United Airlines, Inc. v. TSAUnited Airlines, Inc. v. TSA
On Petition for Review of a Decision of the Transportation Security Administration
Adam P. Feinberg argued the cause and filed the briefs for petitioner.
Leif E. Overvold, Attorney, U.S. Department of Justice, argued the cause for respondent. With him on the brief were Brian M. Boynton, Acting Assistant Attorney General, and Scott R. McIntosh, Attorney.
Before: HENDERSON and WALKER, Circuit Judges, and EDWARDS, Senior Circuit Judge.
Opinion for the Court filed by Circuit Judge HENDERSON.
I. Statutory Framework
The Aviation and Transportation Security Act, Pub. L. No. 107-71, 155 Stat. 597 (2001) (codified at
II. Procedural History
On April 8, 2016, United submitted a refund request to the TSA through its consultant, Ryan Excise Tax Services, LLC (Ryan). United sought the return of security fees that it asserted had been erroneously remitted during the period from January 1, 2010 through February 29, 2012. The asserted overpayments can be separated into two categories. First, United claimed that it had erroneously remitted to the TSA $1,059,743.06 in security fees in connection with passengers who bought their tickets from other airlines but were later involuntarily transferred to United flights. For these Involuntary Transfer (IT) tickets, United maintained that it remitted the security fees despite having never collected the fees from the passengers and that the transferring airline, not United, maintained responsibility for their collection and remittance to the TSA.
Second, United claimed that it had erroneously remitted $478,244.88 in connection with tickets for which United had collected the security fee in a foreign currency but subsequent fluctuations in the foreign exchange rate caused the collected fee to be slightly more or slightly less than the amount required by statute—$2.50, or a multiple thereof—when it was ultimately recorded by United.1 If the converted amount was less than the statutorily required amount, United adjusted upward and remitted the amount required by statute. But if the converted amount was more than the statutory amount, United did not adjust downward, instead remitting the higher amount to the TSA. For these Exchange-Rate-Difference (ERD) tickets, United claimed its practice resulted in a net overpayment to the TSA.
On April 18, 2016, the TSA promptly denied United‘s refund request, concluding that the request was precluded by United‘s failure to express its concerns during an audit conducted by the TSA in 2012. See United Airlines, Inc. v. TSA, 859 F.3d 67, 69–70 (D.C. Cir. 2017). This Court disagreed and remanded for further administrative proceedings. Id. at 70–71.
On remand, United renewed and supplemented its refund request in a letter to the TSA that outlined Ryan‘s methodology. Ryan first identified the two sets of tickets at issue here: one that included all tickets for which another airline‘s ticket stock had been involuntarily used as payment for a United ticket—the IT tickets—and one that included all tickets for which a security fee was deposited into United‘s fee account that was not evenly divisible by the then-applicable statutory fee amount of $2.50—the ERD tickets. Ryan then undertook a “programmatic review” by running a computer formula programmed to determine whether the correct security fee had been remitted for each ticket. For the IT tickets, Ryan treated any payment of a security fee as an overpayment. For the ERD tickets, Ryan treated amounts paid in excess of $2.50 (or a multiple where
Ryan then verified the programmatic review‘s results using a “stratified random sample,” whereby Ryan manually reviewed a sample of 2,135 tickets, calculated the net refund amount for that sample and then extrapolated that amount for the entire ticket sample. Using the stratified random sample, Ryan calculated a similar refund amount as that calculated by the programmatic review. Ryan conducted a similar verification process using the 600-ticket sample used by the TSA during its 2012 audit, again extrapolating a similar refund amount.
During its review process, the TSA worked with the U.S. Customs and Border Protection‘s Office of Trade, Regulatory, Audit and Agency Advisory Services (CBP) to examine and verify the reliability of Ryan‘s methodology and calculations. The CBP, in turn, communicated with Ryan to clarify anomalies in its analysis and methodology. For example, Ryan disclosed the exclusion of the eight ranges of ERD tickets only after the CBP inquired into Ryan‘s search parameters; the actual ticket data for the excluded tickets was not provided to the CBP or the TSA. After concluding its review, the CBP submitted a memorandum summarizing its findings to the TSA. The CBP explained that its team was unable to verify the reliability of Ryan‘s data and analysis, citing its inability to replicate Ryan‘s calculation to arrive at the same net refund amount, its determination that Ryan‘s programmatic review—and, by extension, its stratified random sample—relied on an incomplete universe of tickets and its observation of numerous discrepancies in the accounting records provided by United.
