Duquesne Light Holdings Inc v. Commissioner of Internal Reven

PRECEDENTIAL

UNITED STATES COURT OF APPEALS
FOR THE THIRD CIRCUIT
________________

No. 14-1743

DUQUESNE LIGHT HOLDINGS, INC. & SUBSIDIARIES
f/k/a
DQE, INC & SUBSIDIARIES,

Duquesne Light Holdings, Inc. & Subsidiaries,

Appellant

v.

COMMISSIONER OF INTERNAL REVENUE
________________

Appeal from the Decision of the
United States Tax Court
Docket No. 10-9624
Tax Court Judge: Honorable Carolyn P. Chiechi

Argued July 12, 2016

Before: AMBRO, HARDIMAN,
and KRAUSE, Circuit Judges
 (Opinion filed: June 29, 2017)

James C. Martin, Esquire (Argued)
Reed Smith LLP
225 Fifth Avenue, Suite 1200
Pittsburgh, PA 15222

Counsel for Appellant

Caroline D. Ciraolo
Acting Assistant Attorney General
Diana L. Erbsen
Deputy Assistant Attorney General
Arthur T. Catterall, Esquire (Argued)
Jonathan S. Cohen, Esquire
Gilbert S. Rothenberg, Esquire
Jennifer M. Rubin, Esquire
United States Department of Justice
Tax Division, Room 4333
950 Pennsylvania Avenue, N.W.
P.O. Box 502
Washington, DC 20044

Joseph P. Grant, Esquire
Mary H. Weber, Esquire
Internal Revenue Service
Office of Chief Counsel
312 Elm Street, Suite 2350
Cincinnati, OH 45202

Counsel for Appellee

2
 ________________

OPINION OF THE COURT
________________

AMBRO, Circuit Judge

This appeal concerns the continued vitality of the so-
called Ilfeld doctrine for interpreting the Internal Revenue
Code. Taken from Charles Ilfeld Co. v. Hernandez, 292 U.S.
62 (1934), this doctrine teaches that “the Code should not be
interpreted to allow [the taxpayer] ‘the practical equivalent of
a double deduction’ … absent a clear declaration of intent by
Congress.” United States v. Skelly Oil Co., 394 U.S. 678, 684
(1969) (quoting Ilfeld, 292 U.S. at 68). Duquesne Light
Holdings, Inc. (“DLH”) and its subsidiaries (DLH and its
subsidiaries are variously referred to as the “Duquesne
group,” the “Duquesne entities,” or simply “Duquesne”)
appeal the Tax Court’s grant of summary judgment to the
Internal Revenue Service based on the Ilfeld doctrine. In
particular, Duquesne challenges the Tax Court’s holding that
the consolidated entities affiliated with DLH claimed a
double deduction for certain losses incurred by its
AquaSource subsidiary and thus disallowed $199 million of
those losses (all numbers are rounded). As we conclude that
the Tax Court properly applied the Ilfeld doctrine, we affirm. 1

I. BACKGROUND

The Duquesne entities, including DLH and
AquaSource, filed their tax returns as a consolidated taxpayer,

1
The Tax Court had jurisdiction to hear this case
under 26 U.S.C. § 6213 and we have jurisdiction to review its
judgment per 26 U.S.C. § 7482(a)(1).

3
meaning they filed a single tax return reflecting their joint tax
liability. Despite allowing corporate groups to file a single
return, the applicable tax laws require a mixed approach of
calculating some aspects of the group’s taxes as though the
entities were a single taxpayer and calculating others as if
each member of the group were a separate taxpayer. 13
Mertens Law of Federal Income Taxation § 46:1 (2016 ed.).
This approach—called the “consolidated return regime”—
reflects how the IRS has chosen to exercise its broad
discretion to issue regulations for consolidated returns “to
reflect the income-tax liability” of the group and “to prevent
avoidance of such tax liability.” 26 U.S.C. § 1502.

The possibility of separate treatment nonetheless
creates the potential for the group to deflect its tax liability by
using stock sales to claim a second deduction for a single loss
at the subsidiary (such as a loss on the subsidiary’s sale of an
asset). See Lawrence Axelrod, 1 Consolidated Tax Returns
§ 18:2 (4th ed. 2015). This is known as a double deduction,
or more technically as loss duplication. It may occur because
by definition the parent company in a corporate group owns
all or most of the stock in its subsidiaries. See 26 U.S.C. §
1504(a). All else being equal, the value of the parent’s stock
depends on the value of the subsidiary’s assets. If the
subsidiary’s assets decline in value, the parent’s stock will as
well. If the subsidiary sells those assets (which may include
stock and other securities) at a loss, it is generally able to
claim a deduction for those losses. See 26 U.S.C. § 165(a)
(“General rule.--There shall be allowed as a deduction any
loss sustained during the taxable year and not compensated
for by insurance or otherwise”). If the parent and subsidiary
are viewed as separate entities, the parent is able to sell its
stock of the subsidiary at a loss and claim a deduction for that
loss as well. See Axelrod, supra, § 18:2. But in fact the
overall group has only suffered one economic loss though it
was deducted twice. For example, suppose that parent P has a

4
wholly owned subsidiary S and its investment in S’s stock is
worth $100. After one of S’s assets declines in value by $50,
S sells the asset and deducts a $50 loss under § 165. P’s stock
value in S also declines by $50, and if P and S are viewed
separately, P is able to sell its stock in S and deduct a further
loss of $50 under § 165. The consolidated group is thus able
to deduct $100 for a single economic loss of $50 resulting
from the decline in value of S’s asset.
To prevent double deductions, the IRS has
promulgated numerous regulations requiring that consolidated
taxpayers be treated as a single entity for stock sales. Of
particular relevance to the events of this case is the former
Treas. Reg. § 1.1502-20. Starting in the early 1990s, it
prevented, among other things, double deductions when the
parent’s loss on its sale of stock occurred before the
subsidiary recognized its loss. See Consolidated Return
Regulations; Special Rules Relating to Dispositions and
Deconsolidations of Subsidiary Stock, 55 Fed. Reg. 9426-01,
9427 (Mar. 14, 1990). In July 2001, however, the Federal
Circuit’s decision in Rite-Aid v. United States, 255 F.3d 1357
(Fed. Cir. 2001), invalidated the pertinent portion of
§ 1.1502-20 as beyond the IRS’s power to issue because it
addressed a problem not specifically attributable to the filing
of consolidated returns. Id. at 1360. Though Rite-Aid has not
been construed to annul any other consolidated return
regulation preventing duplicated loss, invalidating § 1.1502-
20 meant that in its immediate aftermath there was no
regulation expressly preventing a double deduction when the
parent’s stock loss occurred before the subsidiary’s asset loss.
In contrast, Rite-Aid left intact the regulatory prohibition on
double deductions where the transactions are structured in
such a way that the losses occur in reverse order, i.e., the
subsidiary’s loss is recognized before the parent’s loss. See
Treas. Reg. § 1.1502-32.

5
 In the aftermath of Rite-Aid, the Duquesne group
arranged a series of transactions in which DLH incurred a
loss on AquaSource stock, and then AquaSource incurred
losses on the sale of its assets (which were stock interests that
AquaSource held directly and indirectly in eight of its own
subsidiaries). DLH formed AquaSource in the late 1990s as a
wholly owned subsidiary to expand into the water utility
business. It funded AquaSource through a series of
contributions in return for AquaSource stock. Through
February 2001, DLH contributed approximately $223 million
in return for 50,000 shares of AquaSource stock. Though
DLH contributed a similar amount to AquaSource in the years
thereafter, it increased its holdings to 1.2 million shares of
AquaSource stock. AquaSource used these contributions to
purchase more than 50 water utility companies, which
became both its subsidiaries and its assets. It began to lose
money, however, and in 2000 the Duquesne group decided to
explore the sale of AquaSource’s assets.

The transactions before us began on December 31,
2001, just before the deadline would expire for the IRS to file
a petition for certiorari in Rite-Aid. DLH transferred 50,000
shares of AquaSource stock to Lehman Brothers, which
Lehman valued at $4 million, as payment for Lehman’s
services rendered to AquaSource. DLH determined that these
particular 50,000 shares of stock were the shares that it had
possessed prior to February 2001 and thus accounted for $223
million of its investment. After various adjustments, DLH
claimed a capital loss of $199 million on the transfer of stock
to Lehman Brothers (the “2001 loss”). On its 2001 tax return,
the Duquesne group carried back $161 million of that loss to
tax year 2000 and claimed a tentative refund of $35 million.

Shortly thereafter, the IRS announced the regulatory
response to Rite-Aid. It declined to litigate further the validity
of Treas. Reg. § 1.1502-20 and instead announced that it

6
would issue new regulations governing stock losses. I.R.S.
Notice 2002-11, 2002-1 C.B. 526 (Jan. 31, 2002). It did so in
early March 2002 by issuing temporary regulations that
included Treas. Reg. § 1.337(d)-2T, which applied to stock
losses occurring on or after March 7, 2002. See Loss
Limitation Rules, 67 Fed. Reg. 11034-01, 11036-37 (Mar. 12,
2002). 2 Though these regulations “[did] not disallow [a]
stock loss that reflects . . . built-in asset losses of a subsidiary
member,” I.R.S. Notice 2002-18, 2002-1 C.B. 644 (Mar. 9,
2002), the IRS published as guidance a draft of a new
regulation barring double deductions in October 2002. See
Guidance Under Section 1502; Suspension of Losses on
Certain Stock Dispositions, 67 Fed. Reg. 65060-01 (Oct. 23,
2002). The new regulation was issued in final form as Treas.
Reg. § 1.1502-35T and applied retroactively to stock sales
occurring on or after March 7, 2002.

In this complex and shifting regulatory environment,
the Duquesne group thereafter incurred further losses on the
sale by AquaSource of its assets. 3 It did so through a series of

2
The IRS also issued Treas. Reg. § 1.1502-20T, which
applied retroactively to stock losses occurring between the
date of the Rite-Aid decision and March 7, 2002. See Loss
Limitation Rules, 67 Fed. Reg. 11034-01, 11037 (Mar. 12,
2002).
3
At oral argument before a prior appellate panel,
Duquesne’s counsel represented that AquaSource had
different subsidiaries in 2001 versus 2002-2003 in support of
its claim that the case should be remanded for further
development of the factual record. Oral Arg. Trans. at 18
(Nov. 18, 2015). When opposing counsel pointed out that the
record demonstrated that AquaSource’s subsidiaries had in

7
transactions after March 2002 and continuing into 2003 in
which AquaSource sold all of its stock in eight subsidiaries,
and thus the resulting losses were also stock losses. 4 In 2002,
these transactions resulted in capital losses for AquaSource
totaling $59.5 million (the “2002 loss”), all of which the
Duquesne group carried back to tax year 2000 on its
consolidated return. Based on the 2002 loss and additional
carrybacks, the group received a tentative refund of $12
million. The sale transactions in 2003 yielded $192.8 million
in capital losses for AquaSource after various adjustments
(the “2003 loss”). Duquesne carried all of the 2003 losses
back to 2000 on its consolidated return and, based on that loss
and additional carrybacks, received a tentative refund of $40
million. Aggregating the 2002-2003 losses, the Duquesne
group deducted approximately $252 million in capital losses
in addition to the $199 million it had already claimed for the
2001 loss.

