BNSF Railway v. FRA

105 F.4th 691

Opinion

United States Court of Appeals for the Fifth Circuit

____________

United States Court of Appeals Fifth Circuit

No. 22-60217 ____________ FILED June 21, 2024

BNSF Railway Company, Lyle W. Cayce Clerk

Petitioner,

versus

Federal Railroad Administration; Amit Bose, in his official capacity as Administrator, Federal Railroad Administration; United States Department of Transportation,

Respondents.

______________________________

Petition for Review of an Order of the Federal Railroad Administration Agency No. 2020-64

______________________________

Before Jones, Smith, and Graves, Circuit Judges. Jerry E. Smith, Circuit Judge:

BNSF Railway Co. (“BNSF”) seeks an expanded waiver from the Federal Railroad Administration (“FRA”) with respect to BNSF’s use of automated track inspection (“ATI”). This petition for review returns to this panel after a limited remand. See BNSF Ry. Co. v. Fed. R.R. Admin. (“BNSF I”), 62 F.4th 905, 911–12 (5th Cir. 2023). After examining the FRA’s reconsidered decision declining to expand the scope of BNSF’s

No. 22-60217

existing waiver, we hold that decision to be arbitrary and capricious. Having already granted BNSF’s petition for review, we reverse the June 2023 decision and remand with instruction to expand BNSF’s existing waiver.

I.

A protracted procedural history can be found in the introduction to BNSF I. See id. at 908–10. When we last addressed this petition, we held that the FRA had failed to “articulate a satisfactory explanation for its action including a rational connection between the facts found and the choice made.” Id. at 910 (quoting Motor Vehicle Mfrs. Ass’n of the U.S. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983)) (internal quotation marks omitted ). We granted the petition for review, vacated the FRA’s decision, and remanded to the agency, noting “[t]his is a limited remand; this panel retains jurisdiction. We direct the FRA to enter its decision no later than one hundred days from the announcement of this opinion.” Id. at 911–12.

FRA responded in a letter dated June 21, 2023. It again denied BNSF’s request for a waiver, providing three reasons:

• that “[t]he public interest and railroad safety favor addressing these issues through the RSAC process,” June 2023 Letter at 5, • that “BNSF has not shown that an expanded waiver would improve railroad safety,” Id. at 7, • and that “BNSF’s implementation of the current waiver does not warrant an expanded waiver at this time,” Id. at 13.

In turn, BNSF asked this court to “direct FRA to modify the existing waiver.” Appellant’s Supp. Br. at 2.

During the pendency of this appeal, on February 14, 2024, the FRA filed a letter with this court indicating that the Railroad Safety Advisory Committee (“RSAC”) process—which sought to “develop a consensus

3

No. 22-60217

recommendation for incorporating ATI technology into the applicable regulatory framework,” BNSF I, 62 F.4th at 909 (cleaned up)—would terminate and that the “Working Group will not be able to reach a consensus recommendation on this issue.” Feb. 2024 Letter. In March, the FRA confirmed “the track inspection task ha[d] been closed.” Mar. 2024 Letter.

II.

The standard of review remains what it was the last time we saw this case. See BNSF I, 62 F.4th at 910–11. In short, “[w]e review final orders of agencies under the standard set out by the Administrative Procedure Act— we 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 law.” Id. at 910 (cleaned up).

It is true that “when an agency gives multiple reasons, we may uphold its decision based on any one of those reasons.” Tex. Tech Physicians Assocs. v. HHS, 917 F.3d 837, 844 n.4 (5th Cir. 2019) (citing Salt River Project Agric. Improv. & Power Dist. v. United States, 762 F.2d 1053, 1060 n.8 (D.C. Cir. 1985)). But the apparently broad language in Tex. Tech Physicians is rendered much narrower by context. The footnote that that panel cites stands for a considerably more circumscribed proposition: “When an agency relies on a number of findings, one or more of which are erroneous, we must reverse and remand only when there is a significant chance that but for the errors the agency might have reached a different result.” Salt River, 762 F.2d at 1060 n.8. Salt River articulates a correct, albeit more specific, statement of the law (and incidentally, the D.C. Circuit’s modern view 1).

_____________________

1 See Nat’l Fuel Gas Supply Corp. v. FERC, 468 F.3d 831, 839 (D.C. Cir. 2006)

(Kavanaugh, J.) (“[W]here [an agency] has relied on multiple rationales (and has not done so in the alternative), and we conclude that at least one of the rationales is deficient, we will

4

No. 22-60217

Moreover, even where multiple reasons are proffered, but the erroneous primary justification for a decision is reflected in “[a]lmost every part” of an agency’s decision, that “error permeates—and therefore infects—the entire” decision. See Chamber of Com. of U.S. v. SEC, 85 F.4th 760, 779 (5th Cir. 2023).

III.

As a threshold matter, the FRA contends that the Hobbs Act precludes our ability to continue to exercise jurisdiction over this case. We disagree.

The portion of the Hobbs Act relevant here, 28 U.S.C. § 2342(7), gives the courts of appeals jurisdiction to review certain final agency actions. And it refers to a specific way to invoke that jurisdiction. Id. But as in Castaneda-Castillo v. Holder, 723 F.3d 48, 64 (1st Cir. 2013), “the government has not pointed us to any language . . . in the Hobbs Act which would support its proposition that this court lacked authority to retain jurisdiction over the post-remand administrative proceedings that followed” our earlier decision.

