REINHARDT, Circuit Judge:
On December 30, 1993, we filed an opinion concerning various disputes between Albertson’s and the Internal Revenue Service. 38 F.3d 1046 (9th Cir.1993). We granted the government’s petition for rehearing as to Part II.B of the opinion, which concerned the appropriate tax treatment of deferred compensation agreements. Today we vacate Part II.B of the original opinion and affirm the Tax Court’s decision.
I.
BACKGROUND
Deferred compensation agreements (“DCAs”) are agreements in which certain employees and independent contractors (“DCA participants”) agree to wait a specified period of time (“deferral period”) before receiving the annual bonuses, salaries, or director’s fees that they would otherwise receive on a current basis. During the deferral period, the employer uses the basic amounts of deferred compensation (“basic amounts”), which accumulate on an annual basis, as a source of working capital. At the end of the deferral period, the employer pays the participating individuals the basic amounts
and
an additional amount for the time value of the deferred payments that have accumulated on the basic amounts (“additional amount”). The time-value-of-money sums are also computed on a yearly basis. The total of these basic amounts and the amounts attributable to compensation for the delay in
payment of those amounts constitutes the whole of the deferred compensation (“deferred compensation”). The time-value-of-money component may be measured by interest rate indices, equity fund indices, or cost of living increases, or it .may simply be included within a lump-sum payment.
Prior to 1982, Albertson’s entered into DCAs with eight of its.top executives and one outside director. The parties agreed that their deferred compensation would include the annual basic amounts plus additional amounts calculated annually in accordance with an established formula.
Albertson’s, Inc. v. Commissioner,
38 F.3d 1046, 1051 (9th Cir.1993).
The DCA participants would be eligible to receive the deferred compensation (the total sum) upon their retirement or termination of employment with Albertson’s. The DCA participants also had the option of further deferring payment for up to fifteen years thereafter. During that extra period, the additional amounts would continue to accrue on an annual basis.
Id.
at 1052.
In 1982, Albertson’s requested permission from the IRS to deduct the additional amounts (but not the basic amounts) during the year in which they accrued instead of waiting until the end of the deferral period.
Id.
In 1983', the IRS granted Albertson’s request. Accordingly, Albertson’s claimed deductions of $667,142 for the additional amounts that had already accrued, even though it had not yet paid the DCA participants any sums under the deferred compensation agreements.
Id.
In 1987, the IRS changed its policy, however, and sought a deficiency for the additional amounts, contending that all amounts provided for in the deferred compensation agreements were deductible only when received by Albertson’s employees. Albertson’s filed a petition with the Tax Court, claiming that the additional amounts constituted “interest” and thus could be deducted as they accrued.
Id.
In a sharply divided opinion,
the Tax Court rejected Albertson’s position.
Albertson’s, Inc. v. Commissioner,
95 T.C. 415, 1990 WL 149185 (1990). The court found that the additional amounts represented compensation, not interest, and were therefore not deductible until the end of the deferral period under I.R.C. § 404(a)(5) &’(d).
We reversed the decision of the Tax Court.
Albertson’s, Inc. v. Commissioner,
38 F.3d 1046 (9th Cir.1993). We held that the additional amounts constituted interest within the definition of I.R.C. § 163(a) and that interest payments were not governed by the timing ^restrictions of section 404. The government petitioned for rehearing due to the significant fiscal impact of the panel’s opinion which it estimates will cause a $7 billion loss in tax revenues.
II.
REHEARING
We agreed to rehear this issue after lengthy consideration and reflection. In our
original opinion, we stated that the plain language of the statute strongly supported Albertson’s interpretation and, accordingly, we adopted it. Nevertheless, we expressed sympathy for the Commissioner’s argument that Congress intended the timing restrictions of I.R.C. § 404 to apply to
all
payments made under a deferred compensation plan and recognized that our plain language interpretation seemed to undercut Congress’ purpose.
