁴

Thus, I disagree with the Majority's conclusion that the change-in-position doctrine did not apply such that FRA had no obligation to explain its changed view of the facts simply because its earlier order had been vacated on other grounds.

I now move to the second step of the change-in-position doctrine and find that FRA did not satisfy it. FRA did not merely change its position when it promulgated the 2024 Final Rule; it made factual findings that contradicted its findings in the 2019 Withdrawal Order—relying on largely the same data it considered in 2019-without offering adequate explanations for the change. See Fox Television , 556 U.S. at 515.

Page 85In 2019, FRA withdrew its 2016 NPRM, which would have required two-person crews. In doing so, it stated that “existing oneperson operations have not yet raised serious safety concerns and, in fact, it is possible that one-person crews have contributed to the railroads’ improving safety record.” 2019 Withdrawal Order, 84 Fed. Reg. at 24739 (alteration adopted) (quotations omitted). FRA determined that “despite studying this issue in-depth and performing extensive outreach to industry stakeholders and the general public, FRA’s statement in the [2016] NPRM that it ‘cannot provide reliable or conclusive statistical data to suggest whether one-person crew operations are generally safer or less safe than multiple-person crew operations’ still holds true today.” Id. at 24737. Among other things, FRA observed that it did “not have information that suggest[ed] that there ha[d] been any previous accidents involving one-person crew operations that could have been avoided by adding a second crewmember.” Id. at 24738. Addressing the six research reports the 2016 NPRM had relied on, FRA noted that “[w]hile the[] reports identif[ied] safety issues that railroads should consider when evaluating any reduction in the number of train crewmembers or a shift in responsibilities among those crewmembers, the reports [did] not indicate that one-person crew operations [were] less safe and therefore [did] not form a sufficient basis for a final rule on crew staffing.' Id. at 24740. FRA also determined that a two-person-crew rule would “unnecessarily impede the future of rail innovation and automation,' because automation would reduce human-error accidents. Id. (capitalization omitted).

Page 86In the 2024 Final Rule, on the other hand, FRA reversed course and claimed that a two-person-crew rule was now “necessary” to keep people safe from railroad accidents. 2024 Final Rule, 89 Fed. Reg. at 25053, 25083 ('This final rule is justified based on . . . safety concerns” about one-person crews.). Because FRA’s “new policy rest[ed] upon factual findings that contradict[ed] those which underlay its prior policy,” it was required to provide “a reasoned explanation . . . for disregarding facts and circumstances that underlay” the 2019 Order. Fox Television , 556 U.S. at 515–16. But FRA did not do so: it did not sufficiently explain how it reached an opposing conclusion while relying on largely the same evidence it discussed in the 2019 Withdrawal Order, and it did not connect the changed circumstances it identified to the safety of one-person crews.

In the 2024 Final Rule, FRA offered two justifications for its changed position. First, it said that the 2019 Withdrawal Order had “de-emphasized safety concerns raised by FRA-sponsored research on the cognitive and collaborat[ive] demands of crewmembers and by commenters on the 2016 NPRM.” 2024 Final Rule, 89 Fed. Reg. at 25083. It offered the general explanation that “the research raises safety concerns regarding one-person train crews, such as the loss of a second crewmember to notice and correct errors,” and that the Final Rule was 'justified based on FRA's reevaluation of those safety concerns and the threat they pose to public safety.' Id.

But the “research” FRA referred to did not draw any link between crew size and railroad accidents or concrete safety risks, and Page 87FRA had recognized as much in the 2019 Withdrawal Order. And FRA did not explain why the same research it had previously found inadequate was now sufficient.⁵ Simply regurgitating the same studies and findings FRA had previously rejected and concluding, without explanation, that, upon reevaluation, they now provided sufficient support is not a “reasoned explanation.” Fox Television , 556 U.S. at 515–16.

Second, FRA maintained that the change in position was justified by changed circumstances such as longer trains, an increase in human-error accidents, and uncertainty about new technologies. 2024 Final Rule, 89 Fed. Reg. at 25083. But FRA’s explanations were conclusory at best and suffered from other underlying flaws. FRA first claimed that “freight train length has increased in recent years.” Id. For support, FRA cited the 2022 NPRM, which declared that “[t]he growth in the number of trains with more than 150 rail cars is a business practice that FRA has observed over the past several years.” 2022 NPRM, 87 Fed. Reg. at 45571. The 2022 NPRM, in turn, cited a study from May 2019—the same month FRA withdrew its previous rulemaking—that considered data from 2008 through 2017. Id. at 45571 n.71; see U.S. Gov’t Accountability Off., Rail Safety: Freight Trains Are Getting Longer, and Additional Information Is Needed to Assess Their Impact 3, 11 (2019). In other Page 88words, there was no evidence that train lengths had increased between 2019 and 2024 such that a policy change was warranted. Plus, FRA did not specify how longer trains heightened safety risks, nor did it explain why increasing crew sizes would mitigate such risks. It merely posited that “this trend may have cascading safety impacts unless mitigated by technology, training, or other processes.” 2024 Final Rule, 89 Fed. Reg. at 25083 (emphasis added). FRA is entitled to rely on “predictive judgments” regarding safety trends, but such predictions “must be based on some logic and evidence, not sheer speculation.” Sorenson Commc’ns Inc. v. FCC , 755 F.3d 702, 708 (D.C. Cir. 2014) (quotations omitted).