Though the IRS initially declined to challenge
Duquesne’s deductions in a 2004 audit, it subsequently

fact remained the same, see Oral Arg. Trans. at 26-27 (July
12, 2016), Duquesne abandoned this claim.
4
That AquaSource incurred stock losses on the sale of
its assets was the source of much terminological confusion in
the litigation of this case. As Duquesne did not make this
clear in its individual briefing, and though the record
remained unclear whether all of AquaSource’s losses were
incurred on the sale of stock, the IRS acknowledged that at
least 90% of the losses were stock losses. IRS Supp. Br. at 4.
As this issue is not ultimately relevant to our analysis, we
accept Duquesne’s representation that all of AquaSource’s
losses were stock losses for purposes of this appeal.

8
determined that the Duquesne group had claimed a double
deduction; that is, the 2001 loss and the 2002-2003 losses
reflected a single economic loss in the form of an underlying
decline in the value of AquaSource’s subsidiaries. Based on
its calculations, the IRS concluded that $199 million of these
losses were disallowed under the Ilfeld doctrine. 5 Duquesne
strongly disagreed, and as the parties were unable to resolve
their dispute out of court, the IRS sent Duquesne a formal
notice of deficiency in 2010.

Duquesne began proceedings in the Tax Court by
filing a petition for relief. Rather than following the normal
pretrial procedure, the parties chose to forgo discovery and
filed cross-motions for summary judgment focusing on
whether the Ilfeld doctrine applied. While these motions
were pending, the Tax Court issued its decision in Thrifty Oil
Co. v. Commissioner, 139 T.C. 198 (2012). There the Court
relied on Ilfeld to disallow certain duplicative losses that a
consolidated taxpayer had claimed based on environmental
remediation expenses. As a result, the Tax Court allowed the
parties to this case to file supplemental briefs addressing
Thrifty Oil.

Thereafter the Tax Court granted the IRS’s motion for
summary judgment and concluded that the Ilfeld doctrine
remained good law in our Circuit. Turning to the factual
record, the Tax Court concluded that the IRS met its burden
to show that the 2002-2003 losses duplicated the $199 million
deduction taken in 2001, as they reflected the same economic
loss. It also rejected Duquesne’s argument that it satisfied the
Ilfeld doctrine because its deductions were authorized under
5
The IRS also asserted that a $228 million excess loss
account should be triggered for tax year 2005, which would
have been taxable as income for that year. As this claim is no
longer at issue, we need not decide it.

9
the applicable statutes and regulations as well as its further
argument that the statute of limitations had, in any event,
lapsed. The Tax Court thus disallowed $199 million of the
2002-2003 losses and ordered Duquesne to repay $36.9
million of the refunds it had received based on those losses.
This appeal followed.

II. ANALYSIS

Did the Tax Court properly rely on the Ilfeld doctrine
to disallow $199 million of the 2002-2003 losses? Duquesne
contends that the Court erred for three reasons: (1) the factual
record was inadequate to support summary judgment; (2) the
Ilfeld doctrine does not support disallowing the losses; and (3)
the IRS’s claims are at least partially barred by the statute of
limitations. We consider each contention in turn.

A. Adequacy of the Record

As the Ilfeld doctrine requires a clear declaration of
intent to allow a double deduction for a single economic loss,
its application is premised on a factual question: Did the
deductions claimed by the taxpayer reflect the same economic
loss? See Thrifty Oil, 139 T.C. at 212; see also Denton &
Anderson Co. v. Kavanagh, 164 F. Supp. 372, 378 (E.D.
Mich. 1958). Here the question is whether the 2001 loss and
the 2002-2003 losses are truly duplicative—that is, did they
both reflect the decline in value of the same AquaSource
assets? The Tax Court held that the IRS proved the absence
of a genuine dispute of material fact on this point and that
$199 million of the losses were deducted twice. Though
Duquesne contends that the IRS failed to submit sufficient
evidence to show the existence of a double deduction or its
amount, we disagree in light of Duquesne’s failure to present
any evidence to the contrary despite having all of the relevant

10
documents in its possession and the Supreme Court’s decision
in McLaughlin v. Pacific Lumber Co., 293 U.S. 351 (1934).

We exercise plenary review over the Tax Court’s grant
of summary judgment, Hartmann v. Comm’r of Internal
Revenue, 638 F.3d 248, 249 (3d Cir. 2011) (per curiam), and
the summary judgment standard in Tax Court Rule 121 is
identical to that contained in Federal Rule of Civil Procedure
56, see Rivera v. Comm’r of Internal Revenue, 89 T.C. 343,
346 (1987). The party moving for summary judgment bears
the initial burden to show the absence of a genuine dispute of
material fact, see Celotex Corp. v. Cattrett, 477 U.S. 317,
323 (1986), and thus all that remains is to resolve legal issues.

Though the initial burden of the IRS may be heavy in
some cases, it is far lighter when, as here, it seeks to collect
unpaid taxes. “It is well established that[,] as a general
matter, the [IRS]'s determination of deficiency is presumed
correct, and the taxpayer bears ‘the burden of proving it
wrong.’” Cebollero v. Comm’r of Internal Revenue, 967 F.2d
986, 990 (4th Cir. 1992) (quoting Welch v. Helvering, 290
U.S. 111, 115 (1933)). “This presumption is a procedural
device that places the burden of producing evidence to rebut
the presumption on the taxpayer.” Anastasio v. Comm’r of
Internal Revenue, 794 F.2d 884, 886 (3d Cir. 1986). This is
so because “an income tax deduction is a matter of legislative
grace and . . . the burden of clearly showing the right to the
claimed deduction is on the taxpayer.” INDOPCO, Inc. v.
Comm’r of Internal Revenue, 503 U.S. 79, 84 (1992). The
IRS is thus essentially in the position of a defendant in a civil
case who may meet its initial burden merely by pointing to
the absence of evidence supporting essential elements of the
non-moving party’s case. See Celotex, 477 U.S. at 322-23;
see also Cal-Farm Ins. Co. v. United States, 647 F. Supp.
1083, 1086-87 (E.D. Cal. 1986) (collecting cases stating that
the presumption alone is sufficient to support summary

11
judgment when the taxpayer fails to rebut it); Mitchell v.
Comm’r of Internal Revenue, 38 T.C.M. (CCH) 854 (T.C.
1979) (describing the IRS’s deficiency determination as
“prima facie correct” and concluding that it must be sustained
where “[the taxpayers] have presented no evidence to show
that they are entitled to additional deductions”) (citing Tax
Ct. R. 142(a)).

Duquesne now challenges the adequacy of the record
before the Tax Court, yet it made numerous tactical decisions
to limit that record. While conceding that it has possession of
the records necessary to determine whether its deductions
reflect the same economic loss, Oral Arg. Trans. at 19-20
(Nov. 18, 2015), it nonetheless agreed to limit the scope of
discovery before summary judgment in order to conserve
resources. Duq. Br. at 43. It then filed a motion for summary
judgment raising “only legal issues,” notably whether the
Ilfeld doctrine applied. Id. at 37 n.10.
The IRS responded with a cross-motion for summary
judgment. It pointed to evidence initially filed in an exhibit
to Duquesne’s motion regarding the size and timing of the
deductions. While Duquesne challenged the method the IRS
used to calculate the amount of duplicated losses, it did not
present any evidence rebutting the latter’s claim of a double
deduction and merely promised that “[it] w[ould] challenge at
trial the [IRS’s] erroneous determination of the amount of
purported duplicated losses.” J.A. at 670. After allowing the
parties to submit supplemental briefs on the effect of the
Thrifty Oil decision, the Tax Court held that the IRS had
pointed to sufficient evidence of duplicate losses; hence
Duquesne’s “bald assertions” of a factual dispute were too
little to merit a trial. Id. at 45.

Duquesne claims that the Tax Court applied an
“irrebuttable presumption” based on Thrifty Oil that the

12
deductions were duplicative. Reply Br. at 18. While we may
affirm on any basis supported by the record, it is worth noting
that this argument misrepresents the Court’s reasoning.
Though it stated that Duquesne’s deductions “represent the
same economic loss[es],” J.A. at 42, borrowing this turn of
phrase from Thrifty Oil was not invoking an irrebuttable
presumption. Nor did its conclusion that Thrifty Oil was
consistent with Third Circuit precedent become a
presumption that barred rebuttal. Instead, the Tax Court
determined that the IRS had met its burden based on the
evidence in the record, including the size and timing of the
deductions at issue. There is thus nothing irrebuttable in the
Court’s analysis; it concluded that Duquesne simply failed to
rebut the IRS’s claims as required by ordinary summary
judgment practice.

Though Duquesne contends that it was not required to
present any evidence because the IRS’s case contained
“unexplained gaps,” Reply Br. at 15 (quoting O’Donnell v.
United States, 891 F.2d 1079, 1082 (3d Cir. 1989)), we agree
with the Tax Court that the IRS demonstrated the absence of a
genuine dispute of material fact. As noted, it could have met
its burden here merely by pointing to the absence of evidence
supporting Duquesne’s position that the losses were not
duplicative.

The Supreme Court’s 1934 decision in McLaughlin
also supports the IRS. There a subsidiary between 1920 and
1923 had claimed operating losses three times greater than the
capital invested in it by the parent, and the parent had
attempted to claim further investment losses when it
liquidated the subsidiary. The Court reasoned that “the
circumstances tend strongly to indicate” that the losses had a
common source in the failing business of the subsidiary. Id.
at 357. In addition, as “[p]resumably [the taxpayer] had
within its control the records showing facts that would fully

13
disclose the relations between such losses,” if they were not
duplicative “it reasonably may be presumed that [the
taxpayer] would have shown that fact.” Id. at 356-57. The
Court thus upheld the determination of the IRS that the
parent’s losses should be disallowed.

Here the IRS’s evidence regarding the size and timing
of the losses similarly reveals that Duquesne claimed losses
significantly greater than its net investment in AquaSource.
Aggregating the 2001 loss and the 2002-2003 losses as
provided on AquaSource’s Stock Sales Adjustment
calculation sheet, J.A. at 744, the IRS determined the
Duquesne group deducted far more in aggregate capital losses
than its net investment in AquaSource, the difference being
$281 million. As in Pacific Lumber, the excess of losses over
investment occurred over a few years before the subsidiary
ceased operation. This “tend[s] strongly to indicate” a double
deduction stemming from the declining value of
AquaSource’s assets. Id. at 357. Because Duquesne has
possession of the relevant documents, we presume that it
would have demonstrated that the losses came from different
sources if that were the case. 6 See id. at 356-57. Though it

6
Our decision in Nat’l State Bank v. Federal Reserve
Bank of N.Y., 979 F.2d 1579 (3d Cir. 1992), is not to the
contrary. There National State Bank (“NSB”) bore the
burden of proof as plaintiff to show that the Federal Reserve
Bank of New York had negligently handled certain checks
that were part of a check-kiting scheme. The District Court
granted summary judgment because the Federal Reserve had
taken too long to inform NSB that it had lost the checks in
dispute. We reversed, as there was no evidence on when the
checks were lost or by whom and the Federal Reserve likely
had the relevant records. This is not our case; Duquesne had

14
attempts to distinguish Pacific Lumber on the grounds that it
was not decided at summary judgment and involved a less
complex set of corporate relationships, that is off target in
light of the low burden the IRS faced here.