Pointing to Abrams v. FDIC, 938 F.2d 22 (2d Cir. 1991), overruled on other grounds by Pa. Dep’t of Env’t Res. v. FDIC, 78 F.3d 97 (2d Cir. 1996), BNSF wants us to treat this case as returning from a remand under Federal Rule of Appellate Procedure 16. Appellant’s Reply Br. at 3. That’s not quite right. The court in Abrams both explicitly invoked Rule 16 and did not vacate the FDIC’s initial decision. Id. at 25–26.

The FRA would have us treat the June 2023 Letter as a wholly “new

_____________________

ordinarily vacate the order unless we are certain that [the agency] would have adopted it even absent the flawed rationale.” (citations omitted)).

5

No. 22-60217

decision.” That’s also not quite right. “New agency action” requires agencies to “comply with the procedural requirements for new agency action.” Dep’t of Homeland Sec. v. Regents of the Univ. of Cal., 591 U.S. 1, 21 (2020) (cleaned up). But here, there was no second notice-and-comment opportunity as required by 49 U.S.C. § 20103(d)(2)(C). Thus, this case seems to fall closer to the other camp described by Regents: a “remand for the agency to . . . offer a fuller explanation of the agency’s reasoning at the time of the agency action.” 591 U.S. at 20 (cleaned up).

The precise procedural posture in which we find ourselves is possibly unique: a limited remand directly to an agency enumerated in 28 U.S.C. § 2342(7), with the panel retaining jurisdiction post-vacatur. 2 There is no precise Fifth Circuit caselaw on this point.

BNSF rightly points out that we have retained jurisdiction over limited remands without requiring a new notice of appeal. See, e.g., Royal Bank of Canada v. Trentham Corp., 665 F.2d 515, 519 (5th Cir. Unit A Dec. 1981). It is also true that “[o]rders retaining jurisdiction ordinarily contemplate that the case will return to the court of appeals without a new notice of appeal.” 16 Charles Alan Wright et al., Federal Practice and Procedure § 3937.1, at 846 (2012) (footnote omitted). But, to the end of that statement, BNSF not so stealthily appends “or petition for review.” Appellant’s Reply Br. at 2. That addition is an obvious extension—and not one we accept uncritically, given the manifold differences between direct review of agencies and our ordinary review of district courts by way of appeal.

BNSF does not point to any in-circuit examples of limited remands to

_____________________

2 No party in this case nor this court sua sponte has identified an analogue. The closest case is Quincy Cable TV, Inc. v. FCC (Quincy Cable I), 730 F.2d 1549 (D.C. Cir. 1984) (per curiam), and its follow-on. See infra note 6.

6

No. 22-60217

agencies directly from this court, 3 let alone any cases doing so with vacatur. Indeed, BNSF I appears to be the first time this circuit has remanded and retained jurisdiction with vacatur. BNSF I, 62 F.4th at 911–12. 4 The FRA apparently noticed that novelty and now attaches a great deal of importance to our having vacated its original decision. The FRA contends that vacatur “re-establish[es] the status quo absent the unlawful agency action.” Texas v. United States, 40 F.4th 205, 220 (5th Cir. 2022) (per curiam). That, the FRA contends, resets the process entirely, requiring a new petition for review under the Hobbs Act. Appellee’s Supp. Br. at 7–8.

We disagree and conclude that we retain jurisdiction. First, the language from Texas v. United States refers to vacatur’s effect on an agency, not on a reviewing court’s jurisdiction. See 40 F.4th at 220 (“[V]acatur neither compels nor restrains further agency decision-making.”). Even if that

_____________________

3 Though, notably, one of this court’s opinions that issued after letter briefing provides an example. See Chamber of Com., 85 F.4th at 780 (granting the petition for review and remanded to the SEC while retaining jurisdiction). And in that case, our court did so without vacatur. See id. (“Short of vacating the rule, we therefore afford the agency limited time to remedy the deficiencies in the rule.”).

4 As a matter of precedent, the mere fact that we purported to retain jurisdiction in BNSF I does not resolve the jurisdictional issue in this case. Our court is not bound by implicit jurisdictional assumptions made in earlier cases. See, e.g., United States v. Garcia- Ruiz, 546 F.3d 716, 718 n.1 (5th Cir. 2008) (“The failure of . . . earlier panels . . . to discuss mootness does not yield an implication that those panels decided the cases were not moot, and we are not bound by any sub silentio determinations there.”); Kershaw v. Shalala, 9 F.3d 11, 13 n.3 (5th Cir. 1993) (“We do not consider our exercise of jurisdiction in [a prior case] to constitute a binding precedent, however, because the jurisdictional issue was neither raised by the parties nor addressed by the Court.”). Cf., e.g., Steel Co. v. Citizens for a Better Env’t, 523 U.S. 83, 91 (1998) (“[D]rive-by jurisdictional rulings . . . have no precedential effect.” (citations omitted)).

7

No. 22-60217

language were read in the broadest possible sense, it does not account for our panel’s explicitly retaining jurisdiction. The FRA identifies no language in a case or a statute that specifically abrogates our ability to retain jurisdiction.

