Jos. Schlitz Brewing Company and the Stroh Brewery Company v. Milwaukee Brewery Workers' Pension Plan, Cross-Appellee

EASTERBROOK, Circuit Judge. Three brewers in Milwaukee — Schlitz, Miller, and Pabst — contributed to a multi-em-ployer pension plan. The brewers also bargained jointly with the union over wages, pensions, and other benefits, and administered the pension plan. In 1981 Schlitz closed its Milwaukee brewery. (Schlitz has since been acquired by Stroh; we use its old name for clarity.) We must decide how much Schlitz owes as withdrawal liability under the Multiemployer Pension Plan Amendments Act of 1980 (MPPAA). I Several sets of documents govern relations between the Milwaukee Brewery Workers’ Pension Plan and the Milwaukee Brewers’ Association (the employers’ collective bargaining group). An Association Agreement provides that the Labor Committee of the Brewers’ Association may make decisions only by consensus: “All decisions and action taken by the ... Labor Committee shall be by unanimous vote of the respective committee, with each brewery having a reasonable opportunity to cast its vote on the question.” The Plan was managed and administered by a Joint Committee consisting of four members appointed by the union and four members appointed by the employers. This committee was bound by the agreement establishing the Plan. That agreement could be amended by the “mutual agreement of the Union and the Employers”. So changes in the Plan required the assent of both labor and management, which had agreed to act by consensus. When the consensus ended the difficulties began. Schlitz withdrew from joint bargaining in January 1981 because it thought the other *997 employers too willing to accept the union’s demands. When giving notice of this withdrawal, Schlitz stated that it would continue to sit on the Plan’s Joint Committee and participate in the Plan. On June 1,1981, the union struck Schlitz, which responded by closing its Milwaukee brewery. On August 14 the Schlitz representative on the Joint Committee resigned, and Schlitz withdrew from the Plan. At the time, the Plan was underfunded to the tune of $111 million. Inflation was high, leading the Plan to estimate big future obligations to pay benefits, while steep interest rates had depressed the capital value of the Plan’s existing assets. Schlitz estimated that it would be responsible for $27 million of this shortfall under the “presumptive method” that the MPPAA specifies when a pension plan has not adopted another method of allocation. 29 U.S.C. § 1391(b). Withdrawal liability under the presumptive method depends on “the proportion of total employer contributions to the plan made by the withdrawing employer during certain 5-year periods. In essence, the withdrawal liability imposes on the withdrawing employer a share of the unfunded vested liability proportional to the employer’s share of contributions to the plan during the years of its participation.” Concrete Pipe & Products of California, Inc. v. Construction Laborers Pension Trust, — U.S. -, -, 113 S.Ct. 2264, 2273, 124 L.Ed.2d 539 (1993) (citations omitted). Under the “direct attribution” method, which the MPPAA permits most plans to elect, each employer is responsible for that portion of the unfunded, vested liabilities earned by its current and former employees. 29 U.S.C. § 1391(c)(4). See Artistic Carton Company v. Paper Industry Union-Management Pension Fund, 971 F.2d 1346, 1349-50 (7th Cir.1992). Miller and Pabst began to discuss adopting the direct attribution method, which altered the employers’ shares of the $111 million. According to their memos, these were the options: Method of Attribution Brewer Direct Presumptive Miller $31,100,000 (28%) $45,500,000 (41%) Pabst 38,800,000 (35%) 38,800,000 (35%) Schlitz 41,100,000 (37%) 26,700,000 (24%) Miller, Pabst, and the union entered into memoranda of understanding in May 1981 purporting to amend the Plan to adopt the direct attribution method. Schlitz’s assent was neither sought nor secured. On August 17,1981, the remaining members of the Joint Committee directed the Plan’s actuary to calculate Schlitz’s withdrawal liability under the direct attribution method. Schlitz first learned of the change in funding method when it received notice in September 1981 that it owed $41,031,406. Schlitz demanded arbitration under 29 U.S.C. § 1401(a)(1). Arbitrator Siegel concluded that the Plan could be amended only by unanimous consent. 