The IRS also met its burden to show the amount of
duplicative losses. As the aggregate excess of deducted
losses over net investment implies a double deduction, the
IRS compared the amount of excess to each deduction. That
excess was greater than the deduction claimed for the 2001
stock loss ($199 million), and thus the IRS concluded that at
least the amount of the 2001 stock loss was deducted twice.
This calculation of $199 million in duplicated loss thus
abandoned any IRS claim in addition to that loss (and, as in
Pacific Lumber, it reflected the parent’s claimed loss in the
transaction). Duquesne cites no case in which greater proof
was required, nor does it present any evidence that the $199
million figure is inaccurate. 7 We thus see no unexplained

the relevant records and did not produce documents to
counter the IRS’s allegations of duplicate losses.
7
In its brief to the Tax Court, Duquesne suggested that
the IRS should have used a different valuation for the
AquaSource subsidiaries as of December 31, 2001, which
Duquesne asserted would have yielded a different amount of
duplicated loss. J.A. at 660. This argument is waived, as
Duquesne failed to raise it in its opening brief on appeal. See
United States v. Hoeffecker, 530 F.3d 137, 162 (3d Cir.
2008). Moreover, the issue is not material, as the amount of
the 2002-2003 losses must be determined by comparing
AquaSource’s adjusted basis in its subsidiaries’ stock to the
sale price. See 26 U.S.C. § 1001(a). If this calculation uses
Duquesne’s preferred valuation, it does not reduce the amount
of duplicated loss. See IRS Br. at 28 n.5.

15
gaps in the IRS’s case that renders summary judgment in its
favor improper.

Duquesne nonetheless contends that the IRS had to
“trace” the losses between Duquesne’s sale of AquaSource
stock and the losses incurred on the sale of particular
AquaSource assets (that is, the sale of particular AquaSource
subsidiaries). Tracing is necessary, Duquesne asserts, to rule
out the possibility that unspecified intervening event(s)
somehow account for a portion of the 2002-2003 asset losses.
But because Duquesne fails to present evidence of any such
events or of any other effect that tracing would have on the
amount of duplicated loss, this is merely the kind of
speculation that does not defeat summary judgment. 8

In the alternative, Duquesne contends that the District
Court abused its discretion by failing to order further
discovery before deciding the IRS’s summary judgment
motion. Like Federal Rule of Civil Procedure 56(d), Tax

8
To the extent Duquesne contends that tracing is still
required as a matter of law, we reject it as unsupported by any
authority. Though it relies on Edward Katzinger Co. v.
Commissioner, 44 B.T.A. 533 (1941), aff’d, 129 F.2d 74 (7th
Cir. 1942), that case held it was the taxpayer who failed to
show that its deductions reflected different losses. See id. at
76. The IRS’s 2007 comments on a proposal to prevent
double deductions by disallowing asset losses are also of
little-to-no use. See Unified Rule for Loss on Subsidiary
Stock, 72 Fed. Reg. 2964-01, 2976 (Jan. 23, 2007). While the
IRS rejected a proposal that would “present considerable
tracing issues,” id., it was never adopted and hence is
irrelevant to its burden at summary judgment.

16
Court Rule 121(e) confers discretion on the trial court to
order further discovery when the non-moving party files a
motion or affidavit stating that more discovery is needed for it
to justify its opposition to summary judgment. Duquesne,
however, never claimed that it could not justify its opposition
without further discovery, and thus it cannot claim the
protection of Rule 121(e).

Despite its failure to comply with the Rule, Duquesne
contends that our decision in Sames v. Gable, 732 F.2d 49 (3d
Cir. 1984), required the Tax Court to order further discovery
anyway. There two former police officers filed a civil rights
claim against local police officials for retaliatory discharge.
While their interrogatory requests were pending, the
defendants moved for summary judgment and the plaintiffs
failed to request further discovery under Rule 56(d) (then
codified as Rule 56(f)). Relying on the well-settled rule that
summary judgment should not be granted while the material
facts remain in the moving party’s possession, we held “that it
was error for the district court to grant defendants’ motion for
summary judgment while pertinent discovery requests were
outstanding” despite the plaintiffs’ failure to comply with the
Federal Rules. Id. at 51-52. Here, however, Duquesne had
possession of the relevant facts and chose to limit discovery
for reasons only it knows. In sum, it gambled that its legal
arguments were strong enough to win without creating the
factual record necessary to rebut the IRS’s position. After that
gamble failed, Sames did not require the District Court to
delay summary judgment.

We thus conclude that the Tax Court properly held that
$199 million of the 2001 loss and 2002-2003 losses were
deducted for the same underlying economic loss. After
determining that there was no genuine dispute that $199
million in deducted losses for 2002-2003 were duplicative,
the Tax Court disallowed that amount of those losses under

17
the Ilfeld doctrine. 9 We thus turn to Duquesne’s next
contention: that although its consolidated deductions may be
duplicative, the Ilfeld doctrine nonetheless does not support
the Tax Court’s decision to disallow them.

B. The Ilfeld Doctrine
Ilfeld requires a clear declaration allowing double
deductions for the same loss on consolidated returns.
Duquesne contends that the text of 26 U.S.C. § 165 provides
the required authority to satisfy Ilfeld; if § 165 alone proves
insufficient, the combination of it with the applicable
regulations in effect provides clear authority; and in any event
the regulations alone—particularly §1.337(d)-2T—suffice. In
evaluating these arguments we review anew the Tax Court’s
legal conclusions. See Anderson v. Comm’r of Internal
Revenue, 698 F.3d 160, 164 (3d Cir. 2012).

1. The Current Status of the Ilfeld Doctrine

Duquesne does not dispute that Ilfeld remains good
law. While there is dispute as to the scope of the Ilfeld
doctrine — that is, whether it applies outside the consolidated
return context — for consolidated taxpayers it continues to
require that a statute and/or regulation specifically authorize a
double deduction for an underlying economic loss.
9
Duquesne also argues that the Tax Court erred in
how it allocated the $199 million in disallowed loss between
the 2002 loss and the 2003 loss. The Tax Court allocated $59
million to the 2002 loss and the remaining $140 million to the
2003 loss. J.A. at 48-49. This allocation was proposed by
Duquesne, however, and agreed to by the IRS. Id. at 47-48.
Moreover, as we reject below the statute-of-limitations
argument made by Duquesne, the allocation of losses between
2002 and 2003 is irrelevant.

18
 Ilfeld, decided in 1934, concerned consolidated tax
returns filed by Charles Ilfeld Co. and two wholly owned
subsidiaries between 1917 and 1929. During this period, the
subsidiaries claimed substantial operating losses and were
eventually liquidated. Ilfeld Co. then attempted to deduct
losses on its investment in the defunct subsidiaries. Though
these investment losses could not be deducted under the
regulations in effect at the time, the Supreme Court
emphasized that, because the investment losses were already
reflected in the subsidiaries’ operating losses, allowing them
“would be the practical equivalent of a double deduction.”
Ilfeld, 292 U.S. at 68. Thus, “[i]n the absence of a provision
of the [applicable statute] definitely requiring it, a purpose so
opposed to precedent and equality of treatment of taxpayers
will not be attributed to lawmakers.” Id. “[D]efinitely
requiring” a provision to authorize a double deduction for the
same economic loss is a very high hurdle, though how high is
debatable when Ilfeld itself stated that “[i]n the absence of a
provision in the [applicable act] or regulation that fairly may
be read to authorize [a double deduction], the deduction
claimed is not allowed.” Id. at 66.

How to interpret this disparity in the degree of
certainty was taken care of when the Supreme Court restated
Ilfeld as requiring a “clear declaration of intent by Congress”
to authorize a double deduction in Skelly Oil Co., 394 U.S. at
684 (citing United States v. Ludey, 274 U.S. 295 (1927)).
Though a declaration by Congress is stated, we do not purport
to rule out clear statements of intent set out in regulations the
IRS Commissioner is empowered to prescribe. See Ilfeld, 292
U.S. at 68. Indeed, Ilfeld itself noted that because “[t]here
[was] nothing in the [Revenue Act of 1928] that purport[ed]
to authorize double deduction of losses or in the regulations
to suggest that the commissioner construed any of its
provisions to empower him to prescribe a regulation that
would permit consolidated returns to be made on the basis

19
now claimed by [Ilfeld Co.],” it could not deduct its
duplicative investment losses. Id. (emphasis added).

Ilfeld quickly became a key precedent in the
consolidated return context. In Pacific Lumber the Supreme
Court relied on Ilfeld for the principle that “[l]osses of [the
subsidiary] that were subtracted from [the group’s] income
are not directly or indirectly again deductible.” 293 U.S. at
355. Our Court similarly relied on Ilfeld to disallow
deductions for consolidated returns. See Grief Cooperage
Corp. v. Comm’r of Internal Revenue, 85 F.2d 365, 366 (3d
Cir. 1936); see also Comm’r of Internal Revenue v. National
Casket Co., 78 F.2d 940, 941-42 (3d Cir. 1935).

Some of our sister Circuits have extended the Ilfeld
doctrine beyond the consolidated-return context — see, e.g.,
Chicago & N.W.R. Co. v. Comm’r of Internal Revenue, 114
F.2d 882, 887 (7th Cir. 1940) (appropriate method of
depreciation for railroad property); Comar Oil Co. v.
Helvering, 107 F.2d 709, 711 (8th Cir. 1939) (deductibility of
anticipated inventory losses); Marwais Steel Co. v. Comm’r
of Internal Revenue, 354 F.2d 997, 997-99 & n.1 (9th Cir.
1965) (investment losses among corporate taxpayers filing
separate returns).

All of the cases questioning the continued viability of
Ilfeld have occurred outside the consolidated-return context.
For example, in a case in our Circuit—Miller’s Estate v.
Commissioner, 400 F.2d 407 (3d Cir. 1968)—the IRS argued
that certain deductions flowing from estates’ donations to
charitable trusts were double deductions. We distinguished
Ilfeld on the ground that it concerned “the peculiar income tax
context of consolidated corporate income tax reporting,” and
thus it “cannot be regarded as a legitimate canon of estate tax
interpretation to assist the court in this case.” Id. at 411
(internal footnotes omitted).

20
 The Supreme Court indicated that Ilfeld might apply
beyond consolidated returns the year following Estate of
Miller in Skelly Oil, where it required a non-consolidated
taxpayer to recalculate certain refunds based on oil and gas
revenues. 394 U.S. at 684. But decades later the Court in
Gitlitz v. Commissioner, 531 U.S. 206 (2001), may have
implied that Ilfeld does not apply other than for consolidated
returns. Addressing an argument abandoned by the IRS—
that, if shareholders of an S corporation (in which income is
passed through to shareholders and then taxed as personal
income) were permitted to pass through a discharge of
indebtedness before reducing tax attributes, the shareholders
may have gotten a “double windfall” by being partially
exempted from taxation yet able to increase their basis and
deduct previously suspended losses—the Court elected not to
address this “policy concern.” Id. at 219-20. Though Justice
Breyer’s dissent criticized the majority for failing to apply
Ilfeld, see id. at 224, the majority did not so much as mention
that decision or Skelly Oil; it certainly did not purport to
overrule them. In any event, after Gitlitz the Ilfeld doctrine
thus remains good law in the consolidated-return context. 10

10
In its supplemental briefing, Duquesne also points to
the Supreme Court’s decision in United Dominion Industries,
Inc. v. United States, 532 U.S. 822 (2001), as an indication of
Ilfeld’s diminished role. Though United Dominion is a
consolidated-return case in which the Court declined to rule
in the IRS’s favor with respect to its double-deduction
argument, it does not apply here because the calculation at
issue “[was] not in and of itself the basis for any tax event.”
Id. at 834. What this means is that the Court rejected the
analogy to Ilfeld because there was no loss deduction at issue.

21
 With this in mind, we proceed to what is currently
required for a statute to satisfy the Ilfeld doctrine. It reflects
“[t]he presumption … that statutes and regulations do not
allow a double deduction” for the same economic loss.
United Telecomm. v. Comm’r of Internal Revenue, 589 F.2d
1383, 1387-88 (10th Cir. 1978). This presumption must be
overcome by a clear declaration in statutory text or a properly
authorized regulation.