We have now changed our minds about the result we reached in our original opinion and conclude that our initial decision was incorrect. The question is not an easy one, however. We have struggled with it unsuccessfully at least once, and it may, indeed, ultimately turn out that the United States Supreme Court will tell us that it is this opinion which is in error. This is simply one of those cases — and there are more of them than judges generally like to admit — in which the answer is far from clear and in which there are conflicting rules and principles that we are forced to try to apply simultaneously. Such accommodation sometimes proves to be impossible. In some cases, as here, convincing arguments can be made for both possible results, and the court’s decision will depend on which of the two competing legal principles it chooses to give greater weight to in the particular circumstance. Law, even statutory construction, is not a science. It is merely an effort by human beings, albeit judges, to do their best with imperfect tools to arrive at a correct result.
There is a question whether, having once decided a case, we should change our decision when we are not entirely certain that the result we reached is wrong. One response is that, if the issue could be resolved with that degree of certainty, it is unlikely that we would have decided the case incorrectly the first time. Moreover, if certainty were the standard, we would probably never reverse ourselves. There is actually no clear set of rules that tells us when a ease warrants our changing our decision on rehearing. We start with the premise that doing so is not generally desirable, and that it runs contrary to the sense of stability and finality that the law seeks to foster. We also know that it is often better to have a definitive answer, whatever it is, than to have continuing reexaminations or self-questioning.
On the other hand, we judges do not just bury our mistakes. We display them publicly in the Federal Reporters and, while we may then as individuals move on to more decisionmaking, the opinions we have published continue to haunt indefinitely not just the parties, but often numerous other persons whose affairs and fortunes will be governed by them.
Because all of us make hundreds of difficult decisions a year involving complex legal questions, we know that we will make a certain number of errors. All that we can do is to try our best to hold them to a minimum. At the same time, if a rehearing is requested and'we have a strong sense that we may have erred in the particular case, we should not hesitate to undertake a reexamination of the issue. This is particularly so when significant individual rights or interests are at stake or when a number of parties may be seriously affected by a decision that may be erroneous. Given all of this, our conclusion is that, while we should not ordinarily abandon the decisions we have just reached following full deliberation, we must be willing to take that unusual step — at least in cases of some significance — when ultimately we are fairly persuaded that our decision is in error. This is such a case.
In its petition for rehearing, the government, far more forcefully and clearly than it did originally, has articulated the purpose of the timing restrictions outlined in I.R.C. § 404: to encourage employers to invest in qualified compensation plans by requiring inclusions and deductions of income and expense to be “matched” for nonqualified plans.
See infra
p. 543. The matching principle, widely recognized to be the key to I.R.C. § 404, provides significant tax incentives for employers to invest in
qualified
deferred compensation plans, which are nondiscriminatory and ensure that employees receive the compensation promised to them.
See 2
Boris I. Bittker & Lawrence Lokken,
Federal Taxation of Income, Estates & Gifts
¶ 60.1'(2d ed. 1990) (noting that the “matching princi-
pie” is “the most consistent feature of the rules for nonqualified plans.”). As the Commissioner forcefully argues, our original interpretation of I.R.C. § 404 undercut the essential purpose of that provision by violating the matching principle and creating a taxation scheme that favors the type of plan that Congress intended to discourage.
See infra
pp. 54SM5. For this reason, we granted the Commissioner’s petition for rehearing. We now withdraw the portion of our earlier opinion that dealt with deferred compensation agreements, published at 38 F.3d at 1051-1056, and affirm the Tax Court’s decision, although not for the reasons upon which the Tax Court majority relied.
III.
ANALYSIS
Albertson’s again urges this court (1) to characterize the additional amounts as interest as defined by I.R.C. § 163(a),
and (2) to find that such “interest” payments are deductible under I.R.C. § 404.
However, we have now concluded that, notwithstanding the statutory language on which Albertson’s relies, to hold the additional amounts to be deductible would contravene the clear purpose of the taxation scheme Congress created to govern deferred compensation plans. As the Supreme Court noted in
Bob Jones University v. United States,
461 U.S. 574, 103 S.Ct. 2017, 76 L.Ed.2d 157 (1983), a term in the Code
“must
be analyzed and construed within the framework of the Internal Revenue Code and against the background of the congressional purposes.”