Second, although combining limited remand with vacatur is unusual, it does happen. For example, in Ucelo-Gomez v. Gonzales, 464 F.3d 163, 172 (2d Cir. 2006) (per curiam), the court issued a remand like ours except to the BIA instead of the FRA. The court vacated a BIA order, and the BIA issued a responsive opinion within a specified timeframe. See Ucelo-Gomez v. Mukasey, 509 F.3d 70, 71–72 (2d Cir. 2007). The original panel then exercised jurisdiction to review the opinion. See id. The panel seemed to justify that posture by citing United States v. Jacobson, 15 F.3d 19, 21–22 (2d Cir. 1994), though that case dealt with an informal remand to the district court. See Ucelo-Gomez, 509 F.3d at 72. 5 It is true that 28 U.S.C. § 2342 does not implicate the BIA as it does the FRA in this case. But as we note above, there is nothing in the relevant portions of the Hobbs Act that prevents a panel from issuing this sort of a limited remand in this case. 6

_____________________

5 See also Asani v. INS, 154 F.3d 719 (7th Cir. 1998) (similarly structured remand to the BIA though retained jurisdiction did not appear to be used).

6 Even if the language of 28 U.S.C. § 2342 were relevant to this inquiry, we might look to Quincy Cable I, 730 F.2d at 1549, where the court issued a similarly structured remand to the FCC—an agency listed in an analogous part of the Hobbs Act, see 28 U.S.C. § 2342(1). The panel retained jurisdiction and remanded to the FCC to issue a “final determination” based on “current facts.” 730 F.2d at 1551. For reasons unknown to us, the panel after remand did differ by one judge. See Quincy Cable TV, Inc. v. FCC (Quincy Cable II), 768 F.2d 1434 (D.C. Cir. 1985).

Unfortunately, Quincy Cable does not cleanly resolve this case because of several meaningful differences between that case and this one. First, Quincy Cable I did not include vacatur. Second, it is unclear whether the court in Quincy Cable II required a new petition. See 768 F.2d at 1447 (referring to “[t]his petition,” thereby implying the court was reviewing a distinct petition from the petition in Quincy Cable I). Third, even if there were

8

No. 22-60217

In short, we retain jurisdiction over this case. That is in line with how we generally treat limited remands, and nothing in the Hobbs Act abrogates our jurisdiction. 7

IV.

It is not evident that any of the FRA’s proffered reasons stands apart as an alternative and independent ground for denial of the waiver. Instead, “there is a significant chance that but for the errors the agency might have reached a different result,” Salt River, 762 F.2d at 1060 n.8. Under Salt River, we need not reach every error here. Even so, we do. The FRA fails to provide even one rationale that survives our review.

A.

The first reason proffered by the FRA is that “[t]he public interest and railroad safety favor addressing these issues through the RSAC process.” June 2023 Letter at 5. That reason “permeates—and therefore infects—the entire” decision. Chamber of Com., 85 F.4th at 780.

In its original decision, FRA merely pointed to the then-ongoing RSAC process and called it a day. See March 2022 Letter at 2–3. We sent the case back to them, implicitly denying the sufficiency of that explanation. See BNSF I, 62 F.4th at 911–12. In its reconsidered decision, the FRA again principally relied on the then-ongoing RSAC process. See June 2023 Letter

_____________________

a second petition in Quincy Cable II, it is not clear that that petition was jurisdictionally required.

7 The FRA would have us order full merits briefing. We reject that request. The FRA relies on Grant v. Cuellar, 59 F.3d 523, 525 (5th Cir. 1995), and United States v. Sineneng-Smith, 590 U.S. 371 (2020). Those cases are about party presentation, but the FRA has not identified in what sense they have been deprived of an “opportunity to address fully all the issues.” Grant, 59 F.3d at 525. As the history of this case demonstrates, the FRA has had ample time to speak to all the issues at play.

9

No. 22-60217

at 5–7. This time, however, it explained why waiting on the RSAC process was important—Congress’s preference for national uniformity in railroad safety regulations. See id. at 6.

Crucially, the FRA now indicates that that very RSAC process terminated without any consensus recommendation. See Feb. 2024 Letter; Mar. 2024 Letter. That independently renders the FRA’s principal explanation for denying the waiver worthless. The FRA had suggested in its reconsidered decision that “if the RSAC process does not produce consensus recommenddations , FRA will consider developing a proposed regulation on this issue.” June 2023 Letter at 6. But “agencies cannot play the administrative law shellgame of offering future rulemaking as a response to a claim of agency illegality .” El Paso Elec. Co. v. FERC, 76 F.4th 352, 366 (5th Cir. 2023) (internal quotation marks and citation omitted).

Even before this news, BNSF was right that the FRA’s appeal to uniformity “fall[s] flat.” Appellant’s Supp. Br. at 3. BNSF points out fatal flaws in FRA’s reasoning. First, FRA “never explains why the purported interest in uniformity did not prevent it from granting the original waiver.” Id. Second, FRA “never addresses the fact that the ‘RSAC process’ was already ongoing when it granted BNSF the original waiver.” Id.

We agree with this criticism. Merely reciting a “uniformity” rationale in conjunction with its original, already-rejected reasoning cannot fix the FRA’s original decision. That is especially true where we ordinarily insist that an agency at least express awareness of a change in position. See, e.g., FCC v. Fox Television Stations, Inc., 556 U.S. 502, 515 (2009).

B.

The second reason proffered by the FRA for denying BNSF’s waiver is that “BNSF has not shown that an expanded waiver would improve railroad safety.” June 2023 Letter at 7. This reason is twice inadequate.

10

No. 22-60217

1.

The parties dispute the standard under which the relationship between railroad safety and the waiver should be evaluated. FRA insists that BNSF has to “show[] that expanding the waiver would improve railroad safety.” June 2023 Letter at 7. But, as BNSF points out, the proper legal standard for granting a waiver is not whether it would improve public safety but, instead, whether the waiver “is in the public interest and consistent with railroad safety.” 49 U.S.C. § 20103(d)(1) (emphasis added); see Appellant’s Supp. Br. at 6.

