dissenting.
My disagreement with the majority is a narrow albeit decisive one. When the Transportation Security Administration calculated passenger- and property-screening costs for the year 2000, it failed to consider another Government agency’s estimate of the total number of persons screened that year. Notwithstanding the Airlines’ protestations, neither TSA nor the consultant whose analysis it relied on even mentioned that critical data. Although TSA’s calculation of the security fee is entitled to broad deference, the agency’s discretion is not unlimited. I respectfully dissent, because I think TSA impermissibly ignored contradictory evidence.
The Department of Transportation’s Bureau of Transportation Statistics estimated that in the year 2000, the Airlines screened 1.812 billion persons — more than double TSA’s estimate of 865.5 million. After we remanded TSA’s first decision, see Southwest Airlines Co. v. TSA (Southwest I), 554 F.3d 1065 (D.C.Cir.2009), the Airlines provided DOT’S estimate to TSA in a consultant’s report (the “Campbell Report”). Accepting DOT’S estimate would have resulted in a substantially lower security fee for the Airlines.1 Not surprisingly, TSA did nothing of the sort. Although TSA had promised to consider the Campbell Report, it adopted wholesale its own consultant’s report (the “SH & E Report”), which arrived at its estimate of 865.5 million persons screened without even mentioning the Campbell Report or the DOT data it cited.
TSA’s decision on remand gives no reason for choosing SH & E’s estimate of the total number of passengers screened over DOT’s estimate. The decision’s treatment of the Campbell report is confined to a two-sentence paragraph. It is in this cursory statement that the court purports to divine TSA’s reasoned consideration of the DOT data:
TSA conducted a thorough review of the Campbell report that included an examination of both the data and methodologies utilized to construct the report find*759ings. TSA concluded that the Campbell report and findings were insufficient for further consideration due to the report’s use of limited data and broad, simplistic methodologies that did not consider the full spectrum of specific cost categories.
Joint Appendix (“J.A.”) 468, quoted in Maj. Op. at 755-56, 756. Aside from this vague gesture toward the Campbell Report and its data, the remand decision does not address the discrepancy between DOT’s estimate of 1.812 billion screened persons and SH & E’s estimate of 865.5 million. The court interprets TSA’s reference to “broad, simplistic methodologies” and “limited data” as an assessment of Campbell’s uncritical adoption of DOT’s estimate. Maj. Op. at 756. TSA’s statement, however, cannot sustain that charitable reading.
TSA explicitly faulted the Campbell Report for its failure to “consider the full spectrum of specific cost categories.” J.A. 468. This has nothing to do with DOT’s estimate of the total number of passengers screened in 2000. To be sure, the SH & E Report attempts a more sophisticated comparative analysis of the respective costs of screening passengers and non-passengers than the Campbell Report. In particular, SH & E accounted for fixed costs that would have remained in the absence of non-passenger screenings, whereas Campbell assumed non-passengers contributed an equal share to the cost of screening persons. Compare J.A. 279-323 (SH & E Report pt. 3), with J.A. 174-79 (Campbell Report). But this issue is beside the point. Accepting SH & E’s estimates of the respective costs of individual passenger- and non-passenger screenings, a higher raw number of screenings would still result in a lower security fee for the Airlines than the one TSA calculated. There is no evidence TSA considered this critical difference between the reports at all.2 Without considering the issue, TSA could not have reasonably decided to credit SH & E’s estimate of total screenings over DOT’s.
Even if TSA’s denigration of the Campbell Report’s “limited data” is interpreted as a reference to the DOT data, such conclusory treatment of alternative evidence “provides no basis upon which we could conclude that it was the product of reasoned decisionmaking.” Butte County, Cal. v. Hogen, 613 F.3d 190, 195 (D.C.Cir. 2010) (quoting Tourus Records, Inc. v. DEA, 259 F.3d 731, 737 (D.C.Cir.2001)). To merit deference, TSA ought to have given some reason for rejecting DOT’s estimate. See Int’l Union, United Mine Workers of Am. v. Mine Safety & Health Admin., 626 F.3d 84, 93 (D.C.Cir.2010) (remanding a rule “because ... it defies the expert record evidence and is unexplained”). Labeling the DOT data “limited” — if indeed that pejorative may be read as a criticism of DOT’s estimate — is not a reasoned explanation. In United Mine Workers, we concluded the challenged rule was arbitrary and capricious because the agency’s only basis for rejecting contrary evidence was the agency’s own “knowledge *760and expertise.” 626 F.3d at 84. TSA’s rejection of the Campbell Report, and its silent neglect of the DOT data contained therein, is no more descriptive. Such “[cjonclusory explanations for matters involving a central factual dispute where there is considerable evidence in conflict do not suffice to meet the deferential standards of our review.” Id. at 94 (quoting AT&T Wireless Servs. v. FCC, 270 F.3d 959, 968 (D.C.Cir.2001)).