2. Deduction of the 2002-2003 Losses Under § 165

Duquesne argues that a specific authorization for
duplication of loss here may be found in the text of § 165. In
particular, it points to subsections (a) and (f) of that provision.
As these subsections “provide[] for deduction of losses,
including capital losses,” and the 2002-2003 losses were
incurred on AquaSource’s sale of stock in its subsidiaries,
Duquesne contends that the “deductions fall squarely within
§ 165’s text” even if they are duplicative. Duq. Br. at 15. We
conclude otherwise.

The general rule of § 165(a) is that “[t]here shall be
allowed as a deduction any loss sustained during the taxable
year and not compensated for by insurance or otherwise.” 26
U.S.C. § 165(a). That the subsection allows “a deduction”
for “any loss” indicates that it allows a single deduction for a
single loss. Id. (emphases added). The brief text of the
subsection certainly contains no express authorization of a
double deduction, and we are unaware of any evidence of
congressional intent to that effect. Moreover, as the Tax
Court noted, § 165(a) is quite broad in authorizing a
deduction for any loss and is thus a poor candidate to satisfy
the Ilfeld requirement of explicit approval for duplicating the
underlying economic loss. See J.A. at 46-47.

22
 Section 165(f)’s limitation on the deductibility of
capital losses also does not support Duquesne’s position. It
provides that “[l]osses from sales or exchanges of capital
assets shall be allowed only to the extent allowed in sections
1211 and 1212.” 26 U.S.C. § 165(f). 26 U.S.C. § 1211(a)
provides that corporations may deduct capital losses only to
offset capital gains from the sale of different assets, and
section 1212(a)(1)(A) allows corporations to carry back
capital losses up to three years. These limitations are
irrelevant to the prospect of a double deduction for the same
economic loss, and in any event they are not at issue here.

Beyond the plain text of the statute, Ilfeld’s case-
specific holding sends a particularly strong signal that § 165
does not authorize a double deduction. At the time Ilfeld was
decided, the Revenue Act of 1928 was in effect and contained
a provision indistinguishable from the current § 165(a): “In
computing net income there shall be allowed as deductions:
… In the case of a corporation, losses sustained during the
taxable year and not compensated by insurance or otherwise.”
Revenue Act of 1928, ch. 852, § 23(f), 45 Stat. 791, 799-800.
The IRS brought this provision to the Court’s attention and
specifically argued that it did not allow Ilfeld Co. to claim a
double deduction. See Brief for Respondent at 22, Charles
Ilfeld Co. v. Hernandez, 292 U.S. 68 (No. 579), 1933 WL
63349 at *22. To uphold that assertion, Ilfeld had to reject
any assertion that this provision demonstrated a clear intent to
allow it.

While Duquesne contends that any consideration of the
Revenue Act by the Supreme Court was a mere dictum
because it had already determined that the taxpayer’s
deductions were barred by applicable regulations, this
contention runs into Pacific Lumber, which also interpreted
virtually identical statutory text. As no applicable regulations
governed the specific deductions, the Court determined

23
whether they were allowed under the applicable statute, the
Revenue Act of 1921. That Act contained the following
provision: “[For purposes of the corporate income tax], there
shall be allowed as deductions: … Losses sustained during
the taxable year and not compensated for by insurance or
otherwise …” Revenue Act of 1921, ch. 136, § 234(a)(4), 42
Stat. 227, 254-55. The IRS also brought this provision to the
Court’s attention in arguing that it did not authorize a double
deduction. See Brief for Petitioner at 20, McLaughlin v.
Pacific Lumber Co., 293 U.S. 351 (No. 125), 1934 WL 60331
at *20. The bottom line: to hold that the taxpayer’s losses
were not deductible a second time, Pacific Lumber rejected a
provision materially identical to § 165(a) without any help
from the consolidated return regulations. Ilfeld and Pacific
Lumber’s rejection of such a strikingly similar predecessor to
§ 165 as a rationale for a double deduction thus provides
further support for our conclusion that the current statute does
not provide that rationale.

We thus turn to Duquesne’s next argument: that even
if § 165 alone does not satisfy Ilfeld’s requirement of a clear
statement of intent to authorize a double deduction, the
combination of that provision with the applicable regulations
in effect does so.

3. Deduction of the 2002-2003 Losses Under § 165 and
Applicable Regulations

Duquesne asserts that it complied with all applicable
regulations in calculating AquaSource’s losses and then
deducted them under § 165. In particular, it calculated the
amount that AquaSource lost on sales of its subsidiaries’
stock per the regulations governing stock losses, primarily
Treas. Reg. §§ 1.1502-32 and 1.337(d)-2T, before deducting
those losses under § 165. As Duquesne thus purports to have
complied with the requirements of both on-point regulations

24
and an on-point statute, it contends that the Tax Court’s
decision in Woods Investment Co. v. Commissioner, 85 T.C.
274 (1985), teaches that Ilfeld’s clear statement rule is
satisfied and we should rule in its favor by applying the
regulations and the Code “as written.” Id. at 282.

In light of the complexity of the regulatory framework,
Duquesne’s argument requires some unpacking. The first
step in determining the amount of the 2002-2003 losses was
calculating AquaSource’s basis in the stock of its subsidiaries.
Under 26 U.S.C. § 1001(a), the amount of loss for tax
purposes on the sale of an asset, including stock, is equal to
the taxpayer’s basis minus the sale price. For stock sales by
consolidated taxpayers, such as AquaSource, Tax Regulation
§ 1.1502-32 requires the parent to adjust its basis to reflect
various events at the subsidiary, including gains and losses. It
is undisputed that the Duquesne group made all the
adjustments required by § 1.1502-32 to AquaSource’s basis in
its subsidiaries’ stock and calculated the amount of loss for
tax purposes accordingly.

The Duquesne group then calculated the amount of
allowable loss on AquaSource’s stock sales per § 1.337(d)-
2T. Paragraph (a)(1) of the regulation provides the general
rule that “[n]o deduction is allowed for any loss recognized
by a member of a consolidated group with respect to the
disposition of stock of a subsidiary.” Paragraph (c)(2),
governing “[a]llowable [l]oss,” then gives a partial exception
to this general rule: “Loss is not disallowed under paragraph
(a)(1) of this section … to the extent the taxpayer establishes
that the loss … is not attributable to the recognition of built-in
gain on the disposition of an asset (including stock and
securities).” Treas. Reg. § 1.337(d)-2T(c)(2). A built-in gain
occurs when an asset increases in value before being sold. As
AquaSource’s subsidiaries had declined in value, none of the
2002-2003 losses reflected built-in gains.

25
 With this backdrop to Duquesne’s contention of
statutory/regulatory compliance in conjunction with its
reading of Woods, the Government asserts that Woods does
not control, as here no statute or regulation permitted a double
deduction with anything approaching the specificity that
compelled the outcome in that case. In Woods the issue was
whether the Ilfeld doctrine applied to the calculation of a
parent’s basis in its subsidiary’s stock. At the time
Regulation § 1.1502-32 required the parent to adjust its basis
in line with the subsidiary’s earnings and profits, and 26
U.S.C. § 312(k) required the parent to calculate those
earnings and profits using a straight-line method of
depreciation. Despite its awareness of the potential for
straight-line depreciation to result in the practical equivalent
of a double deduction, the IRS had failed to amend § 1.1502-
32 for almost twenty years. The Tax Court reasoned that the
details of the consolidated return regime were “essentially a
legislative and administrative matter,” and thus it was not a
court’s institutional role to engage in “judicial interference.”
Woods, 85 T.C. at 282. It thus held that it would “apply [the
consolidated return] regulations and the statute as written”: if
the IRS disliked this result, it “should use [its] broad power to
amend [its] regulations.” Id. Ilfeld did not support the IRS’s
position because “the detailed rules in [§ 1.1502-32], which
were enacted to comprehensively address the problem in
[Ilfeld], together with section 312(k), can fairly be read to
authorize the result herein . . . .” Id. at 283 (internal footnote
omitted). Woods thus teaches that when an on-point statute
and an on-point regulation authorize the taxpayer to act as it
did, courts should apply the statute and regulation as written.

The Tax Court clarified in Wyman-Gordon Co. v.
Commissioner, 89 T.C. 207 (1987), that Woods may not apply
in the absence of a statute and/or regulation clearly
authorizing the disputed deductions. At issue was whether
Code § 312(1) and Regulation § 1.1502-32 authorized the

26
taxpayer to include discharge of indebtedness income in the
calculation of its subsidiaries’ earnings and profits. The
Court distinguished Woods on the ground that the latter’s
holding “was based in large part on § 312(k), which
specifically require[d]” the taxpayer to compute its basis as it
had, whereas “there exist[ed] no comparable statutory
provision that require[d] inclusion of discharge of
indebtedness income in earnings and profits.” Id. at 219. As
no specific provision of any regulation, including § 1.1502-
32, supported the taxpayer’s position, the Court disallowed
the deductions as an impermissible duplication of loss. Id. at
224. Though there was thus no on-point regulation at issue in
Wyman-Gordon, it nonetheless indicates that Woods’
instruction to apply the Code and regulations as written
requires a statute and/or regulation affirmatively permitting
the taxpayer to act as it did. 11

Section 165, however, did not specifically authorize
Duquesne to claim a duplicative deduction of the 2002-2003
losses. As discussed above, its subsection (a) broadly allows
a single deduction for any loss, but it does not contemplate
the possibility of a double deduction. Moreover, as a
materially identical statute was at issue in Ilfeld and Pacific
Lumber, those decisions send a particularly strong signal that

11
The Tax Court’s decision in CSI Hydrostatic
Testers, Inc. v. Commissioner, 103 T.C. 398 (1994), is not to
the contrary. In that case, the Court applied Woods when
legislative history indicated clearly that Congress intended the
statute in dispute to require including relevant income in the
calculation of its subsidiaries’ earnings and profits, and thus
no more specific statement was needed for the taxpayer to
adjust its basis under § 1.1502-32. See id. at 405.

27
§ 165 does not authorize a double deduction. Thus it is out of
play for coupling it to potentially applicable regulations.

4. Deduction of the 2002-2003 Losses Under Solely the
Regulations

Duquesne counters that, even absent § 165, its
regulatory compliance with §§ 1.1502-32 and 1.337(d)-2T
alone was sufficient to overcome any effect of Ilfeld. 12 Yet
we are not persuaded that these regulations are sufficiently
clear to preclude application of the Ilfeld doctrine. At the
outset, § 1.1502-32 does not support Duquesne’s position
because its basis adjustments address the prospect of a
duplicated loss in a transaction where the subsidiary
recognizes its loss before the parent engages in a stock sale.
Consolidated Return Regulations; Special Rules Relating to
Dispositions and Deconsolidations of Subsidiary Stock, 55
Fed. Reg. 9426-01, 9427 (Mar. 14, 1990). Where, as here,
the parent’s stock sale occurs first, the prospect of a double

12
Duquesne also posits that Treas. Reg. § 1.1502-80(a)
precludes application of the Ilfeld doctrine because it requires
separate treatment of parent and subsidiary in the absence of a
consolidated return regulation to the contrary. Though the
Tax Court once interpreted the regulation in this manner
without reasoned analysis, see Gottesman & Co., Inc. v.
Commissioner, 77 T.C. 1149, 1156 (1981), this interpretation
finds no support in its text. See Treas. Reg. § 1.1502-80(a)
(as amended March 2003) (“The Internal Revenue Code, or
other law, shall be applicable to the group to the extent the
regulations do not exclude its application”). We therefore
reject it and simply look to other law, including the Ilfeld
doctrine, to determine the proper treatment of stock losses for
entity groups filing consolidated returns.