Id.
at 586, 103 S.Ct. at 2026 (emphasis added).
A.
A Comparison of Qualified and Non-qualified Plans
An examination of the differences between qualified and nonqualified plans is essential to an understanding of the purpose of the congressional scheme governing deferred compensation agreements. Congress has imposed few restrictions upon nonqualified deferred compensation plans. An employer may limit participation in a nonqualified plan to highly paid executives, and it need not guarantee equal benefits for all participants. In addition, the employer is not required to set aside any funds or provide any guarantees (beyond the initial contractual promise) that its employees will receive the compensation. Thus, promised benefits for unfunded, nonqualified plans are subject to the claims of the employer’s general creditors.
Under a qualified plan, in contrast, an employer may not discriminate in favor of officers, shareholders, or highly compensated employees. I.R.C. § 401(a)(4) & (a)(5). In addition, a qualified plan must satisfy minimum participation and coverage standards concerning eligibility and actual rates of participation. I.R.C. §§ 401(a)(2) & (a)(26), 410. The amounts which an employer may contribute to qualified plans and the benefits which qualified plans may provide are also restricted. I.R.C. §§ 401(a)(17), 415.
A qualified plan also provides significant guarantees that employees will receive the compensation promised to them. It generally must be funded through a trust. I.R.C. § 401(a). Neither the corpus nor the income of the trust may be diverted for any purpose; they can only be used for the exclusive benefit of the participants. I.R.C. § 401(a)(2). Under certain qualified plans, the employer’s contributions must meet strict funding requirements, and minimum standards govern the vesting of participants’ benefits. I.R.C. §§ 401(a)(1) & (a)(7), 411, 412;
see also
29 U.S.C. § 1082.
It is clear that few employers would adopt a qualified deferred compensation plan, with all of its burdensome requirements, if the taxation scheme favored nonqualified plans or treated nonqualified and qualified plans similarly. Although qualified plans provide significant benefits to employees, they allow employers little flexibility in structuring a plan, require them to provide extensive coverage, prevent them from discriminating in favor of highly compensated employees, and involve a significant initial outlay of funds. Thus, the extensive regulations Congress has imposed upon qualified plans would serve little purpose unless employers had an incentive to adopt such plans. As we discuss in the next part, section 404 provides the incentive necessary to encourage employers to adopt qualified plans by providing significantly more favorable tax treatment of qualified plans than of nonqualified ones.
The most significant difference between the two types of plans, for purposes of tax deductibility, is that under a qualified plan the employer must turn over annually to a third party the basic amounts that are deferred and may not use those amounts for' the employer’s own benefit. Thus, the employer, in effect, is required to make the deferred payments at the time the employee is earning the compensation. It is only the employee’s right to receive the funds that is delayed. In contrast, an employer with a nonqualified plan is not required to turn any funds over to anyone until the end of the deferred compensation period. Such an employer may use those funds for its own purposes for a period of many years. In a nonqualified plan, it is not only the. employee’s right to receive the funds that is deferred; the employer’s obligation to part with the funds is deferred as well. If one could simply retain the funds and receive tax benefits similar to those one would receive if those amounts were paid out, there would clearly be little incentive to establish a qualified plan.
B.
The Purpose of Section b-Ob
Congress enacted section 23(p), the forerunner to section 404, in 1942. Prior to 1942, corporations were allowed to deduct DCA-related expenses as they accrued each year, even though employees did not recognize any income until a subsequent taxable year. In 1942, Congress eliminated this favorable treatment for deductions relating to “non-qualified” deferred compensation agreements, such as the DCAs at issue in this case.