9 E.B.C. 2385 (1988). In May 1981, when Miller, Pabst, and the union tried to adopt the direct attribution method, Schlitz was still a contributing employer and held a seat on the Joint Committee. Lack of unanimity left the presumptive method in effect. The parties’ actuaries jointly concluded that under the presumptive approach (and using actuarial assumptions not at issue here) Schlitz owed only $23.3 million; the arbitrator incorporated that figure into his award. 9 E.B.C. 2689 (1988). The Plan petitioned for review in the district court, which agreed with the arbitrator’s use of the presumptive method but added interest for the year in which Schlitz withdrew from the Plan, plus prejudgment interest on this interest. Both sides appeal. The district judge neglected to determine the amount of prejudgment interest, which usually is essential to finality (and thus appealability). Osterneck v. Ernst & Whinney, 489 U.S. 169, 109 S.Ct. 987, 103 L.Ed.2d 146 (1989). But the parties reached an agreement on the subject. The district court’s order of October 20,1992, apparently accepts this agreement, and Schlitz has paid in full. So although the *998 district court has not entered the form of judgment required by Fed.R.Civ.P. 58, its decision of October 20 is nonetheless appeal-able under the approach of Bankers Trust Co. v. Mollis, 435 U.S. 381, 98 S.Ct. 1117, 55 L.Ed.2d 357 (1978). II The extent of Schlitz’s liability turns on whether it is bound by the purported modification of the Plan in 1981. Was the change by the “mutual agreement of the Union and the Employers”, as the Plan’s governing document provides? The Plan does not deny that Schlitz was a participating “employer” in May 1981. Schlitz argues that the ordinary understanding of “mutual”, consistent with the Association Agreement’s rules for deci-sionmaking, means that the Plan remains unchanged unless the employers agree unanimously to change it. Schlitz was not aware of, and did not agree to, the change in May 1981. The Plan tells us, however, that Schlitz’s lack of assent is irrelevant. On its reading, “mutual agreement” does not mean that the Plan’s participating employers must reach consensus. To the contrary, once Schlitz left the Brewers’ Association, it subjected itself to the risk that the remaining members — its competitors — would “mutually” agree with the union to amend the Plan in a way singularly disadvantageous to Schlitz. Arbitrator Siegel agreed with Schlitz, interpreting the phrase “mutual agreement” to mean “agreement by all”— that is to say, all “employers” contributing to the Plan, as opposed to the two employers still engaged in joint bargaining with the union. Although the parties argue this appeal as if this court’s role is to interpret the documents from scratch, neither side explains why. We are a court of review and not of first instance. A vital issue in many appeals, including this one, is the standard of review — that is, the relation between this tribunal and its predecessors. On most issues appellate review is deferential. When an arbitrator makes the initial decision, judges have only a modest role to play. Paperworkers v. Misco, Inc., 484 U.S. 29, 108 S.Ct. 364, 98 L.Ed.2d 286 (1987); W.R. Grace & Co. v. Rubber Workers, 461 U.S. 757, 764, 103 S.Ct. 2177, 2182, 76 L.Ed.2d 298 (1983). Cases under the Arbitration Act, 9 U.S.C. §§ 1-16, or arising out of collective bargaining agreements, usually present a simple question: did the arbitrator interpret the parties’ agreement, or did he apply a rule he preferred to them agreement? If the former, the award is enforced; if the latter, not. Hill v. Norfolk and Western Ry., 814 F.2d 1192, 1194-95 (7th Cir.1987). Arbitrator Sie-gel unquestionably interpreted the Association Agreement and the Plan’s governing documents; he tried to carry out the parties’ rules rather than invent his own, and no one contends that the award is marred by fraud or corruption. See 9 U.S.C. § 10. If this were a standard labor or commercial case, we could end here. Aspects of this case resemble ordinary labor and commercial disputes. The Plan was created by labor and management, and a clause provides for arbitration in the event the Joint Committee is unable to resolve a dispute. Had today’s case arisen before 1980, then, arbitration would have occurred and the judicial role would have been minimal. But when enacting the MPPAA in 1980 Congress complicated matters substantially by simultaneously incorporating most provisions of the Arbitration Act, 29 U.S.C. § 1401(b)(3), while adopting a unique standard of review. Section 1401(c) provides that in a judicial proceeding to review or enforce an award, “there shall be a presumption, rebuttable only by a clear preponderance of the evidence, that the findings of fact made by the arbitrator were correct.” 