Next, FRA asserted that “the rate for all human factor caused accidents has increased in recent years, notably after the 2019 withdrawal.” 2024 Final Rule, 89 Fed. Reg. at 25083. For support, it pointed generally to the Final Rule’s Executive Summary, which observed that “[t]he percentage of train accidents attributed solely to human factors . . . increased from 38.5 percent to 45.6 percent between 2013 and 2022.” Id. at 25053. But, again, FRA failed to link these human-factor-caused accidents to crew size or to demonstrate why adding additional crew members would reduce such accidents. See id. at 25083.

Finally, FRA determined that 'uncertainty related to new operating technologies can affect train safety,' and that current technological innovations had not yet been proven to obviate the need for a second crew member. Id. But once again, FRA offered no evidence that this was a post-2019 development: FRA cited thePage 89 2022 NPRM, which in turn cited a Department of Transportation report from 2009. See id. n.285 (citing 2022 NPRM, 87 Fed. Reg. at 45572–73). In fact, this 2009 report was one of the six reports that FRA had considered during its earlier rulemaking and found, in the 2019 Withdrawal Order, not to support a two-person rule. See 2019 Withdrawal Order, 84 Fed. Reg. at 24739–40 & n.34. So this report could not support a finding that changed conditions justified FRA’s changed position.

Put simply, each of FRA’s proffered reasons for the change relied on no more than mere speculation that requiring an additional crew member would address safety concerns. And '[w]e may not uphold agency action based on speculation.” Nat’l Sports Shooting Found., Inc. v. Jones , 716 F.3d 200, 214 (D.C. Cir. 2013). Because FRA failed to provide “a reasoned explanation . . . for disregarding facts and circumstances that underlay” the 2019 Order, its 2024 Final Order is arbitrary and capricious. Fox Television , 556 U.S. at 516.

“Agencies have long treated cost as a centrally relevant factor when deciding whether to regulate.' Michigan v. EPA , 576 U.S. 743, 752–53 (2015). “Consideration of cost reflects the understanding that reasonable regulation ordinarily requires paying attention to the advantages and the disadvantages of agency decisions.' Id. at 753 (emphasis in original). The Majority contends that FRA was not statutorily required to conduct a cost-benefit analysis. True. But FRA did perform one. And the D.C. Circuit has explained thatPage 90 “when an agency decides to rely on a cost-benefit analysis as part of its rulemaking, a serious flaw undermining that analysis can render the rule unreasonable' and thus arbitrary and capricious. Nat’l Ass’n of Home Builders v. EPA , 682 F.3d 1032, 1040 (D.C. Cir. 2012). “Nonetheless, we review such a cost-benefit analysis deferentially,” and the “burden to show error is high.” Id. (quotation omitted). And it is “up to the Agency to decide (as always, within the limits of reasonable interpretation) how to account for cost.” Michigan , 576 U.S. at 759. In this case, however, FRA’s analysis of the costs associated with the 2024 Final Rule contains a “serious flaw” that makes the Rule arbitrary and capricious: it fails to account for the labor costs associated with staffing every train with two people, rather than just one.

The 2024 Final Rule estimated that the 10-year costs of the Rule would be $6.6 million. 2024 Final Rule, 89 Fed. Reg. at 25057. In its estimate, it accounted for the direct compliance costs the regulatory scheme would impose, including 'costs associated with alerters, operating rules, notification to FRA, risk assessments and special approvals, annual reporting after receipt of special approval, and Government administration.” Id. The $6.6 million estimate did not, however, include any of the labor costs railroads would incur by operating with two-person crews where they previously would have been permitted to operate with one. In fact, FRA’s regulatory impact analysis explicitly stated that the Rule would “not affect railroads operating trains with two-person crews; therefore they will not incur any costs.” It acknowledged that railroads currently operating with two crew members that wanted specialPage 91 approval to operate with one would 'need to maintain current staffing levels” while awaiting special approval from FRA. And it “recognize[d] that a railroad may voluntarily abandon its one-person operation or refuse to improve its operation to prove it meets FRA's safety standard, creating a cost for additional crewmembers,” but it said that “the options in this final rule make that cost avoidable.'

The absurdity of requiring more labor but refusing to acknowledge the cost of that labor is highlighted by the fact that FRA did account for such costs in its first round of rulemaking. When it issued the 2016 NPRM, FRA's cost-benefit analysis accounted for costs “result[ing] from additional crew members on the trains that are currently operating with a one-person crew and from the delay or prevention of instituting new one-person crew operations in the future' (emphasis added). As a result, it concluded that costs associated with the proposed rule could be as high as $27.7 million over 10 years. FRA’s only explanation for why labor costs are no longer relevant is that the 2024 Final Rule, unlike the rule proposed in 2016, offers the possibility that railroads will receive special approval to operate with single-person crews. According to FRA, increased labor costs are only “hypothetical” when railroads may avoid them via the special approval process. But nothing about these costs is hypothetical: railroads who fail to receive special approval must continue to pay a second crew member when, absent the Rule, they would be able to shift to a one-person crew. And even railroads that ultimately receive special approval must 'maintain current staffing levels' while awaiting such Page 92approval. These are all costs to the railroads of meeting regulatory requirements, and FRA’s cost-benefit analysis accounted for none of them.⁶

FRA certainly may decide “how to account for cost,” Michigan , 576 U.S. at 759, but once it decides to perform a cost-benefit analysis, it may not simply disregard an entire category of obvious costs. The Final Rule relies heavily on the allegedly common-sense idea that having more crew members leads to greater safety while ignoring the obvious fact that having more crew members means more labor costs. The total failure of FRA to at least consider these costs is a “serious flaw” in the agency’s cost-benefit analysis that demands vacatur. Nat’l Ass’n of Home Builders , 682 F.3d at 1040.

Page 93For these two independent reasons, I would vacate FRA’s 2024 Final Rule as arbitrary and capricious.⁷ I therefore respectfully dissent.