In so doing, Congress forced employers who chose to retain their funds for their own use to wait until the
end
of the deferral period, when these amounts were includible in plan participants’ taxable income, before they could take deductions for deferred compensation payments. However, employers who maintained a “qualified” plan that met the rigorous requirements of the Internal Revenue Code (and now ERISA), including turning over the sums involved to a trust fund (or purchasing an annuity), were allowed to continue to take the annual deductions even though their employees would not receive the deferred compensation until a
later year.
See, e.g.,
I.R.C. §§ 404(a) & (d); 29 U.S.C. § 1082 (1988).
1.
The Matching Principle
Congress provided a single explanation for the timing restrictions of section 404: to ensure matching of income inclusion and deduction between employee and employer under nonqualified plans. As both the House and Senate Reports note, “[i]f an employer on the accrual basis defers paying any compensation to the employee until a later year or years ... he will not be allowed a deduction until the year in which the compensation is paid.” H.R.Rep. No. 2333, 77th Cong., 2d Sess. (1942), 1942-2 Cum.Bull. 372, 452; S.Rep. No. 1631, 77th Cong., 2d Sess. (1942), 1942-2 Cum.Bull. 504, 609.
Commentators have widely agreed that this “matching principle” is the key to section 404. As Boris Bittker and Lawrence Lokken have observed, “[t]he most consistent feature for the rules for nonqualified plans is that the employer is ordinarily allowed no deduction for contribution, payments or benefits until they are taxed to the employee.” 2 Boris I. Bittker & Lawrence Lokken,
Federal Taxation of Income, Estates & Gifts
¶ 60.1 (2d ed. 1990). Similarly, Daniel Halperin has noted that, in the case of deferred payment of compensation under nonqualified plans, Congress has imposed “a matching requirement, which denies an employer’s deduction until the deferred amount is included in the employee’s income.” Daniel I. Halperin,
Interest in Disguise: Taxing the “Time Value of Money”,
95 Yale L.J. 506, 520 (1986) (discussing section 404). David Davenport also cites section 404 as the primary proof that “[c]urrent law follows a matching principle and defers the employer’s deduction until the year of payment.” David S. Davenport,
Education & Human Capital: Pursing an Ideal Income Tax and a Sensible Tax Policy,
42 Case W.Res.L.Rev. 793, 865 (1992);
see also Merten’s Law of Federal Income Taxation,
Comm. § 404 (1990 & 1994 Supp.); Joseph L. Cummings, Jr.,
The Silent Policies of Conservation and Cloning of Tax Basis and Their Corporate Application,
48 Tax L.Rev. 113,162 n. 245 (1992) (noting that the matching principle governs section 404); Noel B. Cunningham,
A Theoretical Analysis of the Tax Treatment of Future Costs,
40 Tax L.Rev. 577, 610 (1985) (same) Mark P. Ger-gen,
Reforming Subchapter K: Compensating Service Partnerships,
48 Tax L.Rev. 69, 97 n. 91 (1992) (same).
2.
The Significance of the Matching Principle
The significance of section 404’s matching principle becomes evident when one compares the treatment of qualified and nonqualified plans under that section. Because section 404 requires employer deductions for contributions to nonqualified plans to be “matched,” an employer cannot take tax deductions for payments to its employees until the DCA participants include those payments in their taxable income — that is, until the employees actually receive the compensation promised to them.
Qualified plans, in contrast, are
not
governed by the matching principle and consequently generate concurrent tax benefits to employers. Although employees are not taxed upon the benefits they receive from the plan until they actually receive them, an employer’s contributions to a qualified plan are deductible when paid to the trust. I.R.C. §§ 402(a)(1) & 404(a). Thus, the employer may take an
immediate, unmatched
deduction for any contribution it makes to a qualified plan.
By exempting contributions to qualified plans from the matching principle, Congress compensates employers for meeting the burdensome requirements associated with qualified plans by granting them favorable tax treatment. The current taxation scheme thus creates financial incentives for employers to contribute to qualified plans while providing no comparable benefits for employers who adopt plans that are unfunded or that discriminate in favor of highly compensated employees.
C.