29 U.S.C. § 1401(c). Conflating “clear and convincing evidence” with “preponderance of the evidence” to yield “clear preponderance of the evidence” would be bad enough if it referred to the task of a trier of fact; as a standard of review it makes no sense. As the Supreme Court remarked when trying to digest some related provisions of the MPPAA, Congress apparently confused the standards applicable to trial and reviewing bodies, producing an “incoherent” statute. Concrete Pipe, — U.S. at -, 113 S.Ct. at 2281. Repair is essential, and like the Court in Concrete Pipe we reconstruct by supplying the normal *999 rules: the arbitrator acts by a preponderance of the evidence, and a reviewing body inquires whether the facts so found are “clearly erroneous.” See Anderson v. Bessemer City, 470 U.S. 564, 573,105 S.Ct. 1504, 1511, 84 L.Ed.2d 518 (1985). Perhaps the “clearly erroneous” standard of review, rather than the “clear and convincing evidence” burden of persuasion, was the source of “clear” in § 1401(c). No matter the source, the standard is one of deference to the person who hears the evidence — although not as deferential as the “reasonable person” standard that applies to ordinary arbitration. Concrete Pipe, — U.S. at -, 113 S.Ct. at 2279-80. Section 1401(c) addresses findings of fact. What of legal questions? Facts are unique to the parties; law is general. The Pension Benefit Guaranty Corporation has provided that “[i]n reaching his decision, the arbitrator shall follow applicable law, as embodied in statutes, regulations, court decisions, interpretations of the agencies charged with enforcement of the Act, and other pertinent authorities,” 29 C.F.R. § 2641.4(a)(1). This regulation, promulgated under the PBGC’s authority to interpret the MPPAA, Concrete Pipe, — U.S. at -, 113 S.Ct. at 2279, implies non-deferential review. For this and related reasons we held in Trustees of Iron Workers Local 473 Pension Trust v. Allied Products Corp., 872 F.2d 208, 211-12 (7th Cir.1989), that the arbitrator’s legal conclusions are subject to de novo review. Yet the dominant theme remains deference, which accounts for our decision in Chicago Truck Drivers Pension Fund v. Louis Zahn Drug Co., 890 F.2d 1405, 1409-11 (7th Cir.1989), that the arbitrator’s application of law to fact — in the argot, a decision on “mixed questions of law and fact” — may be set aside only for clear- error. Where does the interpretation of pension plans and other agreements fit into this scheme? The answer within the world of litigation would be: It depends. Both the declaration of trust establishing the Plan and the Association Agreement are contracts. When a contract is clear, a court declares its meaning as a matter of law. When a contract is ambiguous, a court sends questions of interpretation to the trier of fact. RTC v. M & L Investments, 984 F.2d 190, 193 (7th Cir.1993); Woodbridge Place Apartments v. Washington Square Capital, Inc., 965 F.2d 1429, 1439 (7th Cir.1992). And the existence of ambiguity can present formidable problems. Pension agreements do not get to be ambiguous because of parol evidence, such as the fact that an employer joined with its fingers crossed behind its back, Central States Pension Fund v. Gerber Truck Service, Inc., 870 F.2d 1148 (7th Cir.1989) (in banc), or because of statements people made to each other about the meaning of otherwise-clear documents, see Bidlack v. Wheelabrator Corp., 993 F.2d 603, 608, 616 (7th Cir.1993) (in bane). Still, ambiguous language (intrinsic ambiguity) or an unexpected application of language that appears clear in its linguistic context (extrinsic ambiguity) may pose problems for a trier of fact. FDIC v. W.R. Grace & Co., 877 F.2d 614, 620-21 (7th Cir.1989). Courts ask whether the contract is “clear” in order to determine who makes the initial decision: judge or jury. Under the MPPAA, however, the arbitrator does the interpretation whether the text is clear or cloudy. Making the sorts of questions courts use to select between judge and jury the key to review of the arbitrator’s decision would add a layer of complexity to what should