On Petition for Rehearing

TATEL, Circuit Judge:

This petition for rehearing asks that we reconsider one aspect of our earlier opinion denying petitions for review of a Surface Transportation Board regulation that provides two simplified methods for resolving rail rate disputes too small to bring under ordinary procedures. As part of its challenge to the regulation, Petitioner Norfolk Southern argued that in violation of the Administrative Procedure Act the Board's notice of proposed rulemaking had failed to give notice of a significant change that surfaced only in the final rule. Given that no party had presented that argument to the Board in a petition for reconsideration, we declined to consider it. Because we now agree with Norfolk Southern that we should have addressed the issue, and because we conclude that the Board failed to provide adequate notice, we vacate the relevant portions of the regulation, as well as of our earlier opinion.

I.

Our earlier opinion describes the background of this case and the two simplified methods for resolving rail rate disputes. CSX Transp., Inc. v. Surface Transp. Bd., 568 F.3d 236, 238-40 (D.C. Cir. 2009). The three-benchmark method, the one at issue here, compares the challenged rate to three benchmark figures, one of which — the R/VC^COMPM — is derived by comparing the rail movement at issue with a group of similar movements. The Board selects this comparison group from groups of movements the parties propose, and the parties in turn choose the comparison movements from a survey of movements across the nation, the so-called waybill sample. Under the proposed rule, parties could suggest comparison groups drawn from the most recent year of waybill sample data. Simplified Standards for Rail Rate Cases (" NPRM"), STB Ex Parte No. 646 (Sub-No. 1), at 33 (served July 28, 2006) (notice of proposed rulemaking). Under the final rule, however, parties may draw from the four most recent years of data. Simplified Standards for Rail Rate Cases (" Decision" or "final rule"), STB Ex Parte No. 646 (Sub-No. 1), at 80, 83 (served Sept. 5, 2007).

In their petition for review, Norfolk Southern and several other railroads argued that the proposed rule violated the Administrative Procedure Act because it failed to provide notice that the Board was considering switching from one year to four years' worth of data. See 5 U.S.C. § 553(b)(3) (requiring agencies to give notice of "the terms or substance of the proposed rule or a description of the subjects and issues involved"). As petitioners pointed out, nothing in the NPRM expressly indicated that the Board was considering expanding the data from which parties can draw comparison groups. The Board did not argue otherwise. Yet because the railroads had failed to present this argument to the Board, we declined to address it. CSX, 568 F.3d at 246-47. Although the railroads argued that they had no way of knowing of any lack of notice until the Board promulgated its final rule, we pointed out that they could have presented the argument to the Board in a petition for reconsideration. See 49 C.F.R. § 1110.10. In so ruling, we acknowledged the holding of Darby v. Cisneros, 509 U.S. 137, 113 S.Ct. 2539, 125 L.Ed.2d 113 (1993), that absent a statutory or regulatory requirement, courts have no authority to require parties to exhaust administrative procedures before seeking judicial review. We distinguished Darby, however, finding that nothing in that case extinguished the general requirement that parties give the agency a chance to rule on all objections in the first instance. CSX, 568 F.3d at 247.

In its petition for rehearing, Norfolk Southern challenges our characterization of Darby. According to Norfolk Southern, Darby bars courts from imposing an exhaustion requirement where agency action has become final under the APA. See Darby, 509 U.S. at 143-53, 113 S.Ct. 2539. Although a party's failure to raise an issue during agency proceedings prior to a final appealable rule or order may result in waiver of that issue, Norfolk Southern argues that a court may not require a party to return to the agency to raise an issue that arises only at the final rulemaking. In response, the Board insists that our earlier opinion correctly interpreted Darby: Darby addresses only exhaustion of remedies, leaving in place the requirement that a petitioner present its argument to the agency at least once before seeking judicial review.

Having reconsidered this issue, we now agree with Norfolk Southern. "Wisdom," Justice Frankfurter once said, "too often never comes, and so one ought not to reject it merely because it comes late." Henslee v. Union Planters Nat'l Bank Trust Co., 335 U.S. 595, 600, 69 S.Ct. 290, 93 L.Ed. 259 (1949) (Frankfurter, J., dissenting).

Darby stands for the proposition that absent a statutory or regulatory requirement to the contrary, courts have no authority to require petitioners seeking judicial review of a final agency action to further exhaust administrative procedures. Here, although Board regulations do permit a petition for rehearing, neither the ICC Termination Act of 1995 nor the Board's regulations requires one. See 49 U.S.C. § 722; 49 C.F.R. §§ 1110.10, 1115.3(f). Under Darby, therefore, we had no authority to require Norfolk Southern to file a petition for rehearing once the agency issued its final rule.