The Effects of Albertson’s Proposal
Albertson’s maintains that section 404 only requires that the basic amounts of compensation be matched; it argues that all additional amounts paid to compensate an employee for the time value of money represent “interest” payments for which an employer may take an immediate deduction. In light of the clear purpose underlying section 404 — to encourage employers to create qualified plans for their employees — we decline to ascribe such an intention to Congress.
First, Albertson’s proposal appears to undermine the effectiveness of the timing restrictions by reducing the significance of the incentive structure created by section 404. In order to adopt Albertson’s proposal and allow employers to take current deductions for additional “interest” payments, we would be required to conclude that Congress created a system in which employers could deduct a substantial portion of the nonqualified deferred compensation package long before its employees had received any of those funds. For example, when the additional amounts are calculated for a compensation package deferred over a fifteen-year period using an interest rate similar to that used by Albert-son’s, an employer can classify more than seventy percent of the deferred compensation package as “interest payments.”
If the additional amounts were calculated at an eight percent interest rate, compounded annually, almost fifty percent of the compensation package could be characterized as “interest” under Albertson’s approach. Even under a deferred compensation package with an interest rate one-third as high as Albert-son’s, one third of the amount paid to the employee at the end of a fifteen-year period would consist of “interest.” Moreover, we note that, under Albertson’s deferred compensation agreement, participants have the option of deferring payment of the total sum available to them upon retirement for an additional period of up to fifteen years.
All
payments during that additional period would also constitute “interest,” and the deductible portion of the final compensation package would thus increase exponentially.
Albertson’s has been unable to explain why Congress, in designing a taxation scheme to encourage the creation of qualified plans, would require an employer that maintains a nonqualified plan to defer taking a deduction on the basic amounts of a promised compensation package but nevertheless allow that employer to take current deductions on amounts that constitute a substantial portion of the compensation package, merely because that portion is classified as “interest.” Given that the interest payments will often constitute the bulk of the total compensation package that an employee under a nonqualified plan ultimately receives, it would make little sense to impose a matching requirement upon “basic” payments but not upon “interest” payments. Albertson’s interpretation of section 404 would seriously undermine the incentive structure designed by Congress to encourage employers to establish qualified plans.
An additional reason to reject Albertson’s statutory interpretation of section 404 is that, in certain cases, Albertson’s approach might actually create an incentive for employers to establish
nonqualified
plans. Whereas an employer who maintains a qualified plan may only take a current deduction for the basic amounts of promised compensation, an amount it actually has paid out, under Al-bertson’s approach an employer that maintains a nonqualified plan could take current deductions for “interest” payments that substantially exceed the basic amounts even though it has paid out none of these funds. Moreover, the employer could take advantage of these tax benefits without being constrained by the burdensome requirements associated with qualified plans. For this reason, characterizing the additional amounts as deductible interest, as Albertson’s suggest, would encourage employers to maintain non-qualified plans and thus directly contradict the statutory purpose underlying I.R.C. § 404.
Thus, Albertson’s proposal runs counter to the congressional scheme. It undermines Congress’ attempts to encourage employers to adopt qualified plans and, in some cases, directly contradicts the purpose of section 404 by creating an incentive to create non-qualified plans.
D.
Albertson’s Response
Albertson’s has not been able to refute the argument that its interpretation of section 404 undercuts the provision’s central purpose. Equally important, it offers us no reason why Congress would have wanted to treat the “interest” part of the deferred compensation package differently from the basic amounts for tax purposes.
Instead, Albertson’s rests its argument upon its contention that, because the plain language of § 404 only refers to “compensation” rather than “interest,” the employers have a statutory right to deduct the additional amounts as interest under § 163. In this connection, Albertson’s points out that section 404 prohibits deduction under sections 162 and 212 but not under section 163, and it is the latter section that governs the deduction of interest. Albertson’s argument as to the plain language of the statute is a strong one. We certainly agree that the additional payments resemble “interest” and that, under a literal reading of the statutory language, the deduction of interest is not affected by section 404. However, holding such payments to be deductible “interest” under section 404 would lead to an anomalous result: a taxation scheme designed to make nonqualified plans less attractive would in many cases provide incentives for adopting such plans, and a provision intended to apply the matching principle to nonqualified deferred compensation agreements would exempt substantial portions of DCA payments from its application.