28
deduction has been addressed by other regulations since §
1.1502-20 was issued in the early 1990s. Id.

We thus turn to the other regulation on which
Duquesne relies, § 1.337(d)-2T. As noted above, subsection
(a)(1) generally forbids any deduction for losses incurred on
sales of subsidiary stock, and subsection (c)(2) provides that
“[l]oss is not disallowed under paragraph (a)(1) . . . to the
extent the taxpayer establishes that the loss . . . is not
attributable to the recognition of built-in gain on the
disposition of an asset (including stock and securities).”
From this double negative (“not disallowed”) on a negative
(“not attributable”), Duquesne asks us to infer that any stock
losses not reflecting a built-in gain, including duplicative
losses, are deductible.

We do not believe that the structure and purpose of the
broader regulatory regime support Duquesne’s interpretation
of § 1.337(d)-2T. Consolidated taxpayers pay based on the
consolidated taxable income (CTI) of the group. United
Dominion Industries, Inc. v. United States, 532 U.S. 822, 826
(2001). CTI is calculated by combining the separate taxable
income (STI) of each member of the group and then
incorporating certain adjustments calculated on a
consolidated basis. Id. Adjustments on a consolidated basis
are an example of the consolidated return regime adopting the
“single entity” approach to prevent distortion of tax liability.
See American Standard, Inc. v. United States, 602 F.2d 256,
261-62 (Ct. Cl. 1979). This treats the entire consolidated
group as a single taxpayer and “reduce[s] the significance of
each member’s separate existence.” Don Leatherman, Why
Rite-Aid Is Wrong, 52 Am. L. Rev. 811, 815-16 (2003). The
Ilfeld doctrine also reflects a single-entity approach by
preventing the group as a whole from claiming duplicative
deductions that the separate existence of parent and

29
subsidiary would otherwise allow. See Axelrod, supra, §
18:2 (4th ed. 2015).

This structure for consolidated returns indicates the
intent that the 2002-2003 losses be dealt with on a
consolidated basis, and thus it reflects the same single-entity
approach as the Ilfeld doctrine. Section 1.1502-32, on which
Duquesne relies, has the express purpose of treating the group
as “a single entity.” Treas. Reg. § 1.1502-32(a)(1). And
Treas. Reg. §§ 1.1502-11(a)(3) and 1.1502-22 provide that
net capital gains and losses — that is, the overall amount of
capital gains or losses in a particular year — are those items
calculated on a consolidated basis.

The calculation of individual stock losses also occurs
on a consolidated basis, see Treas. Reg. 1.1502-12(j), and for
nearly thirty years these calculations have followed detailed
rules designed to prevent tax evasion. Before Rite-Aid,
§ 1.1502-20 limited stock losses based on various loss
disallowance factors. See Corporations; Consolidated
Returns—Special Rules Relating to Dispositions and
Deconsolidations of Subsidiary Stock, 56 Fed. Reg. 47379-01,
47379 (Sept. 19, 1991). Those factors primarily disallowed
stock losses reflecting a built-in gain in the subsidiary’s
assets. See Consolidated Return Regulations; Special Rules
Relating to Dispositions and Deconsolidations of Subsidiary
Stock, 55 Fed. Reg. 9426-01, 9427 (Mar. 14, 1990). Though
the sale of a capital asset with a built-in gain generally results
in tax liability, see 26 U.S.C. § 1231, a formerly common
type of transaction allowed consolidated taxpayers to arrange
a stock loss to offset that gain and thereby avoid tax
liability. 13 Section 1.1502-20 also had the distinct purpose of
13
In what was known as a “son of mirrors”
transaction, a consolidated taxpayer would purchase a new
subsidiary that had assets with built-in gains. See

30
preventing double deductions, see Consolidated Return
Regulations, 55 Fed. Reg. at 9427, which was reflected in the
duplicated loss factors. Specifically, it prevented double
deductions when, as here, the parent’s stock loss occurred
before the subsidiary recognized a loss. Id.

But because Rite-Aid invalidated § 1.1502-20, the IRS
had to issue new regulations limiting stock losses. In doing
so, it separated the rules for stock losses occurring on and
after March 7, 2002, primarily into two temporary
regulations: §§ 1.337(d)-2T and 1.1502-35T. 14 As noted, §
1.337(d)-2T(c)(2)’s exception to the general ban on losses for
subsidiary stock sales in consolidated tax regimes requires the
taxpayer to prove that a loss is not the result of a built-in gain.

Though the IRS acknowledged that § 1.337(d)-2T
“d[id] not disallow stock loss[es] that reflect[] . . . built-in
asset losses of a subsidiary member,” it informed taxpayers
that it “and Treasury believe that a consolidated group should
not be able to benefit more than once from one economic
loss” and would issue another regulation to prevent double
deductions. I.R.S. Notice 2002-18, 2002-1 C.B. 644 (Mar. 9,

Leatherman, supra, at 846-47 & n.172. It would then trigger
a gain by transferring those assets to another member of the
group, which increased the parent’s basis in the new
subsidiary’s stock under § 1.1502-32. Id. But because the
transfer of the assets also reduced the value of the new
subsidiary’s stock, the parent could then sell its stock and
claim a loss to offset the gain realized on the assets. Id.
14
Though § 1.1502-35T was not issued in final form
until 2003, it applied retroactively to stock sales starting on
March 7, 2002 (with certain exceptions not relevant here).
See Treas. Reg. § 1.1502-35T(j).

31
2002). The IRS thus had the prospect of issuing another
regulation to bar double deductions when it issued § 1.337(d)-
2T, and doing so was consistent with the consolidated return
practice of imposing multiple restrictions on stock losses (as
in the former § 1.1502-20).

Section 1.1502-35T was the pertinent regulation, and it
was expressly intended “to prevent a group from obtaining
more than one tax benefit from a single economic loss.”
Treas. Reg. § 1.1502-35T(a). It accomplished this through a
loss suspension rule:

Any loss recognized by a member of a
consolidated group with respect to the
disposition of a share of subsidiary member
stock shall be suspended to the extent of the
duplicated loss with respect to such share of
stock if, immediately after the disposition, the
subsidiary is a member of the consolidated
group of which it was a member immediately
prior to the disposition (or any successor
group).

Treas. Reg. § 1.1502-35T(c)(1) (as issued in 2003 retroactive
to March 7, 2002). Yet this regulation failed to prevent a
double deduction here because of how the Duquesne group
structured the relevant transactions in the wake of Rite-Aid.
The loss suspension rule of paragraph 1.1502-35T(c) applies
only when a member of a consolidated group, such as
AquaSource, sells stock in its subsidiaries and those
subsidiaries remain members of the group after the sale. But
because AquaSource sold all of the stock in its subsidiaries to
third parties, they were no longer members of the group after
the 2002-2003 losses. Paragraph 1.1502-35T(c) thus did not
prevent AquaSource from deducting the 2002-2003 losses.
Though Duquesne may nonetheless claim that it complied

32
with § 1.1502-35T in the sense that it did not violate it, we
agree with the Tax Court that this is not enough to meet the
clear authorization test of Ilfeld.

So this case comes down to whether, under the
interpretive principle of Ilfeld, § 1.337(d)-2T clearly
authorizes the losses in 2002-03 for the Duquesne
consolidated group that duplicates the loss it took for 2001.
Its argument is that, even if Ilfeld applies, while paragraph
(a)(1) bars a deduction for any losses by any member of a
consolidated group with respect to the disposition of stock by
a subsidiary (here the sale by AquaSource of the entire
interest in eight of its direct and indirect subsidiaries),
paragraph (c)(2) does not disallow those losses if AquaSource
shows that they do not result from built-in gain on that sale,
and here they do not.

If literally applied to the exclusion of all other
provisions of the Internal Revenue Code, Duquesne (and our
dissenting colleague) make a strong argument. By regulation
there is excluded —that is, disallowed—any deduction for a
loss that occurs when a member of a consolidated group sells
stock in a subsidiary. 26 C.F.R. § 1.337(d)-2T(a)(1). Yet that
deduction disallowance “is not disallowed” under paragraph
(c)(2) when the sale does not result in the recognition of built-
in gain (an assertion we accept here).

So is the case closed? No. That is because the partial
excision in paragraph (c)(2) is one-time only, for there is no
mention in the regulation of approval for a loss deduction that
duplicates another already taken for the underlying economic
loss. We know that it has nothing to do with loss duplication
because, at the same time paragraph (c)(2) was issued
notwithstanding its exception to loss disallowance, the Notice
accompanying it warned that the “IRS and Treasury believe
that a consolidated group should not be able to benefit more

33
than once from one economic loss.” IRS Notice 2002-18,
2002-1 C.B. 644 (Mar. 9, 2002). In this context, the
blinkered approach of Duquesne does not stand so long as
there is Ilfeld. It remains a valid interpretative principle for
consolidated returns until the Supreme Court tells us
otherwise. 15

We therefore confront circumstances fundamentally
unlike those before the Tax Court in Woods. This is not a
case of the IRS failing to act when its own regulations and
related statute specifically authorize the result sought by the
taxpayer. Here, unlike Woods, the IRS has never had a
regulatory scheme in place that would authorize Duquesne to
take both deductions it has claimed for the same economic
loss. And the text of § 1.337(d)-2T does not clearly allow in
the future losses already taken on consolidated returns. Put
another way, future losses cannot be added to past losses
already deducted for the same group of assets. The dissent
would have us apply Woods and hold that the IRS implicitly
authorized Duquesne’s double deductions when they were not
explicitly banned. For consolidated-return taxpayers, implied
authorizations are not enough; they must be clear statements

15
What our dissenting colleague characterizes as our
deference to the IRS’s “position regarding the status or
strength of judicial precedent,” Dissent at 13, is no deference
at all, but our independent assessment that possibly relevant
regulations did not provide the required authorization under
Ilfeld for Duquesne to take a double deduction. Because we
happen to agree with the IRS as to whether Ilfeld’s
requirements were satisfied in this circumstance does not
mean we defer in any way to its reading of judicial precedent.

34
that can fairly be read to allow double deductions for the
same economic loss, and here that did not occur. 16

C. Statute of Limitations

Finally, Duquesne contends that the IRS is time-barred
from ordering repayment of at least some of its tentative
refunds resulting from the carryback of the 2002 and 2003
losses to tax year 2000. Though we agree with the Tax Court
that it is difficult to understand what Duquesne is attempting
to argue on this issue, its briefing on appeal enables us to
discern the basic contours of the argument. Duquesne asserts
that, to the extent losses were carried back from 2002 to 2000

16
Duquesne also makes the argument that even if its
interpretation of the consolidated return regulations proved
incorrect, it is still entitled to claim a double deduction
because its interpretation of those regulations was reasonable.
Though the Tax Court allowed a reasonableness defense in
Gottesman, this was because the IRS sought to impose a
penalty tax that is strictly construed against the IRS. See 77
T.C. at 1156. Indeed, the Court repeated that its reasoning
rested on the penalty tax at issue “for emphasis.” Id.
Duquesne is unable to cite any case in which reasonableness
was allowed as a defense for non-payment of ordinary
corporate taxes such as those before us today (despite its
insistence that Applied Research Assocs., Inc. v.
Commissioner, 143 T.C. 310 (2014), somehow implies this).
Though Duquesne also argues that Gottesman teaches
courts not to fill gaps in the consolidated return regulations,
Gottesman’s language to that effect is also off point. See 77
T.C. at 1158. It concerned not only a penalty tax, but also, as
in Woods, a situation in which the IRS had failed to amend its
regulations despite having years to do so. See id.