On April 21, 2020, the TSA again denied United‘s refund request. With respect to the IT tickets, the TSA concluded that United‘s bare assertion that it had no statutory obligation to remit a security fee did not address whether the transferring airline had already remitted the associated fee or whether United received anything less than all funds the passenger originally paid to the transferring airline, including the fee. In the TSA‘s view, United‘s submission created only the possibility that United might be entitled to a refund for this category of ticket but otherwise fell short of demonstrating that a refund was warranted for all IT tickets as a categorical matter. For the ERD tickets, the TSA determined that United‘s submission did not substantiate United‘s underlying conclusion that it had made a net overpayment of security fees. The TSA cited the CBP‘s concerns with Ryan‘s methodology and calculations, including the exclusion of specific ranges of tickets from the programmatic review, the failure of the stratified random sample to verify the result from an otherwise incomplete universe, the
III. Analysis
United contends that the TSA‘s rejection of its refund request was arbitrary and capricious. United first maintains that it was under no legal duty to remit the security fee associated with the IT tickets, meaning that every remittance was categorically an overpayment. In United‘s view, the TSA‘s assumption that the transferring carrier might have either transferred the security fee to United or not remitted the fee to the TSA is therefore unsupported and irrelevant. With respect to the ERD tickets, United argues that the perceived computational and analytical errors in Ryan‘s methodology cannot provide a basis for the TSA to deny or materially reduce its refund request. United finally argues that the TSA acted arbitrarily and capriciously in denying any refund rather than calculating an alternative refund amount using the data before it.
Our review “is limited to determining whether the TSA acted arbitrarily or capriciously, abused its discretion, or acted contrary to law.” Alaska Airlines, Inc. v. TSA, 588 F.3d 1116, 1120 (D.C. Cir. 2009); see also
A. Involuntary Transfer Tickets
In the involuntary transfer context, the allocation of legal liability for unremitted security fees is clear. The statute provides that any security fee “shall be collected by the air carrier that sells a ticket for transportation.”
But this is only the starting point. This is not a case about United‘s failure to remit security fees that it was required to collect. Rather, United is seeking a refund of security fees it erroneously—and inexplicably—remitted to the TSA despite having no statutory responsibility to do so. The TSA‘s authority to issue such refunds is discretionary: “The [TSA] may refund any fee paid by mistake or any amount paid in excess of that required.”
Placing the burden on United, however, does not relieve the TSA of its ordinary burden under the Administrative Procedure Act—i.e., its duty to provide a reasoned explanation for its decision. See, e.g., State Farm, 463 U.S. at 42–52. Even when denying an interested party‘s request via informal adjudication, an agency cannot merely state a “conclusion” but rather “must ‘articulate a satisfactory explanation’ for its action.” Butte Cnty., Cal. v. Hogen, 613 F.3d 190, 194 (D.C. Cir. 2010) (quoting Tourus Records, Inc. v. DEA, 259 F.3d 731, 737 (D.C. Cir. 2001)); see also
In denying United‘s request, the TSA concluded that United had failed to demonstrate that any overpayment occurred. More specifically, it determined that United failed to show that the transferring airline had already remitted the relevant security fee to the TSA or that United had not received from the transferring airline anything less than all of the funds the involuntarily transferred passenger had originally paid to the transferring airline, including the fee amount. But the problem with the TSA‘s rationale is this: If the transferring airline remains legally obligated to collect and remit the security fees for the tickets it sells, even if the passenger is involuntarily transferred to United, there is little reason to suppose that the transferring airline would pass along the security fee to United instead of remitting it to the TSA in proper course. Why would the transferring airline entrust United to satisfy the transferring airline‘s legal responsibility, thereby risking noncompliance if United failed to do so? The TSA‘s hypothetical about airlines transferring security fees among themselves therefore appears logically incongruent with the allocation of liability under the statute and the TSA otherwise makes no effort to rely on industry practice or past practice to validate its concern. See United Airlines, 859 F.3d at 71 n.11 (“For those overpayments due to involuntary transfers, there is no reason to suspect that the carrier that sold the original ticket did not also pay TSA, i.e., it is equally likely that TSA was paid double.“). The TSA‘s reasoning therefore strikes more as a largely unsupported hypothetical than a “satisfactory explanation” rooted in logic or practice. See Butte Cnty., 613 F.3d at 194 (quoting Tourus Records, 259 F.3d at 737).