be a simple and expeditious system. Arbitration under the MPPAA is supposed to accelerate decision, but it has potential to delay matters by increasing from two to three the number of layers through which each dispute must pass; a debate in the district and appellate courts about whether the contract was clear (implying de novo review) or ambiguous (implying deferential review) would add to the potential for delay. Arbitrators serve a vital function in addition to speedy action: arbitrators may be chosen from persons knowledgeable about the industry and the kind of pension plan at issue, promoting accurate resolution of disputes. I. AM. National Pension Fund Benefit Plan C v. Stockton Tri Industries, 727 F.2d 1204, 1208 (D.C.Cir.1984). What appears “clear” to a judge alien to the industry may be ambiguous to an expert versed in its nuances; what appears *1000 cloudy to a judge may speak plainly to an arbitrator who understands the problems and solutions of the trade. Thus an effort to tailor the standard of review to the judge’s estimate of clarity could misfire. What is more, every time a court substitutes its judgment for that of the arbitrator, the benefit of proficiency is lost and disposition becomes more sluggish. Because the parallel to labor and commercial arbitration is imperfect given the MPPAA’s unique review provision — and because arbitration under the MPPAA sometimes is involuntary, and the standard of review must govern these cases as well as those in which the pension plan has its own arbitration clause — we adopt a standard that is slightly more demanding than the norm in labor law. Instead of asking only whether the arbitrator interpreted the contract (as opposed to making up his own rule), we ask whether the arbitrator adopted a reasoned interpretation of the documents. To avoid proliferation of standards (one novelty, in § 1401(c), is quite sufficient) we draw an analogy to the relation between agency and court under the Administrative Procedure Act. An agency must give thoughtful consideration to the. opposing arguments, but a reviewing court does not ask whether the agency was “right.” Instead it asks whether the decision was arbitrary or capricious, or otherwise not in accordance with law. That familiar standard should serve well under the MPPAA, a complex field in which the arbitrator serves in part as a surrogate for the PBGC. (The provisions of the MPPAA were designed in large measure to avoid a sequence of withdrawals that would shift liability to the guaranty fund maintained and administered by the PBGC.) Is deferential review of the arbitrator’s interpretation inconsistent with 29 U.S.C. § 1401(a)(3)(A), which provides that “any determination made by a plan sponsor under [§§ 1381-99, 1405] is presumed correct unless the party contesting the determination shows by a preponderance of the evidence that the determination was unreasonable or clearly erroneous”? Not at all. Section 1401(a)(3)(A) establishes a presumption that the arbitrator must consider. It does not affect the allocation of interpretive authority between the arbitrator and a reviewing court. Just what § 1401(a)(3)(A) does do was one of the subjects the Supreme Court considered in Concrete Pipe, where it understood this language to mean that in the proceeding before the arbitrator the employer must “disprove a [plan sponsor’s] challenged factual determination by a preponderance” of the evidence. — U.S. at -, 113 S.Ct. at 2283 (emphasis added). There are no factual disputes in our case, and the Plan did not make any factual determinations to which the arbitrator owed deference. Arbitrator Siegel’s interpretation of “mutual agreement” to mean “by unanimous agreement of all parties concerned” — that “the Union and all the Employers agree to any amendment of the Plan”, 9 E.B.C. at 2394 — is neither arbitrary nor capricious. The “mutual agreement” language applies to all changes in the pension plan. A conclusion that Miller and Pabst could change the rules for computing withdrawal liability without Schlitz’s assent would imply an equal power to increase the pensions and other benefits promised to labor, to increase the required rate of contributions, and to make any other alteration. No sane firm gives its business rivals such power, and the arbitrator did not act whimsically in concluding that Schlitz had not done so. “Mutual agreement” reads naturally just as the arbitrator understood it, and the parties’ own