35
and it received tentative refunds for tax year 2000 as a result,
26 U.S.C. § 6501(k) places those refunds beyond the statute
of limitations. As Duquesne agreed to extend the statute of
limitations for losses carried back from 2003 to 2000,
however, well-established law provides that the Government
could demand repayment of any refunds issued for tax year
2000. As we see nothing in § 6501(k) that overcomes the
effect of Duquesne’s agreement with the IRS, we reject the
former’s argument that the tentative refunds here are outside
the limitations period.

Section 6501 provides the framework for determining
the statute of limitations. When a tax year is within the
statute of limitations, it is said to be “open;” when the
limitations period expires, it is “closed.” Under § 6501(a), a
tax year generally remains open for three years after the
taxpayer files its return.

But when a corporate taxpayer carries losses back to
an earlier year, determining the statute of limitations is more
complex. A loss carryback uses capital losses incurred in one
year to offset capital gains in an earlier year. See 26 U.S.C.
§§ 1211-12. Though the statute of limitations is typically
measured from the earlier year, § 6501(h) provides that the
limitations period is extended to include the period applicable
to the later year from which losses were carried back. For
example, if losses were carried back from 1992 to 1989, §
6501(h) would extend the limitations period for 1989 until at
least 1995 (that is, three years after 1992). Subsection
6501(k), on which Duquesne relies, provides in essence that
the extended period provided by § 6501(h) also applies when
the taxpayer carries back losses in order to claim a tentative
refund. 17

17
Subsection 6501(k) reads in full:

36
 Regardless whether the taxpayer carries back any
losses at all, it may agree under § 6501(c)(4) to extend the
limitations period for a given tax year. If the taxpayer does
so, the IRS may demand payment of additional taxes for that
year on any ground. See Calumet Industries, Inc. v. Comm’r
of Internal Revenue, 95 T.C. 257, 278 (1990) ( “So long as
the period for assessment is open under some provision in the
year under consideration, we have jurisdiction to consider all
items that may affect that taxable year”). This rule has been
applied repeatedly and specifically to allow the IRS to
disallow losses carried back from a closed year. See id.; see
also Pacific Transport Co. v. Comm’r of Internal Revenue,
483 F.2d 209, 214-15 (9th Cir. 1973); Mecom v. Comm’r of
Internal Revenue, 101 T.C. 374, 392-93 (1993) (collecting
cases), aff’d, 40 F.3d 385 (5th Cir. 1994) (table). As a result,

Tentative carryback adjustment assessment
period.--In a case where an amount has been
applied, credited, or refunded under section
6411 (relating to tentative carryback and refund
adjustments) by reason of a net operating loss
carryback, a capital loss carryback, or a credit
carryback (as defined in section 6511(d)(4)(C))
to a prior taxable year, the period described in
subsection (a) of this section for assessing a
deficiency for such prior taxable year shall be
extended to include the period described in
subsection (h) or (j), whichever is applicable;
except that the amount which may be assessed
solely by reason of this subsection shall not
exceed the amount so applied, credited, or
refunded under section 6411, reduced by any
amount which may be assessed solely by reason
of subsection (h) or (j), as the case may be.
26 U.S.C. § 6501(k).

37
when the taxpayer agrees to extend the statute of limitations
for a particular year, the IRS may disallow within the
extended period any losses carried back to that year and
demand repayment of the resulting refunds.

In this case, Duquesne agreed to extend the statute of
limitations for tax year 2000. After Duquesne carried back
losses from 2003 to 2000 and received a tentative refund, the
statute of limitations for tax year 2000 was extended until at
least 2006 under § 6501(k). It is undisputed that, before that
period lapsed, Duquesne agreed to extend the statute of
limitations and the IRS timely served Duquesne with a notice
of deficiency based on losses carried back to 2000. Though
Duquesne contends that § 6501(k) somehow requires that tax
year 2002 be within the statute of limitations as well, we see
nothing in the subsection’s text to disturb well-settled law on
the effect of a taxpayer’s agreement with the IRS. As 2000
was open by agreement with respect to the 2003 losses and
the 2002 losses were carried back to 2000 as well, we
conclude that the IRS is not time-barred from demanding
repayment of the refunds resulting from any losses carried
back to 2000.

* * * * *

Though it is tempting to dismiss this case as merely
further proof of the “difficult and torturous path” revealed
when “we are constrained to enter the labyrinthine structured
tax laws,” Ambac Industries, Inc. v. Comm’r of Internal
Revenue, 487 F.2d 463, 464 (2d Cir. 1973), it also serves to
remind us that double deductions for consolidated taxpayers
are treated differently from other aspects of tax law. Rite-Aid
created a gap in the regulations preventing them, and thus
Duquesne concluded correctly that a door had opened in the
consolidated-return regime. It took a deduction for losses it
incurred in 2001. That is a ticket for only one ride; Duquesne

38
cannot do so again for the same economic loss. It especially
cannot do so when the IRS told it, when Reg. 1.337(d)-2T
went effective, that a second ride on the same-loss train is
closed to consolidated taxpayers. Yet that is what Duquesne
attempted in the hope that not paying for that ride would go
unchallenged or, if challenged, its position would be upheld
by a court. Duquesne and our dissenting colleague believe
that the authority for a loss given in this temporary regulation
looks forward only, with no accounting for the same loss
already taken. We believe Ilfeld counsels otherwise and thus
affirm the Tax Court’s judgment.

39
HARDIMAN, Circuit Judge, dissenting.

According to the Majority, this case “concerns the
continued vitality of the so-called Ilfeld doctrine for
interpreting the Internal Revenue Code.” Majority Op. 3. Yet
there can be no doubt that Ilfeld retains its vitality as a
precedent of the Supreme Court, and Appellant Duquesne
Light acknowledges as much. As I see it, the true question
presented is whether Ilfeld applies where, as here, a hastily
issued regulation authorizes the very actions that Ilfeld
cautions against. The answer to that question depends on
Ilfeld’s scope, the meaning of the applicable regulations
(specifically § 1.337(d)-2T), and the significance of
Duquesne’s compliance with those regulations. Because I see
those issues differently than my colleagues, I respectfully
dissent.

I. The Scope of Ilfeld

Ilfeld seeks to prevent double deductions. To that end,
the Supreme Court declared that “the practical equivalent of a
double deduction” should not be sanctioned absent “a
provision of the Act definitely requiring it.” Charles Ilfeld
Co. v. Hernandez, 292 U.S. 62, 68 (1934). The Majority
interprets this language to require the taxpayer to identify
three distinct layers of authorization. The first and second
layers would require statutory authorization for each of the
two deductions, while a third layer would need to provide
“explicit approval” for the taxpayer to take both deductions.
See Majority Op. 22. That means even if the Code separately
allows Deduction A and Deduction B, the taxpayer could not
take both deductions unless a provision authorized them to be
taken simultaneously. This triple-authorization requirement, I

1
believe, goes above and beyond any rule envisioned by the
Supreme Court.

As my colleagues acknowledge, disapproval of double
deductions was not Ilfeld’s “case-specific holding.” See
Majority Op. 23. In fact, the Ilfeld Court disallowed the
taxpayer’s claimed deduction not on these overarching policy
grounds, but because the then-applicable regulations
prevented the taxpayer from recognizing the second claimed
loss. 292 U.S. at 67. In that same vein, other courts have
recognized that Ilfeld’s preclusion of double deductions
except where “definitely require[d]” was not essential to the
Court’s disposition. Cf. Marwais Steel Co. v. Comm’r of
Internal Revenue, 354 F.2d 997, 998 (9th Cir. 1965) (“We
follow taxpayer’s argument that part of what was there said
was dicta.”). And when discussing the actual facts and
regulations at issue in the case, the Ilfeld Court seemed to lay
down a less rigorous rule, holding that: “In the absence of a
provision in the Act or regulations that fairly may be read to
authorize it, the deduction claimed is not allowable.” 292
U.S. at 66 (emphasis added).

Thus, two distinct standards (“definitely requires” or
“fairly may be read to authorize”) can be extracted from
Ilfeld. While it is true that the “definitely requires” standard
supports the Majority’s triple-authorization rule, the “fairly
may be read to authorize” standard suggests something
between a watered-down clear statement rule and an
interpretive aid. The Majority acknowledges that choosing
between these standards is both difficult and crucial. See
Majority Op. 19 (explaining that Ilfeld’s standard is
“debatable” and that “‘[d]efinitely requiring’ a provision to
authorize a double deduction . . . is a very high hurdle”). For

2
the reasons I shall explain, we should adopt the “fairly may
be read to authorize” standard.

A.

The disparate standards just mentioned can be
harmonized, despite their apparent inconsistency. See, e.g.,
N.J. Air Nat’l Guard v. Fed. Labor Relations Auth., 677 F.2d
276, 282 (3d Cir. 1982). As a general rule, courts should
preclude double deductions unless “definitely require[d].”
Ilfeld, 292 U.S. at 68. And when is a double deduction
definitely required? When a statute or regulation “fairly may
be read to authorize it.” Id. at 66. The taxpayer lost in Ilfeld
because the Court could point to no authority that “purports
to authorize double deduction of losses.” Id. at 68 (emphasis
added). By requiring merely that the law “purport to
authorize” a double deduction, the Court eschewed a reading
like the Majority’s triple-authorization rule.

B.

Apart from the language of Ilfeld itself, courts
interpreting it have never required the triple authorization that
the Majority articulates today. Just a year after Ilfeld was
decided, our Court explained that the doctrine prohibits
double deductions “except where act and regulation so
provide.” Comm’r of Internal Revenue v. Nat’l Casket Co., 78
F.2d 940, 941 (3d Cir. 1935). Thirty years later, we stated that
despite an obvious double (or even triple) deduction, we had
“no choice . . . other than to apply [the applicable provision]
literally as it is worded.” Miller’s Estate v. Comm’r of
Internal Revenue, 400 F.2d 407, 410 (3d Cir. 1968). We went
on to say that “[t]he line of cases cited by the Commission
descending from [Ilfeld], and allegedly supporting the rule of

3
tax interpretation that double deductions are not permitted
absent express statutory mandate, is merely a variation on the
‘avoid absurd results’ rule.” Id. at 411. We cautioned that
“using broad equitable consideration[s], such as preventing
multiple deductions, to solve problems raised by a tax statute
is a dangerous course.” Id. at 411 n.12.1 Since this weak
embrace, we have not cited Ilfeld.2

1
The Majority says our holding in Miller
“distinguished Ilfeld on the ground that it concerned the
peculiar income tax context of consolidated corporate income
tax reporting.” Majority Op. 20 (internal quotations omitted).
This reading of Miller ignores its primary holding that we
must apply the law “literally as it is worded” and seizes on an
additional reason provided by the court to distinguish Ilfeld.
Miller, 400 F.2d at 411 (“In addition, this tax ‘rule’ has rarely
been applied outside the peculiar income tax context of
consolidated [returns].” (emphasis added)). In fact, the
language quoted by the Majority was not meant to reaffirm
Ilfeld in the consolidated returns context, but to express
skepticism toward the rule generally, as indicated by the
Miller court’s use of scare quotes as a rhetorical device. See
id.
2
Our sister courts have taken a similarly modest view
of Ilfeld’s scope. See Transco Expl. Co. v. Comm’r of Internal
Revenue, 949 F.2d 837, 841 (5th Cir. 1992) (“[Ilfeld] is only
an interpretive principle. It does not require or license us to
rewrite the Code or the Secretary’s regulations. This court
and others have balked in the past at revision of the tax code
to reach what appears to be a more sensible result.”); Transco
v. Comm’r of Internal Revenue, 561 F.2d 1023, 1026 (1st Cir.