That said, the TSA‘s concern is nevertheless understandable. If United simply passed along a security fee received from the transferring airline, there was no overpayment, meaning that a refund would leave the TSA shortchanged for that passenger.4 For its part, United appears to have made no effort to verify that it did not receive a passenger‘s security fee as part of the funds it received from the transferring airline or to cite to industry practice highlighting why such verification would be unnecessary. Instead, United chose to repeat its assertion that it had no
We are therefore confronted with a factual dispute with important implications for United‘s refund. On the one hand, United claims that it never transfers security fees—a practice that appears correct in view of the allocation of liability under
B. Exchange-Rate-Difference Tickets
We turn next to the ERD tickets. The TSA, relying on the CBP‘s analysis of United‘s submission, determined that “material limitations in Ryan‘s overarching methodology” rendered the submission insufficiently reliable to warrant a refund. J.A. 8. In particular, the TSA focused on deficiencies in the programmatic review and the stratified random sample as well as discrepancies inherent in United‘s accounting data for the security fees. We find the TSA‘s conclusions regarding these deficiencies, viewed collectively, provide a reasonable basis for the TSA‘s denial and find United‘s attempts to minimize these flaws unavailing.
Programmatic review: With respect to the programmatic review, the TSA‘s denial relied chiefly on two key limitations. First, the TSA noted that United elected to exclude from its submission—without informing the TSA or the CBP at the outset—those tickets it deemed insufficiently close to statutory fee amounts to be accurately classified as overpayments or underpayments by the programmatic review. According to the TSA, “Ryan was necessarily excluding from its search query tickets for which [United] may have made an over- or under-payment, meaning the universe of tickets that Ryan identified for its refund calculations was necessarily incomplete.” J.A. 9. For example, United excluded tickets for which United remitted a net security fee amount of $0.01 to $1.48, tickets the TSA noted “would appear to be comprised solely of tickets for which [United] under-remitted the Fee (as every ticket in that group falls below the minimum Fee amount of $2.50).” J.A. 9.5 The TSA therefore reasonably concluded that because
Second, the TSA noted that not all of the fee amounts within this universe of tickets could be attributed to exchange rate fluctuations, which Ryan‘s programmatic review necessarily assumed. A more detailed review by the CBP revealed accounting discrepancies suggesting that a ticket may have a fee that deviates from that required by statute for reasons entirely unrelated to exchange rate fluctuations. As one example, in response to a CBP inquiry into two tickets, Ryan explained that United had initially (and correctly) collected $5.00 for each but unexplained refunds had been issued in the amounts of $3.06 and $0.83, bringing the remittance amounts to $1.94 and $4.17 and causing the programmatic review to erroneously flag the tickets as ERD tickets. Although Ryan chalked up the identified discrepancies to “operator errors” and “field refunds” made at airports, it did not otherwise elaborate on their frequency or aggregate impact on the calculated net refund amount. In short, the TSA reasonably concluded that Ryan‘s methodology “necessarily depends on the assumption that the Fee attributable to a ticket was always a multiple of $2.50 or within $1.00 thereof,” but “[t]he vagaries of [United]‘s accounting practices . . . reveals that tickets may easily have unusual Fee amounts assigned to them in [United]‘s ledger,” which would not necessarily warrant a refund. J.A. 9–10 n.15.
We find United‘s arguments to the contrary unpersuasive. United maintains that the number of excluded tickets was “insignificant” but it never substantiated this claim to the TSA or CBP by providing, for example, the relevant ticket-level data, the number of excluded tickets or the effect of those tickets on the net refund calculation.6 United‘s failure to provide this information is particularly glaring given that the CBP asked specifically about the excluded tickets and Ryan‘s search parameters. United had every incentive to substantiate its assertion that the excluded tickets were in fact “insignificant” in the net refund calculation. United ultimately faults the TSA for speculating about the existence of underpayments but the fact remains that United did not provide the TSA with the information necessary to do anything but reasonably hypothesize about a known but undefined pool of potential underpayments within the ERD ticket universe. We cannot fault the TSA for declining to take United‘s word that the excluded tickets were “insignificant,” much less find that its decision was arbitrary and capricious.
United further argues that the TSA, not United, had the burden to establish the existence of underpayments that would offset the total net refund amount, characterizing this burden as an affirmative defense.
Stratified random sample: The TSA correctly concluded that Ryan‘s use of a stratified random sample to verify its programmatic review could not make up for the excluded tickets because a sample “drawn from an incomplete universe will, of necessity, tend only to confirm the results drawn from the incomplete universe itself.” United does not contest the TSA‘s conclusion and we see no reason to disturb it.
The TSA further noted that the stratified random sample contained 32 tickets—within a total sample of 2,135 tickets—that were absent from the programmatic review. The TSA concluded that “the inclusion of exemplars in the stratified random sample that do not appear in the universe is a fundamental flaw in the reliability of the random sampling effort itself.” J.A. 11. United points out that Ryan acknowledged this discrepancy during the CBP‘s review and explained that it had removed the 32 tickets from the programmatic review because the tickets were determined to be neither overpayments nor underpayments, meaning that they would have no effect on the net refund calculated by the programmatic review. Ryan further explained that it did not remove the 32 tickets from the stratified random sample because it did not believe they would change the outcome of the new programmatic results.