BNSF has the better of this argument. The statutory text is clear.

Applying the wrong statutory standard renders a rationale arbitrary and capricious. 8

2.

Even assuming, arguendo, that FRA applied the correct legal standard, the reconsidered decision is still arbitrary and capricious. The FRA correctly frames the two worlds to be compared: (1) a world with automated inspections and a full load of visual inspections and (2) a world with automated inspections but reduced visual inspections. See June 2023 Letter at 7–8. Since BNSF can use automated inspections even without a waiver, the question is whether a reduction in visual inspections “is in the public interest and consistent with railroad safety.” 49 U.S.C. § 20103(d)(1).

FRA says it is not. It notes that automated inspections “cannot directly detect concerns pertaining to track drainage, vegetation conditions, crosstie conditions, joint bar defects, and track components such as switches

_____________________

8 See, e.g., Gen. Land Office of Tex. v. U.S. Dep’t of Interior, 947 F.3d 309, 321 (5th Cir. 2020) (holding that an agency’s basing a decision on the “incorrect legal standard” was arbitrary and capricious).

11

No. 22-60217

and crossings.” June 2023 Letter at 8. BNSF responds that some of those concerns can be addressed by the bi-monthly visual inspections that were still required under the original waiver. See Appellant’s Supp. Br. at 6. BNSF also notes that the FRA considered these concerns and rejected them in its January 2021 waiver grant. BNSF rightly points out that “FRA’s June 21 letter ruling grapples with none of its prior reasoning and conclusions.” Appellant’s Opening Br. at 6 (emphasis omitted). The decision in Fox is plainly applicable here:

[W]hen, for example, its new policy rests upon factual findings that contradict those which underlay its prior policy . . . [i]t would be arbitrary or capricious to ignore such matters. In such cases it is not that further justification is demanded by the mere fact of policy change; but that a reasoned explanation is needed for disregarding facts and circumstances that underlay or were engendered by the prior policy.

556 U.S. at 515–16 (citations omitted). Under Fox, FRA’s failure to address its prior reasoning is easily arbitrary and capricious.

C.

The FRA’s third proffered reason is that “BNSF’s implementation of the current waiver does not warrant an expanded waiver at this time.” June 2023 Letter at 13. This reason is also deficient in two ways.

1.

First, absent any other reasoning—and assuming arguendo that the FRA is right—this is not a freestanding reason to deny BNSF’s waiver.

Imagine the following: ASDF Railway Co. applies for a waiver on the same facts as BNSF except its lacks a pre-existing understanding with the FRA. The FRA responds, “We have denied your request because we never previously promised you an expansion.” That would fail to “articulate a sat-

12

No. 22-60217

isfactory explanation for its action including a rational connection between the facts found and the choice made.” BNSF I, 62 F.4th at 910 (cleaned up).

2.

Even if it were a substantively independent reason to reject BNSF’s waiver, this reason is inadequate in another way.

BNSF argues that the FRA impermissibly “chang[ed] the conditions it established for expanding the waiver.” Appellant’s Supp. Br. at 4. For its part, “FRA recognizes that, when it granted the original waiver, it anticipated that BNSF could petition to expand the waiver, if implementation was successful.” June 2023 Letter at 14. FRA contends that this recognition did not limit it to merely considering BNSF’s performance under the waiver. Id. It also asserts the BNSF should have been aware FRA would have to review a new petition with “consideration of the legally required factors” and points out that BNSF has not identified any costs of reliance. Id.

BNSF points out (1) that FRA, in its 2021 grant, “provid[ed] specific conditions, which if met, will allow BNSF to expand implementation,” and (2) that FRA recognizes in the June 2023 denial that such a denial “may entail economic costs for BNSF,” June 2023 Letter at 10. FRA does not seem to argue that these conditions were not met, and in BNSF we concluded that “[t]he implementation of ATI pursuant to the prior waiver appears to have been an unqualified success.” BNSF I, 62 F.4th at 911.

BNSF also has the better of this argument. It is true that there has been no clear accounting of BNSF’s reliance interests. But it is also eminently reasonable for BNSF to rely on a specific condition that guaranteed a future expansion of the waiver. An agency is “required to assess whether there were reliance interests, determine whether they were significant, and weigh any such interests against competing policy concerns.” Regents,

13

No. 22-60217

591 U.S. at 33. FRA obviously failed to do that. 9

V.

As for remedy, BNSF asks us “to direct FRA to comply with its prior decision giving BNSF a ‘path’ for expanding the scope of the existing waiver, under the same terms and conditions the agency has already determined would be ‘in the public interest and consistent with railroad safety.’” Appellant ’s Supp. Br. at 2 (citations omitted). We grant that request.

Ultimately, whether we can compel the FRA in this way turns on whether the FRA is “required” to reach the result of granting the waiver, or whether this is a “discretionary judgment.” Calcutt v. FDIC, 598 U.S. 623, 630 (2023) (per curiam) (emphasis omitted). We need not remand an inadequate decision for reconsideration where there is “no realistic possibility that the [agency] would reach another outcome.” Maniar v. Garland, 998 F.3d 235, 240 (5th Cir. 2021) (citation omitted). Where “[t]o remand would be an idle and useless formality,” we need not “convert judicial review of agency action into a ping-pong game.” Morgan Stanley Cap. Grp. Inc. v. Pub. Util. Dist. No. 1, 554 U.S. 527, 545 (2008) (quoting NLRB v. Wyman-Gordon Co., 394 U.S. 759, 766–67, n.6 (1969) (plurality)). 10 Importantly, the question is not whether the agency has the statutory authority to deny BNSF’s waiver request. 11 Instead, the query is whether the

_____________________

9 Moreover, this language from Regents suggests that it is the FRA’s responsibility, and not BNSF’s, to account for reliance interests.