practice under this language was one of negotiating unanimous agreement before making any alterations in the Plan. According to the Plan, ¶ 12.08(b)(13) of the Plan Document, which gives the Joint Committee power to interpret ambiguous language, means that the arbitrator could not disagree with the Joint Committee’s assessment of withdrawal liability. Concrete Pipe scotches that contention by observing that a person ordered to pay a large sum of withdrawal liability is entitled to an independent review by a neutral decisionmaker. — U.S. at -, 113 S.Ct. at 2283. Withdrawal liability depends on the MPPAA, not on the parties’ agreement, and the statute does not' permit one person to make an unreviewable assessment against another. What is more, *1001 it distorts the events of 1981 to say that the Joint Committee “interpreted” any part of the Plan Document. Schlitz disputes the validity of an amendment to the Plan, not the interpretation of that amendment if the direct attribution method was properly adopted. Remember how the Joint Committee got involved: months after Miller, Pabst, and the union purported to amend the Plan, the Joint Committee instructed the Plan’s actuary to calculate Sehlitz’s liability pursuant to the direct attribution method. The Plan has not provided us with any reason to believe that, in doing so, the Joint Committee itself interpreted “mutual agreement.” We can imagine one last argument. Might it be possible that, putting the machinations of May 1981 to one side, the remaining employers and the union properly amended the Plan in September 1981 — after Schlitz’s representative had withdrawn from both the Association and the Plan? Perhaps, but Schlitz would not be the worse for it. The MPPAA provides that amendments adopted after January 31, 1981, changing the method of calculating withdrawal liability do not apply to an employer that withdrew before the amendment. 29 U.S.C. § 1394(a). The rationale for this rule is the same as the rationale for unanimity in the Association Agreement: otherwise businesses can gang up to squeeze their competitor. Thus we accept the arbitrator’s decision that the acts of Miller, Pabst, and the union in May and September 1991 did not adopt the direct attribution method — at least, did not adopt that method in a way that permits the Plan to apply it to Schlitz. Ill The Plan has a fallback position, with several steps. (1) When selecting a method of withdrawal liability, all three employers acted as fiduciaries to the Plan. (2) A fiduciary acting in May 1981 would have been required to prefer the direct attribution method to the presumptive method. (3) Had Schlitz attended the meetings in May 1981 it would have voted for the direct attribution method even though that disserved its own interests (and had Schlitz violated its fiduciary responsibility, a court would have intervened). (4) Because the direct attribution method would have been adopted by unanimous consent had Schlitz fulfilled its responsibilities, the arbitrator should have treated the Plan as if it had been amended. There is a certain logic to this, but also an air of unreality. Pabst and Miller did not even invite Schlitz to the meetings. Having deprived it of an opportunity to act, they are scarcely in a position to dictate the outcome. Consider a parallel in corporate law. Members of the board of directors are fiduciaries. Corporations also have rules for determining the validity of the board’s acts— quorum requirements and sometimes super-majority rules. Could a single member of the board change the corporation’s bylaws, without calling a meeting (and necessarily without satisfying quorum and other requirements), by declaring that all fiduciaries must support his proposal, making a meeting superfluous? Most doubtful, yet this is what the Plan’s position entails. At all events, the Plan’s position fails at the first step. Employers do not act as fiduciaries when establishing the terms of a pension plan. According to the Employee Retirement Income Security Act (ERISA), to which the MPPAA belongs, a person is a fiduciary of a pension plan if: he exercises any discretionary authority or discretionary control respecting management of such plan or exercises any authority or control respecting management or disposition of its assets.... 29 U.S.C. § 1002(21)(A)(i). The fiduciary “shall discharge his duties with respect to a plan solely in the interest of the participants and beneficiaries”. 