TSA does not even argue its decision on remand meets the standard we applied in United Mine Workers. Instead, TSA asks us to limit the holding of that case to situations where the neglected “contrary evidence” is “set forth in a congressionallyordered study conducted by an independent federal agency with expertise in the subject matter.” Respondent’s Br. at 43. TSA does not explain why it is essential that the study be “congressionally-ordered” or why DOT’s Bureau of Transportation Statistics lacks “expertise in the subject matter” of air transportation security statistics. Regardless of the specific character of the contrary evidence, an agency is required by “[bjasic principles of administrative law” to “examine the relevant data and articulate a satisfactory explanation for its action including a rational connection between the facts found and the choice made.” AT&T Wireless, 270 F.3d at 968 (quoting Motor Vehicle Mfrs. Ass’n v. State Farm Mwt. Auto. Ins. Co., 463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983)).
Our deference to the substance of an agency’s decision does not permit us to ignore the process by which the agency makes it. The court rightly observes that an agency is entitled to especially strong deference where the relevant statute turns on “the [agency’s] determination that certain conditions are present” — here the cost of screening passengers and property in 2000, as determined by the TSA — rather than “the objective existence of [those] conditions.” Maj. Op. at 756 (quoting Southwest I, 554 F.3d at 1071). But even when our review is at its most deferential, we may not allow an agency to shirk its duty to provide a reason for choosing one body of evidence over another. In AEP Texas North Co. v. STB, 609 F.3d 432 (D.C.Cir.2010), the court noted that the Surface Transportation Board was “entitled to particular deference” because the rate-setting decision at issue was one in which the Board acts “at the zenith of its powers.” Id. at 438. Nevertheless, the court held that the Board acted arbitrarily and capriciously when it “entirely failed to consider an important aspect of the problem.” Id. at 441. For the same reason, TSA should not be able to hide its neglect of DOT’s estimate behind our standard of review.
The court’s “most important! ]” response to the Airlines’ argument that TSA failed to consider the DOT data is that “the airlines presented no evidence that the figure in the [DOT] report was at all reliable.” Maj. Op. at 757. The court conducts its own analysis of the DOT estimate, finds it to be based on self-serving, industry-reported data, and then concludes the SH & E Report’s estimate of total screenings was more reliable. Id. at 756-57. This is exactly the sort of analysis TSA should have undertaken, and it is exactly the sort of reasoning to which a court may defer.3 But TSA’s remand decision makes none of these findings, which appear for the first time as arguments in the agency’s appel*761late brief. See Respondent’s Br. at 38-42. Such post hoe justifications cannot satisfy the agency’s obligation to give reasons for rejecting alternative evidence. See SEC v. Chenery Corp., 332 U.S. 194, 196-97, 67 S.Ct. 1575, 91 L.Ed. 1995 (1947); United Mine Workers, 626 F.3d at 94.
Especially where an agency adopts the reasoning of an outside consultant in toto, as TSA did here, the agency must articulate its reasons for rejecting evidence the consultant ignored. TSA’s failure to do so leaves me with the sneaking suspicion that neither the agency nor its consultant ever seriously considered DOT’S estimate. I would remand once again for further consideration and an explanation. See United Mine Workers, 626 F.3d at 94; AT&T Wireless, 270 F.3d at 968.
. The security fee is capped at the year 2000 cost of screening persons and property ($448 million) minus the cost of screening non-passengers. See Maj. Op. at 754; Southwest I, 554 F.3d at 1069 (D.C.Cir.2009). The parties agree there were 527 million passengers in 2000.
. A letter written by TSA's Acting Assistant Chief Counsel after the agency's final remand decision confirms that the agency misunderstood the fundamental difference between the Campbell and SH & E Reports. Like the remand decision, the letter failed to mention — much less refute — DOT’s estimate of 1.812 billion screened persons. In response to the airlines' "concern regarding SH & E’s estimate of the ratio of passenger to non-passenger screenings,” the letter points out that the Campbell report "extrapolated the experience of just six aviation industry representatives.” J.A. 504. But Campbell relied on those six industry representatives to construct an estimate of the relative average cost of individual passenger and non-passenger screenings. Those surveys were irrelevant to the DOT data on the relative volume of passenger and non-passenger screenings.
. This is not to say the court's reasoning is self-evidently correct. The SH & E Report's estimate of the ratio of passenger to non-passenger screenings was based on "passenger surveys" conducted at two airports and on interviews with airport personnel. It is not obvious that the resulting estimate is more reliable than DOT’s. The Campbell Report gives an intuitive explanation for DOT’s large estimate of the number of non-passen*761gers screened in 2000, relative to the present. Before 9/11, it was common for non-passengers to drop off and pick up passengers; screenings were less onerous; and airline, airport, vendor, and contractor employees passed easily and often through screenings. Most persuasively, the Campbell Report shows that the large reduction in screened persons between 2000 and 2006 (a drop from 1.812 billion to 708 million) corresponds to the increased transaction costs associated with TSA’s management of the screening process and new rules restricting non-passengers from the "sterile” area of the airport. It is impossible to perform a similar comparison with the SH & E data TSA relied on, because SH & E did not estimate screening volume for any year besides 2000. Of course, TSA may have had a reasonable basis for favoring SH & E’s much smaller estimate of year 2000 screenings, but the agency was obliged to explain its reasons.