Our earlier opinion relied on ExxonMobil Oil Corp. v. FERC, 487 F.3d 945 (D.C. Cir. 2007), in which we rejected petitioners' argument that the absence of a rehearing requirement in the Interstate Commerce Act (which was implicated because the case involved oil pipelines) meant that they had no obligation to present their arguments to FERC. But neither we nor any party noticed that the ExxonMobil petitioners had never alleged an inability to raise their arguments before issuance of the final rule. Because they could have presented their arguments to the agency in the first instance, ExxonMobil applied the well-established doctrine of issue waiver, which permits courts to decline to hear arguments not raised before the agency where the party had notice of the issue. See, e.g., United States v. L.A. Tucker Truck Lines, 344 U.S. 33, 35-37, 73 S.Ct. 67, 97 L.Ed. 54 (1952); Appalachian Power Co. v. EPA, 251 F.3d 1026, 1036 (D.C. Cir. 2001). As ExxonMobil explains, petitioners' "error was not failing to seek rehearing, but rather failing to raise the issue at all." ExxonMobil, 487 F.3d at 962.

Unlike the ExxonMobil petitioners, Norfolk Southern insists that it had no way to raise the notice argument until the Board issued its final rule. This is clearly correct. The NPRM mentions providing only one year's worth of data from which parties could draw comparison groups and nowhere indicates that the Board might consider expanding that to four years' worth of data. Given that, and given that Norfolk Southern had no obligation to file a petition for reconsideration, it had a right under Darby to seek judicial review of its argument that the Board failed to give adequate notice of the change from one-year to four-year data samples.

II.

To satisfy the APA's notice requirement, the NPRM and the final rule need not be identical: "[a]n agency's final rule need only be a `logical outgrowth' of its notice." Covad Commc'ns Co. v. FCC, 450 F.3d 528, 548 (D.C. Cir. 2006). A final rule qualifies as a logical outgrowth "if interested parties `should have anticipated' that the change was possible, and thus reasonably should have filed their comments on the subject during the notice-and-comment period." Ne. Md. Waste Disposal Auth. v. EPA, 358 F.3d 936, 952 (D.C. Cir. 2004) (citations omitted). By contrast, a final rule fails the logical outgrowth test and thus violates the APA's notice requirement where "interested parties would have had to `divine [the agency's] unspoken thoughts,' because the final rule was surprisingly distant from the proposed rule." Int'l Union, United Mine Workers of Am. v. Mine Safety Health Admin., 407 F.3d 1250, 1259-60 (D.C. Cir. 2005) (internal citations omitted).

In this case, the Board offers a two-part argument that its final rule represents a logical outgrowth of the NPRM. It first claims that the release of four-year data — albeit for another purpose — was "a foreseeable and reasonable result of other changes advocated by the railroads." Resp't Br. 33. Under the previous three-benchmark approach, two benchmarks — the RSAM and R/VC₁₈₀ — were calculated using four years' worth of private data. See Rate Guidelines — Non-Coal Proceedings, S.T.B. Ex Parte No. 347 (Sub-No. 2) at 16, 20 (served Dec. 31, 1996). By contrast, the NPRM proposed to calculate those two benchmarks based on the railroads' public filings. NPRM at 23. Although no release of private data would be necessary under the proposed rule, the NPRM indicated that if the method for calculating the two benchmarks remained the same, parties would be unable to verify the benchmarks unless the Board released the data. Id. at 23 n. 41. According to the Board, given that the railroads themselves persuaded the Board not to adopt the public filings proposal, see Decision at 80-82, railroad petitioners should have foreseen that the Board would release the four-year data to enable parties to verify the benchmarks.

Second, the Board argues that the railroads had notice that parties would draw comparison groups from whatever data the Board released. In support, the Board notes that the NPRM and the final rule contain identical language regarding the comparison groups. Compare NPRM at 20 ("The [comparison] movements would be drawn from the Waybill Sample provided to the parties by the Board."), with Decision at 18 ("The [comparison] movements must be drawn from the Waybill Sample provided to the parties by the Board at the outset of the case."), and Decision at 83 ("As explained in the NPRM, we will select the comparison group based on information contained in the Waybill Sample released to the parties at the outset of the case. . . .") (internal citations omitted). As a result, the Board argues, the railroads should have foreseen that the same data released for the other two benchmarks would be used for comparison groups.