10 We fully appreciate the differences that this case has with Wyman-Gordon and Maniar; we do not suggest that either of those cases is controlling here. Inter alia, the courts in those cases upheld the agency ruling; here, we reverse it. That said, the principle in those cases is applicable: We need not remand for reconsideration of an agency decision that can have only one outcome.

11 Though it is worth noting that our court seems to have remanded to an agency

14

No. 22-60217

conditions for expanding BNSF’s waiver were adjudicated by the agency in 2021. It is true that “[a]n agency is not forever bound by its prior determinations .” Creppel v. U.S. Army Corps of Eng’rs, 670 F.2d 564, 571 (5th Cir. 1982) (emphasis added). It is also true that adjudications deal with a specific dispute between specific parties. See McDonald v. Watt, 653 F.2d 1035, 1041– 42 (5th Cir. Unit A Aug. 1981) (citing United States v. Fla. E. Coast Ry. Co., 410 U.S. 224, 245 (1973)). Our holding in this case does not run afoul of any of those constraints. This adjudication binds only the parties involved (the FRA and BNSF) regarding a specific dispute (the track-inspection waiver) for a time specified by the agency (five years). 12 The language from the FRA’s 2021 waiver is certain: “The Board is . . . providing specific conditions, which if met, will allow BNSF to expand implementation of the relief in a consistent and safe manner.” (Emphasis added.) It is therefore undeniable that part of the FRA’s 2021 waiver was a set of conditions under which the FRA would be compelled to expand the waiver’s scope. That “waiver is valid for 5 years from the date of [the] letter .” We are well within that window. The FRA—even with a second bite at the apple—fails to explain why it can now escape its promise to expand BNSF’s waiver conditioned on the success the earlier waiver. Therefore, given that “[t]he implementation of ATI pursuant to the prior waiver appears to have been an unqualified success,” BNSF I, 62 F.4th at 911, the only remaining path for the FRA is to expand the waiver.

_____________________

with instructions to reverse its decision even where the underlying statutory language is discretionary. See Hall v. FERC, 691 F.2d 1184, 1196–99 (5th Cir. 1982).

12 We do not opine as to whether the agency can or did reopen its 2021 adjudication.

The FRA does not suggest that, either in its 2023 decision or its briefing before this court. That argument is waived. See Kubala v. Supreme Prod. Servs., 830 F.3d 199, 202 n.1 (5th Cir. 2016) (“[I]ssues not briefed on appeal are waived.” (citation omitted)).

15

No. 22-60217

“To remand [without specific instruction] would be an idle and useless formality.” Morgan Stanley, 554 U.S. at 545 (citation omitted). There is no ambiguity as to what the FRA must do on a second remand. We need not doom BNSF to an endless loop of regulatory activity. In such circumstances, it is appropriate to remand with instruction to grant the waiver expansion rather than “convert judicial review of agency action into a ping-pong game.” Id. (citation omitted). 13

* * * *

In summary, having heretofore granted the petition for review, we REVERSE the FRA’s June 2023 reconsidered decision and REMAND with instruction to expand BNSF’s existing waiver.

_____________________

13 In this case, judicial review is much more like tennis. One faulty serve is given grace. Two are not.

16

No. 22-60217

James E. Graves, Jr., Circuit Judge, concurring in part and dissenting in part:

This panel initially concluded that the FRA failed to fully articulate a rational reason for denying BNSF’s expanded waiver. After limited remand, it is clear now that the agency’s original denial was fatally flawed from the start. Accordingly, I agree with the majority that the FRA’s denial of the expanded waiver was improper.

But I cannot join the remainder of the majority’s analysis nor its ultimate disposition of this petition. I find unnecessary and overly constraining the majority’s conclusion that the FRA abused its power by reaching a different conclusion on one waiver petition than it had on another. Compounding its error, the majority orders the agency to grant the waiver without briefing on the implications of that order and without due regard for the agency’s safety expertise.

Accordingly, while I agree that the FRA’s denial of BNSF’s expanded waiver fails to pass muster, I respectfully dissent from the remainder of the majority opinion, which wrongly undercuts the agency’s authority.

I.

The FRA is entrusted with the safety of America’s railroads. In furtherance of that responsibility, its regulations require railroad companies to assign workers to visually inspect tracks for hazardous defects according to a prescribed schedule, on foot or aboard a slow-moving vehicle. 49 C.F.R. §§ 213.233(b),(c).

New technology offers improvement over this system. Automated track inspection, or ATI, uses a laser that scans tracks for defects and, the parties agree, is significantly more effective than visual inspections at finding most defects. Given that benefit and others, the FRA does not restrict railroad companies from employing ATI. See 49 C.F.R. § 213.233(b). But it

17

No. 22-60217

also has not allowed them to replace visual inspections with ATI. One concern, which has been persistently expressed by the railroad workers’ union, is that ATI is less adept than visual inspection at spotting certain track hazards. 1 Some have advocated for a measured and incremental approach to scaling down visual inspections, particularly in rail corridors where high volumes of hazardous materials are transported or where more people live. 2 Wary of rolling out the technology too quickly, the FRA initiated a piecemeal test program through which railroads could apply for regionspecific waivers of the agency’s visual inspection requirements. Federal law allows the FRA to grant such waivers if doing so is “in the public interest and consistent with railroad safety.” 49 U.S.C. § 20103(d)(1).