4
 The Supreme Court’s recent treatment of factually
similar cases indicates that any such “clear statement” rule
yields in the face of compliance with applicable laws and
regulations. In Gitlitz v. Commissioner of Internal Revenue,
the shareholders of an S corporation received a double tax
benefit by excluding a debt cancellation from their income
under one provision (rendering it nontaxable) and then
including that same amount when calculating their stock basis
through another provision (so that it increased their
deductions for losses). 531 U.S. 206, 209–10 (2001). Because
the shareholders’ deductions complied with sequencing
provisions “expressly addressed in the [applicable] statute,”
the Supreme Court held that the statute imposed no restriction
on the deduction. Id. at 218–19. Without citing to Ilfeld, the
majority closed by addressing Ilfeld’s principal concern:

[C]ourts have discussed the policy concern that,
if shareholders were permitted to pass through
the discharge of indebtedness before reducing
any tax attributes, the shareholders would
wrongly experience a “double windfall”: They
would be exempted from paying taxes on the
full amount of the discharge of indebtedness,
and they would be able to increase basis and
deduct their previously suspended losses.
Because the Code’s plain text permits the

1977). But cf. Marwais Steel Co., 354 F.2d at 998 (applying
Ilfeld even though “the argument of [the taxpayer] is very
difficult to answer [and] seems near perfect in logic[,
because] in human experience, most logic can be carried only
so far”).

5
 taxpayers here to receive these benefits, we
need not address this policy concern.

Id. at 219–20 (citation omitted).

I find it significant that the only citation to Ilfeld in the
Gitlitz case was in Justice Breyer’s lone dissent. He opined
that the statute was ambiguous and should be interpreted in
favor of “closing, not maintaining, tax loopholes.” Id. at 223
(Breyer, J., dissenting). And he lamented the fact that “[the
majority’s] interpretation of the Code results in the ‘practical
equivalent of [a] double deduction.’” Id. at 224 (quoting
Ilfeld, 292 U.S. at 68). Like the Majority here, Justice Breyer
would have precluded the double deduction because he could
find no “clear declaration of intent by Congress” to allow it.
Id.

Thus, the Gitlitz Court broke along the same
theoretical lines as we do here. Justice Breyer’s view of
Ilfeld’s scope would have required an explicit authorization in
the Code allowing for both tax benefits to be reaped
simultaneously. See id. The eight-justice majority, on the
other hand, was satisfied that the combined provisions
“permit[ted] the taxpayers to receive the[] benefits,” without
requiring an additional statement that the two benefits could
be concurrently utilized.3 Id. at 220. It goes without saying

3
The Majority notes that the Court in Gitlitz “did not
purport to overrule [Ilfeld].” Majority Op. 21. This makes
sense, given that the two cases are perfectly consistent. Each
indicates that the judiciary’s policy concern of precluding
double deductions must yield to positive law that provides
otherwise. The Majority finds support in the fact that the
Gitlitz majority “did not so much as mention” Ilfeld. Id.

6
that we must adhere to the opinion of the Court. And Gitlitz’s
“permit” standard is quite closer to my “fairly may be read to
authorize” standard than it is to the Majority’s triple-
authorization rule.

C.

Next, the Majority’s triple-authorization test should be
rejected because reasonable minds can differ as to when and
how it would be satisfied. The decision of the Tax Court in
Woods Investment Co. v. Commissioner of Internal Revenue,
85 T.C. 274 (1985), is instructive in this regard. The Majority
cites Woods as a case that clears Ilfeld’s “high hurdle.” See
Majority Op. 19, 25–26. I cannot agree. Woods involved the
interplay of a regulation and statute that, when applied
simultaneously, could result in a double deduction. As the
Majority explained:

[Section] 1.1502-32 required the parent to
adjust its basis in line with the subsidiary’s
earnings and profits, and 26 U.S.C. § 312(k)
required the parent to calculate those earnings
and profits using a straight-line method of
depreciation. Despite its awareness of the
potential for straight-line depreciation to result
in the practical equivalent of a double
deduction, the IRS had failed to amend §
1.1502-32 for almost twenty years.

However (as noted above), the Gitlitz majority felt that it
“need not address [the] policy concern” of duplicative
benefits (i.e., the Ilfeld doctrine) and the citation to Ilfeld by
Justice Breyer in dissent emphasizes Ilfeld’s applicability to
the case. 531 U.S. at 219–20.

7
Majority Op. 26. It is difficult to understand how Woods—
where the IRS was aware of the potential of a loophole but
failed to amend the scheme—is one where an on-point statute
and an on-point regulation provide “explicit approval for
duplicating the underlying economic loss,” as the Majority
claims is necessary under Ilfeld. Majority Op. 22. If Woods
passes muster, the Majority’s test is meaningless. Presumably
for that reason, the Woods court did not make such a claim.
Rather, it primarily relied on Ilfeld’s other standard—the
“fairly may be read to authorize” standard that I advocate
here—and wrote that the taxpayer should prevail because the
regulations and statute “can fairly be read to authorize the
result herein.” 85 T.C. at 283.

One final point of policy is worth mentioning: If we
were to adopt the “fairly may be read to authorize” standard,
we would not be opening the double-deduction floodgates.
For example, I agree with the Majority’s excellent analysis
that § 165 does not authorize a double deduction. See
Majority Op. 22–23 (noting, among other things, that the
language of § 165 “allows a single deduction for a single
loss” and that the Court in Ilfeld rejected a similar provision).
In fact, the Majority’s careful parsing of § 165 is just what the
“fairly may be read to authorize” standard encourages—an
examination of the text, history, and scheme to see whether
the taxpayer’s interpretation is a fair one.4

4
The preceding section hopefully makes clear that,
contrary to the Majority’s claim, I do not interpret Ilfeld to
permit double deductions merely “when they [are] not
explicitly banned.” Majority Op. 34. A double deduction is
not authorized by silence; rather it is allowed when a statute

8
 II. The On-Point Regulations

The Majority does not dispute that Duquesne’s returns
complied with the applicable regulations: § 1.1502-32 (basis
calculation) and § 1.337(d)-2T (loss allowance). Majority Op.
24. The meaning of those regulations is a point of contention,
however.

A. Section 1.337(d)-2T

Section 1.337(d)-2T says in paragraph (a)(1): “No
deduction is allowed for any loss recognized by a member of
a consolidated group with respect to the disposition of stock
of a subsidiary.” Paragraph (c)(2) qualifies this rule by stating
that the “[l]oss is not disallowed under paragraph (a)(1) of
this section . . . to the extent the taxpayer establishes that the
loss or basis is not attributable to the recognition of built-in
gain.”5

Section 1.337(d)-2T’s quadruple-negative construction
leaves much to be desired, but its clumsy syntax doesn’t
absolve us from enforcing it as written. Specifically, the
provision says that losses such as the one Duquesne took here
are “not disallowed.” While the Majority claims this language

or regulation “fairly may be read to authorize” it. See Ilfeld,
292 U.S. at 66.
5
In my view, the Majority’s discussion of § 1.337(d)-
2T was off track from the outset because its triple-
authorization standard obliges the regulation to bear much
more weight than the caselaw demands for the reasons I
described in Part I. What is actually demanded of § 1.337(d)-
2T is that it can reasonably be interpreted (i.e., read fairly) to
mean what the taxpayer claims it means.

9
does not positively authorize any deductions, I read it
differently. My disagreement with the Majority on this score
is simple: “not disallowed” can be read fairly to mean
“allowed.” It has long been a convention of the English
Language—as with arithmetic and logic—that two negatives
make a positive. See, e.g., Robert Lowth, A Short
Introduction to English Grammar: With Critical Notes 162
(1762) (“Two Negatives in English destroy one another, or
are equivalent to an Affirmative.”); Henry Watson Fowler, A
Dictionary of Modern English Usage 374 (1926) (“You may
treat a double negative expression as though it were formally
as well as virtually a positive one.”). And the convention
remains prevalent today, both in formal writing and in the
vernacular. See, e.g., Martha Rose Shulman, Focaccia: One
Basic Bread, Endless Delicious Options, N.Y. Times, May
13, 2013 (“Focaccia is a flatbread, not unlike a very thick-
crusted pizza”); Tad Friend, Blowback, New Yorker, Oct. 25,
2010 (“He grinned, seeming not displeased to be
complicating the issue.”); Tom Jones, It’s Not Unusual, on
Along Came Jones (Decca Records 1965) (explaining that
“it’s not unusual” to do various everyday activities, such as
“to go out at any time”).

If one accepts that “not disallowed” can reasonably be
interpreted to mean “allowed,” then § 1.337(d)-2T—having
been reduced from a quadruple negative to a double
negative—becomes clearer. The combination of paragraphs
(a)(1) and (c)(2) would read something like this: “No
deduction is allowed for any loss recognized with respect to
the disposition of stock of a subsidiary; however, that
deduction is allowed to the extent the taxpayer establishes
that the loss or basis is not attributable to the recognition of
built-in gain.” Put more simply, a subsidiary’s stock loss is

10
allowed unless it relates to a built-in gain. It is undisputed that
Duquesne would meet this standard.

The Majority acknowledges that the foregoing
interpretation results when § 1.337(d)-2T is “literally
applied,” but claims that “[w]e know” this literal
interpretation cannot be correct because it conflicts with an
IRS notice issued alongside § 1.337(d)-2T. Majority Op. 33.
Specifically, the Majority gives great weight to the IRS’s
statement that it “believe[s] that a consolidated group should
not be able to benefit more than once from one economic
loss.” Id. (quoting IRS Notice 2002-18, 2002-1 C.B. 644
(Mar. 9, 2002)).

I see three principal issues with the Majority’s reliance
on this notice. First, the full context of the language quoted by
the Majority shows that the IRS did not mean to imply that
§ 1.337(d)-2T—or any part of the then-in-effect regulatory
scheme—prevented the deductions taken by Duquesne.
Rather, the IRS’s belief that “a consolidated group should not
be able to benefit more than once from one economic loss”
was the impetus for its forthcoming regulations, namely,
1.1502-35T:

These rules [§§ 1.337(d) and 1.1502] do not
disallow stock loss that reflects net operating
losses or built-in asset losses of a subsidiary
member. Nonetheless, the IRS and Treasury
believe that a consolidated group should not be
able to benefit more than once from one
economic loss. Accordingly, the IRS and
Treasury intend to issue regulations that will
prevent a consolidated group from obtaining a
tax benefit from both the utilization of a loss

11
 from the disposition of stock (or another asset
that reflects the basis of stock) and the
utilization of a loss or deduction with respect to
another asset that reflects the same economic
loss.

I.R.S. Notice 2002-18 (March 9, 2002). Thus, the IRS
identified a defect in its regulations and explained that it
would take corrective action to cure that defect. If anything,
this shows that the IRS—at the time Duquesne took its
deductions—did not believe there was a regulatory
mechanism in place to prevent a double deduction.