In the end we are forced, therefore, to reject Albertson’s approach. We may not adopt a plain language interpretation of a statutory provision that directly undercuts the clear purpose of the statute. In
Brooks v. Donovan,
699 F.2d 1010 (9th Cir.1983), we refused to adopt a plain language interpretation of a statute governing pension funds. We reasoned that the “court must look beyond the express language of a statute where a literal interpretation Vould thwart the purpose of the overall statutory scheme or lead to an absurd or futile result.’ ”
Brooks,
699 F.2d at 1011 (quoting
International Tel. & Tel. Corp. v. General Tel. & Elec. Corp.,
518 F.2d 913, 917-918 (9th Cir.1975)). In reaching our conclusion, we followed the Supreme Court’s approach in
United States v. American Trucking Ass’ns.,
310 U.S. 534, 60 S.Ct. 1059, 84 L.Ed. 1345 (1940). There the Court noted that “[w]hen [a given] meaning has led to absurd results ... this Court has looked beyond the words to the purpose of the act. Frequently, however,
even when the plain meaning did not produce absurd results but merely an unreasonable one ‘plainly at variance with the policy of the legislation as a whole, ’ this Court has followed that purpose, rather than the literal words.” American Trucking Ass’ns.,
310 U.S. at 543 (emphasis added; citations omitted).
The Supreme Court’s decision in
Bob Jones University v. United States,
461 U.S. 574, 103 S.Ct. 2017, 76 L.Ed.2d 157 (1983), provides especially useful guidance in this regard. In
Bob Jones University,
the Supreme Court addressed the question of whether I.R.C. § 501(c)(3) — which provides that “corporations ... organized and operated exclusively for religious, charitable ... or educational purposes” are entitled to tax exemption — included educational institutions with racially discriminatory policies. Even
though Congress had explicitly outlined eight categories of exempt organizations in section 501(e)(3) without making any mention of additional requirements beyond those outlined in the statute, the Supreme Court interpreted the statute to include an additional requirement: the organization in question must serve a valid charitable purpose.
Bob Jones University,
461 U.S. at 592 n. 19, 103 S.Ct. at 2029 n. 19. Thus, the Court was willing to read additional language into the text of the statute because an alternative interpretation would undermine the fundamental purpose of the legislation. As the Court itself noted, “[i]t is a well-established canon of statutory construction that a court should go beyond the literal language of a statute if reliance on that language would defeat the plain purpose of the statute.”
Id.
at 586, 103 S.Ct. at 2025.
In rejecting Albertson’s appeal, we take heed of the Supreme Court’s instructions concerning the proper interpretation of the Internal Revenue Code when the plain language of the provision leads to an unreasonable result and directly contradicts its underlying purpose: the provision
“must
be analyzed and construed within the framework of the Internal Revenue Code and against the background of the congressional purposes.”
Id.
(emphasis added). For the reasons we have expressed, we conclude that, despite the literal wording of the statute, Congress could not have intended to exclude interest payments, a substantial part of the deferred compensation package, from the rule prohibiting deductions until such time as the employee receives the benefits. Indeed, the matching principle would not be much of a principle if so substantial a part of the deferred compensation package were excluded from its operation.
IV.
CONCLUSION
In sum, we decline to adopt Albertson’s interpretation of I.R.C. § 404. Whether or not the additional amounts constitute interest, allowing Albertson’s to deduct them pri- or to their receipt by their employees would contravene the clear purpose of the taxation scheme governing deferred compensation agreements. Accordingly, we vacate the portion of our original opinion dealing with deferred compensation agreements and affirm the Tax Court’s holding that Albertson’s may not currently deduct the additional amounts.
AFFIRMED.