In 2018, BNSF petitioned to reduce visual inspections on seven segments of track—1,348 miles total—in a rural territory covering parts of Nebraska and Wyoming. The FRA granted BNSF’s petition, explaining that it was instituting a “limited, temporary suspension” of 49 C.F.R. § 213.233(c) “as necessary to carry out the Test Program.” 83 Fed. Reg. at 55,449.

BNSF applied for another waiver in July 2020 to roll out ATI systemwide. In January 2021, the FRA responded by granting a waiver limited to BNSF’s Southern Transcon route between Chicago and Los Angeles, covering an additional 4,635 miles of track. The agency explained that “unrestrained system-wide implementation” was not yet appropriate because of “the different weather, geographical, and operational challenges

_____________________

1 Comments from the Brotherhood of Maintenance of Way Employes Division (Aug. 22, 2021), https://downloads.regulations.gov/FRA-2020-0064- 0017/attachment_1.pdf; Appellant’s Br. at 7.

2 Comments from the Association of State Railroad Safety Managers (Sept. 25, 2020), https://downloads.regulations.gov/FRA-2020-0064-0006/attachment_1.pdf.

18

No. 22-60217

found throughout BNSF’s rail system.” Instead, the agency explained, BNSF should focus its resources on monitoring the waiver territories “to avoid unnecessary risk.” But BNSF could petition to add new territories “contingent on successful implementation.”

BNSF requested another expanded waiver in June 2021 for its Northern Transcon route, spanning from Chicago to Seattle, and its Orin Subdivision territory, in Wyoming—another 4,717 track miles in total. After notice and comment, the FRA denied the petition. It informed BNSF that its Railroad Safety Advisory Committee (“RSAC”), a group established to make safety recommendations, was developing “a consensus recommendation for incorporating ATI technology into the applicable regulatory framework” and “short-circuiting this evaluation process on individual railroads is not in the public interest and consistent with railroad safety at this time.” BNSF Ry. Co. v. FRA (“BNSF I”), 62 F.4th 905, 910 (5th Cir. 2023).

BNSF petitioned our court for review. This panel concluded that we could neither approve nor reject the FRA’s sparse rationale. Quite simply, the agency had made little-to-no effort to articulate a “rational connection” between its denial of the waiver and its consideration of the public interest and railroad safety. Id. at 911 (quoting Motor Vehicle Mfrs. Ass’n of U.S., Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983)). The importance of the RSAC process in furthering railroad safety or the public interest was unclear at best. We remanded to give the FRA another shot at explaining its decision.

The FRA issued a new denial letter to BNSF in , June 2023 that was substantially longer and more detailed than its 2022 denial. The rationale there—at least, the rationale that we are permitted to review—boils down to one word: uniformity. We are tasked now with determining whether

19

No. 22-60217

uniformity constitutes the rational connection that was missing from the FRA’s initial denial.

II.

First, I join the majority’s analysis of the jurisdictional question raised by the FRA. The agency failed to point to authority that would prohibit this panel from picking up the case where we left it. Moreover, retaining jurisdiction makes sense given the specific issues that the panel identified and our experience with the case.

I also agree that because the June 2023 letter does not meet the requirements of a new agency action, we should treat it as a “fuller explanation” of the FRA’s original waiver denial. See DHS v. Regents of the Univ. of Cal., 591 U.S. 1, 20 (2020). I therefore would proceed, as the majority does, to evaluate whether the FRA’s fuller explanation meets the rational connection test.

In my view, because the FRA’s June 2023 letter was a fuller explanation and not a new agency action, we must not review any rationale in the letter that does not aim only to clarify the agency’s original reason for denial, i.e., the RSAC process. See Texas v. United States, 40 F.4th 205, 226 (5th Cir. 2022); see also Regents, 591 U.S. at 21 (“When an agency’s initial explanation indicates the determinative reason for the final action taken, the agency may elaborate later on that reason (or reasons) but may not provide new ones.”) (cleaned up). Only the first rationale in the letter—the uniformity rationale, directly invoking the RSAC process—is up for review. See Texas, 40 F.4th at 226–27 (“‘[A]n agency’s action must be upheld, if at all, on the basis articulated by the agency itself,’ not reasons developed post hoc.”) (quoting Motor Vehicle Mfrs., 463 U.S. at 50).

Even though the RSAC process has concluded, we are not foreclosed from considering that rationale. As the FRA explained in the June 2023 letter,

20

No. 22-60217

RSAC was to “develop[] recommendations for ATI regulations that would apply nationally.” Even though the RSAC process has concluded, the agency states that it is “continuing to consider a comprehensive approach to” ATI through rulemaking. Because the RSAC and rulemaking processes implicate the same “comprehensive” uniformity interest, and rulemaking remains on the table, we should evaluate the merits of the uniformity rationale.

Ultimately, however, the FRA fails to explain exactly how uniformity justifies denying BNSF’s expanded waiver. Uniformity is about railroad safety, the agency wrote in the June 2023 letter. RSAC and rulemaking, it wrote, further uniformity because they involve a “comprehensive technical review” where the FRA considers the perspectives of all stakeholders and examines ATI data from its test programs “on a national basis.” Expanding BNSF’s waiver “in a piecemeal fashion” would be contrary to those efforts.