Second, even if the IRS notice meant what the
Majority claims it does, I would not afford it such deference.
While the IRS may “believe that a consolidated group” should
never receive a double deduction, id. (emphasis added), the
IRS must do more than believe something for it to become
law. See generally 5 U.S.C. §§ 551–559 (The Administrative
Procedure Act). No matter what the IRS says in its informal
notice publications, it cannot alter the plain meaning of its
regulations. To allow otherwise would frustrate the scheme of
the Administrative Procedure Act, essentially permitting the
IRS “to create de facto a new regulation” through the back
door. Christensen v. Harris Cty., 529 U.S. 576, 588, (2000).
“Making regulatory programs effective is the purpose of
rulemaking, in which the agency uses its ‘special expertise’ to
formulate the best rule. But the purpose of interpretation is to
determine the fair meaning of the rule. . . . Not to make
policy, but to determine what policy has been made and
promulgated by the agency[.]” Decker v. Nw. Envtl. Def. Ctr.,
133 S. Ct. 1326, 1340 (2013) (Scalia, J., concurring in part
and dissenting in part).

12
 And third, the Majority’s reliance on this IRS notice
goes over and above any canon of judicial deference to
agency interpretation. In this respect, it’s important to ask:
what authority was the IRS interpreting to support its view
that all double deductions should be disallowed? The IRS
does not claim to be “implementing a statute,” so we know
that it is not entitled to Chevron or Skidmore deference. See
United States v. Mead, 533 U.S. 218, 227–28 (2001). And
since the IRS concedes that §§ 1.337(d) and 1.1502 did not
disallow these deductions, its belief could not have been
derived from those regulations—and therefore is not entitled
to deference under Auer v. Robbins. 519 U.S. 452, 461 (1997)
(holding that an agency’s interpretation of its “own
regulations” is entitled to deference). Rather, the Majority
seems to claim that the IRS notice was a gloss on Ilfeld itself,
writing that the notice shows that Duquesne’s approach fails
“so long as there is Ilfeld.” Majority Op. 33. I am aware of no
caselaw that demands (or permits) a court to give such
deference to an agency’s position regarding the status or
strength of judicial precedent. Nor do I see any principled
reason to do so. After all, “[i]t is emphatically the province
and duty of the judicial department to say what the law is.”
Marbury v. Madison, 5 U.S. (1 Cranch) 137, 177 (1803).

B.

Linguistics aside, the Majority’s reading of § 1.337(d)-
2T is inconsistent with the IRS’s comprehensive regulatory
scheme. The Majority argues that “the blinkered approach of
Duquesne,” i.e., that it was entitled to follow the regulations
that were before it, “does not stand so long as there is Ilfeld.”
Majority Op. 33. But it seems unnatural for the IRS to write a
regulation that literally authorizes a specific action, only to
expect taxpayers to appreciate that the regulation is

13
undermined by common-law doctrines lurking in the
shadows. And this interpretation would seem to vitiate the
need for much of the IRS’s prior and subsequent regulations.
Why have § 1.337(d)-2T or § 1.1502-32 in place if Ilfeld is
the panacea for all consolidated-return, double-deduction ills?

Indeed, if Ilfeld carried such weight, many IRS
regulations promulgated during the past several decades
would be rendered superfluous. For example, after Woods, the
IRS overhauled the consolidated return rules that had
permitted the taxpayer to prevail in that case. T.D. 8560,
1994-2 C.B. 200 (Aug. 15, 1994). Surely this reform was not
just busywork. And after Rite Aid, the IRS scrambled to
implement temporary regulations and warned taxpayers that it
would soon adopt permanent ones to “prevent duplication of
losses within a consolidated group on dispositions of member
stock.” IRS Notice 2002-18, 2002-1 C.B. 644 (Mar. 9, 2002).
It seems unlikely that the IRS would warn taxpayers that it
would soon pass regulations to ban that which was already
banned.

Finally, the Majority counters that the temporary
regulations created a “gap” that Duquesne exploited. Majority
Op. 38. But the regulations were not so porous. Section
1.337(d)-2T explicitly addresses the scenario that occurred
here—losses by a consolidated group incurred on “disposition
of stock of a subsidiary”—and then describes which losses
are disallowed (those based on built-in gains) and which
losses are not disallowed (those for which the “taxpayer
establishes that the loss or basis is not attributable to the
recognition of built-in gain”). This is a “gap” only in some
philosophical sense. Because of the interstitial nature of
regulatory schemes, not every factual permutation will be
specified in the law. But a “gap” doesn’t exist whenever an

14
on-point regulation inadvertently authorizes a result that the
agency later comes to regret.

C. Section 1.1502-35T

Section 1.1502-35T doesn’t affect the meaning of
§ 1.337(d)-2T. The Majority argues that § 1.1502-35T was
the regulation actually intended to prevent Duquesne’s double
deductions but “failed to prevent a double deduction here
because of how the Duquesne group structured the relevant
transactions in the wake of Rite Aid.” Majority Op. 32. But it
doesn’t matter that § 1.1502-35T would—under a different
set of facts—disallow deductions like the ones Duquesne took
here. After all, neither the Tax Court nor the Government
claims that § 1.1502-35T precluded the deductions actually
taken by Duquesne in the relevant time period. And the
Majority concedes that Duquesne “complied with § 1.1502-
35T in the sense that it did not violate it.” Majority Op. 32. I
would require no more.

In this regard, the Majority seems to criticize
Duquesne for acting during the Rite Aid interregnum. But it
matters not whether “[t]he only purpose of [Duquesne] here
was to escape taxation. . . . The fact that [Duquesne] desired
to evade the law, as it is called, is immaterial, because the
very meaning of a line in the law is that you intentionally may
go as close to it as you can if you do not pass it.” Superior Oil
Co. v. Mississippi ex rel. Knox, 280 U.S. 390, 395–96 (1930).
Duquesne was entitled to order its “affairs that [its] taxes
shall be as low as possible;” and was “not bound to choose
that pattern which will best pay the Treasury; there is not
even a patriotic duty to increase one’s taxes.” Helvering v.
Gregory, 69 F.2d 809, 810 (2d Cir. 1934) (Hand, J.) (citations
omitted). For that reason, I disagree with my colleagues’

15
implication that Duquesne should have accommodated the
IRS by executing its transaction at a time when the regulatory
scheme was more favorable to the Government.

III. Duquesne’s Compliance

The foregoing discussion establishes (hopefully) that:
(1) Ilfeld yields to a law that can fairly be read to authorize a
double deduction; and (2) § 1.337(d)-2T can fairly be read to
authorize a double deduction. These propositions lead
ineluctably to the conclusion that, by complying with
§ 1.337(d)-2T, Duquesne’s deductions did not run afoul of
Ilfeld.

A.

The Majority’s analysis is not so straightforward,
looking as it does to the “structure and purpose of the broader
regulatory regime,” Majority Op. 29, instead of the language
of § 1.337(d)-2T. In that regard, I have no quarrel with the
Majority’s convincing argument that Duquesne’s reading of
§ 1.337(d)-2T runs contrary to the provision’s context and
broad purposes. “Let us not forget, however, why context
matters: It is a tool for understanding the terms of the law, not
an excuse for rewriting them.” King v. Burwell, 135 S. Ct.
2480, 2497 (2014) (Scalia, J., dissenting). And “even the
most formidable argument concerning [a law’s] purposes
could not overcome the clarity [found] in the [law’s] text.”
Kloeckner v. Solis, 133 S. Ct. 596, 607 n.4 (2012); see also
Hanover Bank v. Comm’r of Internal Revenue, 369 U.S. 672,
687 (1962) (“[W]e are not at liberty, notwithstanding the
apparent tax-saving windfall bestowed upon taxpayers, to add
to or alter the words employed to effect a purpose which does
not appear on the face of the statute.”).

16
 Section 1.337(d)-2T may not be a model of clarity, but
its tortured prose does not necessarily beget ambiguity. When
the IRS writes that a deduction is “not disallowed,” we should
accept that it is not. And without that ambiguity, it is not our
place to investigate the structure and purpose of the scheme in
order to restyle the language of the regulation. See Cent. Tr.
Co., Rochester, N.Y. v. Official Creditors’ Comm. of Geiger
Enters., 454 U.S. 354, 359 (1982) (“It is elementary that the
meaning of a [law] must, in the first instance, be sought in the
language in which the [law] is framed, and if that is plain, . . .
the sole function of the courts is to enforce it according to its
terms.” (quoting Caminetti v. United States, 242 U.S. 470,
485 (1917))).

Nor does the seemingly unfair result mean that we
should ignore § 1.337(d)-2T’s text. Duquesne’s duplicative
deductions “would show only that the statutory scheme
contains a flaw; they would not show that [§ 1.337(d)-2T]
means the opposite of what it says.” King, 135 S. Ct. at 2503
(Scalia, J., dissenting). And we are not at liberty to revise a
regulation “just because the text as written creates an apparent
anomaly.” Michigan v. Bay Mills Indian Cmty., 134 S. Ct.
2024, 2033 (2014).

Perhaps the Majority’s forecast of the IRS’s intent is
correct and § 1.337(d)-2T’s plain meaning is inadvertent.
That’s beside the point. If the IRS “enacted into law
something different from what it intended, then it should
amend the [law] to conform it to its intent. It is beyond our
province to rescue [lawmakers] from [their] drafting errors,
and to provide for what we might think . . . is the preferred
result.” Lamie v. U.S. Tr., 540 U.S. 526, 542 (2004) (internal
quotations and citation omitted).

17
 B.

Until today Ilfeld had never displaced both the Code
and the relevant regulations that allowed the double
deduction at issue. In fact, Ilfeld itself suggests that it would
give way to permissive regulations. See 292 U.S. at 68
(“There is nothing in the Act that purports to authorize double
deduction of losses or in the regulations to suggest that the
[IRS] construed any of its provisions to empower [the IRS] to
prescribe a regulation that would permit consolidated returns
to be made on the basis now claimed by [the parent].”
(emphases added)). The deductions in Ilfeld violated the
existing regulations, which was the basis of the decision. Id.
at 67.

Nor do any other Supreme Court cases make such a
leap. See McLaughlin v. Pac. Lumber Co., 293 U.S. 351,
355–56 (1934) (holding that the claimed deductions were
inconsistent with the governing statute, where no on-point
regulations permitted the duplicative deductions); United
States v. Skelly Oil Co., 394 U.S. 678, 683–84 (1969) (ruling
against the taxpayer who did not refer to an on-point
regulation—just to a bare, general Code provision that the
Court interpreted, under Ilfeld, not to allow double-dipping).
The same goes for our decisions applying Ilfeld in National
Casket and Greif Cooperage: in neither case did an on-point
regulation actually permit the double deduction challenged by
the IRS. See Nat’l Casket, 78 F.2d at 942 (offering the
qualification: “except where act and regulation so provide,
double deduction of the same losses . . . is not permissible”
(emphasis added)); Greif Cooperage Corp. v. Comm’r of
Internal Revenue, 85 F.2d 365, 365 (3d Cir. 1936).

18
 To the contrary, courts have set Ilfeld aside where, as
here, two separate provisions combine to (perhaps
accidentally) authorize the taxpayer’s double deduction. See
Gitlitz, 531 U.S. at 219–20 (explaining that a taxpayer may
receive a “double windfall” because the “Code’s plain text”
contained separate provisions that “permitted” taxpayers to
“be exempted from paying taxes on the full amount of the
discharge of indebtedness and [to] be able to increase basis
and deduct their previously suspended losses”); Woods, 85
T.C. at 282 (“If respondent believes that his regulations and
section 312(k) together cause petitioner to receive a ‘double
deduction,’ then respondent should use his broad power to
amend his regulations.”).
* * *

Duquesne concedes that it took a double deduction.
Although that action might be criticized on policy grounds, it
complied with the laws and regulations applicable at the time.
For that reason, I respectfully dissent.

19