29 U.S.C. § 1104(a)(1). Managing a plan is some distance from establishing the terms of a plan. See Young v. Standard Oil (Indiana), 849 F.2d 1039, 1044-45 (7th Cir.1988); United Independent Flight Officers, Inc. v. United Air Lines, Inc., 756 F.2d 1262, 1268-69 (7th Cir.1985); Amato v. Western Union International Inc., 773 F.2d 1402, 1416-17 (2d Cir.1985); Hozier v. Midwest Fasteners, Inc., 908 F.2d 1155, 1161 (3d Cir.1990); Musto v. American General Corp., 861 F.2d 897, 911-13 (6th Cir.1988); Cunha v. Ward Foods, Inc., 804 F.2d *1002 1418, 1432-33 (9th Cir.1986). These and other cases hold that firms may set the terms of a pension plan, such as the level of benefits achieved after a given length of service, as employers rather than fiduciaries, a position that not only makes sense of the “management” language in ERISA but also recognizes that pension benefits are deferred compensation. Employers no more have a fiduciary obligation to assure larger pensions for their workers than they do to pay higher wages. The Plan concedes that in establishing the level of benefits and the employers’ contributions the brewers do not act as fiduciaries. Nonetheless it contends that amendments “which impact vested benefits in a defined benefit plan are fiduciary functions,” (emphasis added), and that the choice of a method of determining withdrawal liability is one such amendment: [Ajlloeation may bear upon the collectability of the funds. Delay and additional administrative costs associated with pursuing a financially weakened or bankrupt employer ... may reduce the fund’s ability to make the vested, earned benefit payments .... Allocation impacts future wages and benefits to plan participants as the remaining employers struggle to cover inherited unfunded liability or are relieved from that burden. The method selected may trigger an employer to withdraw from a plan in order to take advantage of a method which it believes is favorable. Implementing one method may result in significant administrative costs to a plan which are not associated with other methods. Of course one could say the same thing about levels of funding. If employers have a fiduciary obligation to use the method of computing withdrawal liability that maximizes the pension plan’s current wealth, do they not have an even clearer obligation to increase their contributions so that the plan is not underfunded in the first place? Is not “underfunding” the most obvious breach of duty? Yet at oral argument the Plan (speaking for Miller and Pabst, no doubt) stoutly denied that employers have any fiduciary obligation to increase their payments to a plan in order to bring it to full funding. If selecting a level of funding is not a fiduciary obligation, how can selecting a method of computing withdrawal liability be? Withdrawal liability is to be calculated pursuant to the presumptive formula in 29 U.S.C. § 1391(b), but parties may select the direct attribution method, one of two other methods, or make up their own method with approval by the PBGC. 29 U.S.C. § 1391(c)(5). Construction industry plans must use the presumptive method, § 1391(c)(1), but others may select freely; the statute does not establish criteria under which a plan must adopt one method or another. The absence of criteria leaves the choice in private hands — which is to say that like levels of benefits it is a subject for contract rather than legal obligation. A plan established jointly by labor and management is governed by the statutory obligation of good faith bargaining. 29 U.S.C. § 158(d). But ERISA also governs pension and welfare plans established unilaterally, and employers are free to alter or disestablish these plans at will. Electronic Workers v. Murata Erie North America, Inc., 980 F.2d 889 (3d Cir.1992); McGann v. H & H Music Co., 946 F.2d 401 (5th Cir.1991). For what it is worth, we add that the second step of the Plan’s argument is no more sound than the first. Each method produces a total of $111 million in withdrawal liability; only the allocation among employers is at issue. The Plan’s expert testified before the arbitrator that the direct attribution method was in the Plan’s best interest in 1981 because it accelerated the rate at which the Plan could become fully funded. Fundamentally, the expert testified that a bird in the hand is worth two in the bush. Miller, Pabst, and union representatives could have concluded, as this witness did, that direct attribution served the