But the FRA does not explain how expanding BNSF’s waiver is inconsistent with comprehensive review or nationwide coordination. Nor does it explain the relationship to railroad safety. That was the FRA’s job on remand. See FCC v. Fox Television Stations, Inc., 556 U.S. 502, 513–14 (2009).

Nor can we accept uniformity for the sake of uniformity. The FRA argues that Congress expressed a generic preference for uniformity when it passed the Federal Railroad Safety Act. But that statute represented Congress’s intent “to promote safety in every area of railroad operations and reduce railroad-related accidents and incidents.” Lane v. R.A. Sims, Jr., Inc., 241 F.3d 439, 442 (5th Cir. 2001) (quoting 49 U.S.C. § 20101) (emphasis added); see also 49 U.S.C. § 20106. Again, the FRA fails to identify a nexus between safety, the uniformity interest advanced by RSAC and rulemaking, and the denial. Moreover, if uniformity alone is acceptable as an overriding interest, any waiver of the FRA’s regulations would be impossible to justify, despite the FRA’s clear authority to grant them. See 49 U.S.C. § 20103(d)(1).

21

No. 22-60217

In the end, the FRA essentially conceded that RSAC and rulemaking do not justify denying the expanded waiver. In the June 2023 letter, the FRA admitted that “if BNSF develops evidence that expanding the waiver would improve railroad safety . . . FRA would consider a renewed waiver application during the pendency of the RSAC and/or rulemaking process.” But it was the RSAC process that the FRA originally cited in rejecting BNSF’s waiver petition; it is, therefore, the only reason we may review now. See Texas, 40 F.4th at 226. If the RSAC process and the related rulemaking process do not bar the granting of a waiver, then we are left with one conclusion: the FRA acted arbitrarily and capriciously by resting its waiver denial on that ground.

I join the majority in concluding that the FRA’s denial of BNSF’s petition was arbitrary and capricious. The agency failed to demonstrate a reasonable connection between its interest in uniformity, as represented by the RSAC and rulemaking processes, and its denial of BNSF’s expanded waiver.

III.

I cannot sign on to the majority’s alternative conclusion, however, that because the FRA granted some of BNSF’s prior waiver petitions, it was obligated to grant another.

First, the majority faults the FRA for failing to “display awareness that it chang[ed] position.” Fox, 556 U.S. at 515. In the majority’s view, the FRA changed its position by making promises to BNSF that it failed to keep.

It is not accurate, in my view, to interpret the FRA’s denial of BNSF’s expanded waiver as a change in position. When the agency first granted BNSF’s waiver, it did not, as the majority writes, “guarantee a future expansion of the waiver.” The majority relies on language in the FRA’s January 2021 letter granting BNSF’s second waiver. The agency wrote that “[the FRA] is providing specific conditions, which if met, will allow BNSF

22

No. 22-60217

to expand implementation of the relief in a consistent and safe manner.” The agency also wrote that “BNSF may petition to include other territories in the waiver, contingent on successful implementation” of the ATI Program under the second waiver.

Those sentences simply do not create a guarantee. They set forth a condition necessary for BNSF to petition to expand the waiver, but not a sufficient one. Other parts of the letter make clear that the FRA’s approval of subsequent waivers was not solely contingent on whether BNSF satisfied those conditions. The agency noted, for example, that weather or geographical concerns in certain parts of the country might limit expansion. It also warned that the test program was “limited” and “temporary” and emphasized that any expansion of BNSF’s waiver would be incremental. It did not promise that BNSF would be permitted to expand the waiver system- wide, or to certain territories, or on a certain timeline.

Moreover, the conclusion that the FRA did obligate itself to grant future waivers, if taken to its logical end, would lead to the absurd. Under BNSF’s reading, the FRA must grant an automatic expansion to BNSF for any track area, at any time, regardless of other conditions. That would surely constitute an abdication of the FRA’s responsibility to consider safety and the public interest. But it is reasonable that the agency responsible for the safety of all the country’s railroads would find it prudent, in the name of safety and the public interest, to be cautious of implementing new technology too rapidly. Indeed, the FRA could have cited such concerns, about the types of safety issues that only become apparent over the course of years, when it first denied BNSF’s petition. But that did not happen.

The FRA does not have the authority to guarantee outcomes on waiver petitions. It is statutorily required to examine each waiver petition on its own merits to determine if it is in the public interest and consistent with

23

No. 22-60217

railroad safety. 49 U.S.C. § 20103(d)(1). Its regulations require waivers to be “limited in scope and application to such relief as may be necessary to facilitate the conduct of the test program.” 49 C.F.R. § 211.51(a)(2). It could not have granted a waiver that did not meet those criteria even if it had told BNSF that it would.

Next, the majority faults the agency for failing “to assess whether there were reliance interests, determine whether they were significant, and weigh any such interests against competing policy concerns.” Ante at 13 (quoting Regents, 591 U.S. at 33). But the majority’s analysis omits an element of the principle from Regents. There, the Supreme Court wrote: “When an agency changes course . . . it must ‘be cognizant that longstanding policies may have engendered serious reliance interests that must be taken into account.’” Id. at 30 (quoting Encino Motorcars, LLC v. Navarro, 579 U.S. 211, 212 (2016)) (emphasis added).