But this explanation misses the point. If the 32 tickets were in fact neither underpayments nor overpayments, United is correct that their exclusion from the programmatic review would have no effect on the calculated net refund. The TSA, however, was focused on the stratified random sample, not the programmatic review, and whether it was reflective of the universe of tickets from which it was ostensibly drawn—і.е., the universe of tickets comprising the programmatic review. The reliability of random sampling, namely its randomness, decreases if the random sample is not drawn from the universe it purportedly samples. See RICHARD L. SCHEAFFER ET AL., ELEMENTARY SURVEY SAMPLING 8–9 (7th ed. 2012). Given that neither Ryan nor United attempted to answer the CBP‘s concerns over the inclusion of the 32 tickets in the stratified random sample, it was reasonable for the TSA to conclude that the validity of the stratified random sample was diminished as a result.
The TSA also cited the CBP‘s inability to replicate Ryan‘s calculations to reach the requested refund amount, prompting the CBP to conclude that Ryan utilized a dataset other than the one provided to the TSA. The TSA characterized the differing calculations as “inexplicabl[e].” J.A. 11. But, as United points out, the record does provide an explanation. When CBP voiced concern over the discrepancy, Ryan explained that it stemmed from duplicated data and confirmed that the CBP‘s calculation was correct. Thus, we find that it was unreasonable for the TSA to rely on this later-reconciled error, at least absent any reason or findings to the contrary. Nevertheless, even if this single basis for denying United‘s refund request may not be reasonable, we cannot demand perfection nor vacate the TSA‘s decision on this basis alone. See Dickson, 68 F.3d at 1404 (noting that an agency‘s decision need not be “a model of analytic precision to survive a challenge“).
In short, because the TSA‘s decision to deny United‘s request for a refund for the ERD tickets was reasonable, we deny United‘s petition regarding the ERD tickets.
C. The TSA‘s Duty to Calculate an Alternative Refund Amount
Finally, United contends that the TSA‘s decision to deny its entire refund request rather than calculate a revised refund amount in light of the omissions and errors detected by the TSA and CBP was arbitrary and capricious. Rather than rely on administrative law principles to support its assertion, United relies primarily on Cohan v. Commissioner, 39 F.2d 540 (2d Cir. 1930), a federal tax case. In Cohan, the Second Circuit concluded that the U.S. Board of Tax Appeals—the predecessor of the modern U.S. Tax Court—could not deny a business-expense deduction altogether when it was clear the taxpayer “had spent much and that the sums were allowable expenses.” Id. at 543. Although the court noted that “[a]bsolute certainty in such matters is usually impossible and is not necessary,” it reasoned that the Board “should make as close an approximation as it can, bearing heavily if it chooses upon the taxpayer whose inexactitude is of his own making.” Id. at 543-44; see also United States v. Marabelles, 724 F.2d 1374, 1383 (9th Cir. 1984) (citing Cohan and stating that “if it is clear that the taxpayer is entitled to some deduction, but he cannot establish the full amount claimed, it is
At the outset, United makes no effort to anchor the Cohan principle to
To the extent that Cohan may operate as a background principle of fairness, we have found the Cohan principle inapplicable if “there are no reliable figures from which to calculate or extrapolate a reasonable estimate’ of taxpayers’ entitlements.” Green Gas Del. Statutory Tr. v. Comm‘r, 903 F.3d 138, 144 (D.C. Cir. 2018) (quoting Plisco v. United States, 306 F.2d 784, 787 (D.C. Cir. 1962)); see also Coloman v. Comm‘r, 540 F.2d 427, 431–32 (9th Cir. 1976) (cautioning that undue application of the Cohan principle “would . . . in essence condone the use of that doctrine as a substitute for burden of proof“). In Green Gas, this Court declined to require a ”Cohan estimate” of deductible landfill gas production because the Tax Court reasonably concluded that the estimation methods proffered by landfill owners were insufficiently reliable. Id. at 144. For example, the Tax Court determined that the owners’ site logs were too infrequent and data contained in them was “statistically improbable,” and that software used by the owners to monitor landfill gas emissions was not designed to provide an accurate measurement of landfill gas production. Id. at 143. Green Gas supports the conclusion that a Cohan estimate is unwarranted when the reviewing body—whether the Tax Court or an agency—has valid concerns about the reliability of the proffered methodology or data. We therefore decline to find that the TSA acted arbitrarily or capriciously when it elected to deny United‘s submission rather than approximate a refund amount based on data and methodology it questioned.7
IV. Conclusion
For the foregoing reasons, we grant United‘s petition in part, vacate the TSA‘s decision with respect to the Involuntary Transfer tickets and otherwise deny the
So ordered.