The railroads respond that the Board's "convoluted `explanation' of its heretofore undisclosed, `complex' path" to the final rule's use of four-year data stands as "cogent proof that this change was anything but a logical outgrowth of the Board's proposal." Railroad Petr's' Reply Br. 15. The railroads point out that the NPRM proposed drawing comparison groups from the most recent year's worth of data and never mentioned the possibility that the Board might consider using data for a longer period of time. See NPRM at 33. Indeed, the railroads tell us, and the Board nowhere disagrees, that not one commenter indicated that it understood the proposal to mean that the Board might consider using more than one year's worth of private data. Railroad Pet'rs' Br. 8.

Responding to the Board's second argument, the railroads contend that the proposal to calculate two benchmarks from public data "had nothing to do with the number of years from which comparable movements would be drawn." Railroad Pet'rs' Reply Br. 15-16. Noting that the final rule expanded the data available for comparison movements "in a very oblique and indirect manner," the railroads protest that "two separate sections of the final rule, which decide wholly unrelated issues, must be cobbled together" to conclude that the rule authorizes the use of four years' worth of data for comparison movements. Railroad Pet'rs' Br. 12. According to the railroads, because the NPRM nowhere suggested that the two sections were linked, it failed to give adequate notice that the Board was considering using anything other than the most recent year's data to derive comparison groups.

As mentioned above, a final rule qualifies as a logical outgrowth of the proposed rule if interested parties "`should have anticipated' that the change was possible." Ne. Md. Waste Disposal Auth., 358 F.3d at 952. We have found that a final rule represents a logical outgrowth where the NPRM expressly asked for comments on a particular issue or otherwise made clear that the agency was contemplating a particular change. For example, in Owner-Operator Independent Drivers Ass'n v. Federal Motor Carrier Safety Administration, 494 F.3d 188 (D.C. Cir. 2007), we considered a rule that allowed long-haul truck drivers to satisfy their ten-hour off-duty requirement in two separate resting periods as long as one period was at least eight hours long. We concluded that the final rule was a logical outgrowth of the NPRM, which stated that "FMCSA will consider a variety of possible changes . . . including . . . establishing a minimum time for one of the two `splits,' such as 5 hours, 8 hours, or some other appropriate level." Id. at 209-10. Similarly, in City of Portland v. Environmental Protection Agency, 507 F.3d 706 (D.C. Cir. 2007), we concluded that a final rule requiring uncovered reservoirs to be treated for a particular parasite was a logical outgrowth of the proposed rule because the NPRM expressly requested comments on whether the agency should consider requiring cities to inactivate the parasite. Id. at 715.

By contrast, our cases finding that a rule was not a logical outgrowth have often involved situations where the proposed rule gave no indication that the agency was considering a different approach, and the final rule revealed that the agency had completely changed its position. For example, in International Union, we concluded that a final rule setting a maximum mine belt air velocity of 500 feet per minute was not a logical outgrowth of a proposed rule providing that "[a] minimum air velocity of 300 feet per minute must be maintained." 407 F.3d at 1259. We explained that "the Secretary could not have expected interested parties to realize that she would consider abandoning her proposed regulatory approach . . . simply because she invited commentary on a proposed rule that included a minimum air velocity." Id. at 1260. We reached a similar result in Environmental Integrity Project v. Environmental Protection Agency, 425 F.3d 992 (D.C. Cir. 2005), in which EPA published a proposed rule clarifying that a set of regulations operate independently of one another. In its final rule, however, EPA adopted just the opposite position, declaring that those regulations are in fact not separate regulatory standards. Id. at 994-95. We rejected EPA's argument that it had satisfied its notice-and-comment obligations by "repudiat[ing] its proposed interpretation and adopt[ing] its inverse" in the final rule. Id. at 998.