The FRA’s previous granting of two other waivers was not the kind of “longstanding policy” that courts have deemed sufficient to generate significant reliance interests. Regents concerned the Department of Homeland Security’s wholesale rescission of the DACA program—a sustained national policy that governed the legal status of some 700,000 undocumented people. Id. at 18, 31–32. Encino Motorcars concerned a threedecade -old Department of Labor policy that governed overtime pay for auto sales employees across the nation. Encino Motorcars, LLC v. Navarro, 579 U.S. 211, 222–23 (2016). The D.C. Circuit’s decision in Southwest Airlines v. FERC concerned an agency’s deviation from its “consistent practice” over more than twenty years of using particular data to decide the appropriateness of oil pipeline rate increases. 926 F.3d 851, 853 (D.C. Cir. 2019).

This case, by contrast, is not about a policy sustained over an extended period of years or applicable to a broad class of individuals or businesses. This

24

No. 22-60217

case is about two discrete, individualized waiver determinations that applied to a single business.

Moreover, requiring the agency to scrupulously account for all conceivable reliance interests any time it acts within its statutory authority would prevent it from acting flexibly in pursuit of its mandates. Circumstances change. “[A]gencies ‘must be given ample latitude to adapt their rules and policies to the demands of changing circumstances.’” FDA v. Brown & Williamson Tobacco Corp., 529 U.S. 120, 157 (2000) (quoting Motor Vehicle Mfrs., 463 U.S. at 42).

The FRA’s actions were also predictable. BNSF knew it was participating in a regional “test program,” the ultimate goal of which was to create a nationwide regulatory framework. 3 It had no reason to expect that it was entitled to avoid the FRA’s discretionary review.

The FRA’s crucially important mandate is to consider—and, when necessary, reconsider—the safety and public interest impact of any new railroad technology. See 49 U.S.C. § 20103(d)(1). That includes ATI, with its significant potential to improve our railroad system for the better. We should not undermine that mandate by forcing the agency to strictly adhere to its earlier actions. As the majority itself notes, “[a]n agency is not [to be] forever bound by its prior determinations[,] for they are neither congressional directives nor scriptural admonitions.” Creppel v. U.S. Army Corps of Eng’rs, 670 F.2d 564, 571 (5th Cir. 1982).

IV.

Now that we have concluded that the FRA’s rationale for denying BNSF’s expanded waiver was arbitrary and capricious, we should require

_____________________

3 See, e.g., RSAC, Track Standards Working Group Update (Dec. 8, 2021) at 2.

25

No. 22-60217

briefing from the parties on whether to direct the FRA to grant the waiver or to remand. Post-remand briefing has been minimal. We asked the parties only to submit letter briefs explaining what we should do in light of the June 2023 letter. The party presentation principle, even if not binding here, counsels in favor of hearing the parties’ views as to our next course of action. See United States v. Sineneng-Smith, 590 U.S. 371, 375 (2020) (“[W]e rely on the parties to frame the issues for decision . . . .”) (quoting Greenlaw v. United States, 554 U.S. 237, 243 (2008)).

If we must act now, remand is more appropriate. Remand is the appropriate course except in “rare” circumstances, and even then, only when “the agency will inevitably adhere to its prior decision.” BizCapital Bus. & Indus. Dev. Corp. v. Comptroller of Currency of U.S., 467 F.3d 871, 874 (5th Cir. 2006). The majority concludes that the FRA’s granting of the expanded waiver is inevitable because BNSF satisfied the conditions in the January 2021 letter. As I stated above, I disagree that BNSF’s satisfaction of those terms entitles it to an expanded waiver. Moreover, the FRA disputes that BNSF has satisfied the terms. It argues, convincingly, that it cannot fully evaluate BNSF’s performance because BNSF has not reduced visual inspections to the level permitted under the waiver.

Further, the FRA has not made reviewable findings on other material aspects of the expanded waiver that seem to have clear safety and public interest implications. For example, there should be some consideration of how the Northern Transcon territory’s geography and weather affects the appropriateness of an expanded waiver. The same goes for whether, as it argues, the FRA should engage in a slow, incremental rollout of its ATI waivers, given that negative effects of reducing visual inspections could take years to spot. Those determinations are better left to the FRA’s expertise. See Ctr. for Biological Diversity v. EPA, 749 F.3d 1079, 1087–88 (D.C. Cir. 2014) (“We must look at the decision not as the chemist, biologist or

26

No. 22-60217

statistician that we are qualified neither by training nor experience to be, but as a reviewing court exercising our narrowly defined duty of holding agencies to certain minimal standards of rationality.”) (quoting Ethyl Corp. v. EPA, 541 F.2d 1, 36 (D.C. Cir. 1976) (en banc)). Accordingly, remand is appropriate. See Hall v. FERC, 691 F.2d 1184, 1198 (5th Cir. 1982) (concluding that remand was appropriate when agency had outstanding matters to consider).

I acknowledge that a remand might make the current process more tedious. But given this case’s unusual posture, it seems appropriate. We originally remanded to the FRA to “provide further justification for its rejection” of BNSF’s expanded waiver. BNSF I, 62 F.4th at 911. The agency followed the spirit of that order by offering a fuller explanation of its RSAC justification rather than engaging in a new agency action. Now, because that RSAC justification is unavailing, a new agency action is required. See Biden v. Texas, 597 U.S. 785, 811 (2022) (explaining that in undertaking new agency action, “the agency is ‘not limited to its prior reasons’ in justifying its decision”) (quoting Regents, 591 U.S. at 21).

As inconvenient as another remand may be, it is preferable to a train derailment caused by a track defect that would have been spotted but for the lack of a visual inspection. In short, when railroad safety is at issue, we should generally let America’s railroad safety authority make the call.

V.

I concur with the majority opinion insofar as it rejects the FRA’s fuller explanation of its original denial of BNSF’s waiver petition. As to the remainder, I respectfully dissent.