This case presents a closer question. Although the NPRM neither asked for comments on a particular issue nor otherwise indicated that the Board was contemplating a particular change, the final rule did not amount to a complete turnaround from the NPRM. That said, we think this case far more like those in which we found that agencies had failed to give adequate notice. In essence, the Board contends that the mere mention of the release of one-year data for comparison groups gave notice that the amount of data available for that purpose might change. We rejected just that argument in both International Union and Environmental Integrity Project, and we do so here as well. Although the NPRM proposed several revisions to the existing system, it nowhere even hinted that the Board might consider expanding the number of years from which comparison groups could be derived. Unlike the notices in Owner-Operator and City of Portland, in which we found that the final rules qualified as logical outgrowths, the Board's NPRM requested comments on no particular issue at all. To be sure, expanding from one to four years' worth of data is less dramatic than adopting a maximum velocity cap where a minimum was proposed ( International Union) or a completely different reading of a set of regulatory standards ( Environmental Integrity Project). Even so, we see no way that commenters here could have anticipated which "particular aspects of [the Board's] proposal [were] open for consideration." Envtl. Integrity Project, 425 F.3d at 998 (emphasis omitted); see also Fertilizer Inst. v. EPA, 935 F.2d 1303, 1312 (D.C. Cir. 1991). Indeed, were we to conclude that commenters had notice merely because the NPRM mentioned one year's worth of data, the Board could issue broad NPRMs "only to justify any final rule it might be able to devise by whimsically picking and choosing within the four corners of a lengthy `notice.'" Envtl. Integrity Project, 425 F.3d at 998. Such a rule would hardly promote the purposes of the APA's notice requirement.

We are similarly unpersuaded by the Board's argument that making private data available to verify the other two benchmarks gave commenters notice that the same data would be used to derive comparison groups. Although both the NPRM and the final rule note that comparison groups will be drawn from data released to the parties, neither makes clear that the Board was referring to all data released to the parties for any purpose. Indeed, the language regarding the release of data appears in portions of the NPRM discussing comparison groups, suggesting that it refers only to data released for that purpose. Moreover, even under the Board's broader reading, commenters could hardly be expected to pick this single sentence out of a sixty-four page NPRM absent any indication that the comparison group data might change.

To be sure, in retrospect we might be able to discern the Board's reasoning, i.e., that the railroads' other unrelated comments suggested keeping the four-year averages for two benchmarks, that as a result the Board might release four-year data, and that the four-year data might then be used to calculate comparison groups. Under the APA, however, notice must come from the NPRM. See 5 U.S.C. § 553(b). Here, because nothing in the NPRM (1) indicated that the Board might consider expanding the comparison group data from one to four years, or (2) linked data released for other purposes to the comparison groups, we are unable to conclude that the final rule qualifies as a logical outgrowth of the NPRM.

The Board next argues that even if we conclude that it failed to give adequate notice, there is no reason to vacate the rule because the change was neither prejudicial to the railroads nor important. See 5 U.S.C. § 706 (requiring courts to take "due account" of "the rule of prejudicial error"); First Am. Discount Corp. v. Commodity Futures Trading Comm'n, 222 F.3d 1008, 1015 (D.C. Cir. 2000) (declining to decide whether final rule represented a logical outgrowth where petitioner suffered no prejudice); Transmission Access Policy Study Group v. FERC, 225 F.3d 667, 729 (D.C. Cir. 2000) (declining to apply logical outgrowth analysis to minor change), aff'd sub nom. New York v. FERC, 535 U.S. 1, 122 S.Ct. 1012, 152 L.Ed.2d 47 (2002). In support, the Board points out that the final rule allows parties to demonstrate in individual cases that comparison movements drawn from older data are "unreasonable." Although this is certainly true, the railroads' point, with which we agree, is that they were prejudiced by their inability to persuade the Board not to adopt the four-year rule in the first place, thus requiring them to litigate the issue in individual proceedings. "Had the Board given notice that it proposed to include four years' historical Waybill Samples," the railroads tell us, they "would have made additional objections and presented significantly more (and different) evidence (concerning, for example, changes in market conditions over four-to-six year periods) to support those objections." Railroad Pet'rs' Br. 10. The railroads also point out that because the Board needs one to two years to gather and release the data, see Decision at 84, expansion to four years' worth of data means that comparison groups could be drawn from movements that are up to six years old, and older data increases the "likelihood of distorted comparisons and results." Railroad Pet'rs' Br. 12. We thus agree with the railroads that the change from one year to four years' worth of data was important and potentially prejudicial.

III.

For the foregoing reasons, we conclude that the Board failed to comply with the APA's notice and comment requirements. We therefore vacate (1) the portion of our earlier opinion rejecting the railroads' notice argument, see CSX, 568 F.3d at 246-47, and (2) the portion of the final rule that makes four years of data available for comparison groups, see STB Ex Parte No. 646 (Sub-No. 1), at 83 (served Sept. 5, 2007); see also Allied-Signal, Inc. v. U.S. Nuclear Regulatory Comm'n, 988 F.2d 146, 150-51 (D.C. Cir. 1993). We also vacate the portion of our opinion rejecting the railroads' argument regarding regulatory lag, see CSX, 568 F.3d at 247-48, an issue we had no need to reach given our conclusion here that the Board failed to provide the required notice.

So ordered.