Plan’s interests. We say “could have” because nothing in the record reflects that those who met in May 1981 acted as fiduciaries or, indeed, gave the slightest weight to the interests of employees with vested benefits. None of the contemporaneous documents so much as mentions the best interests of the Plan. Although the Plan talks on appeal about the necessity of *1003 matching the amount of withdrawal liability with each employer’s financial strength, no such concerns surfaced during the discussions. Instead, each employer acted in its self-interest — Sehlitz by bargaining independently, and Miller and Pabst by voting to shift more of the burden of the Plan’s unfunded liability onto Sehlitz. The papers document a series of attempts by each brewery to minimize its financial exposure while maximizing the obligations of its competitors. Maybe transferring a greater share to Sehlitz increased the probability of collection, but Sehlitz insists that in May 1981 Miller was best able to pay, and Miller’s withdrawal liability went down. Nothing in the record demonstrates that firms with only the employees’ interests at heart would have selected the direct attribution method. Parties to a multi-employer pension plan may agree that decisions regarding allocation of withdrawal liability must be made in the plan’s best interests, or subject to some other heightened standard. The parties to this Plan, however, reserved the subject to the collective bargaining process, in which each looked out for its own interests. Federal law does not prohibit that decision. IV The Plan’s actuary includes, as part of Schlitz’s withdrawal liability, interest for all of 1981. Arbitrator Siegel disallowed this interest as inconsistent with the MPPAA. The district judge disagreed and ordered Sehlitz to pay the interest, plus prejudgment interest on that interest. The parties agree that the propriety of assessing this interest is a question of law, on which our review is plenary, rather than a question on which the arbitrator and courts must defer to the actuary under 29 U.S.C. § 1401(a)(3)(B). See also Concrete Pipe, — U.S. at -, 113 S.Ct. at 2284-86. Sehlitz tells us that the district court’s reading of the MPPAA conflicts with that of the PBGC, which has interpreted the statute to prohibit the assessment of such interest. If the PBGC had issued a regulation or other formal interpretation on the question, we would respect its judgment to the extent Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984), requires. The PBGC has been delegated powers to enforce the MPPAA, see 29 U.S.C. § 1303(e)(1), which usually leads courts to accept its understanding of that complex (and all too frequently ambiguous) law. E.g., Mead Corp. v. Tilley, 490 U.S. 714, 726, 109 S.Ct. 2156, 2164, 104 L.Ed.2d 796 (1989); Allied Products Corp., 872 F.2d at 210 n. 2; Robbins v. McNicholas Transportation Co., 819 F.2d 682, 686 (7th Cir.1987). Yet the PBGC has not issued any regulation or opinion on this subject. The interpretation to which Sehlitz points appears in a brief filed as amicus curiae in Huber v. Casablanca Industries, Inc., 916 F.2d 85 (3d Cir.1990). Courts do not defer to “agency litigating positions that are wholly unsupported by regulations, rulings, or administrative practice.” Bowen v. Georgetown University Hospital, 488 U.S. 204, 212, 109 S.Ct. 468, 474, 102 L.Ed.2d 493 (1988). Only the agency, and not its lawyers, exercises delegated power. SEC v. Chenery Corp., 318 U.S. 80, 88, 63 S.Ct. 454, 459-60, 87 L.Ed. 626 (1943). Did that brief articulate the policy of the agency, or was it a concoction of appellate counsel? The PBGC did not participate in this case. Even were we sure that the position taken in Huber was the PBGC’s rather than its lawyer’s, see Georgetown University Hospital, 488 U.S. at 212, 109 S.Ct. at 473-74; Atlanta College of Medical and Dental Careers, Inc. v. Riley, 987 F.2d 821, 828 n. 10 (D.C.Cir.1993); cf. Martin v. Occupational Safety and Health Review Commission, 499 U.S. 144, 156-57, 111 S.Ct. 1171, 1179, 113 L.Ed.2d 117 (1991), we could not be sure that it subscribes to this view today. See Mead Corp., 490 U.S. at 726 n. 11, 109 S.Ct. at 2164 n. 11. We must decipher the MPPAA for ourselves. Multi-employer plans calculate their funding status only once per year, on December 31. If they are underfunded on that date, an employer withdrawing during the next year owes withdrawal liability in an amount determined by the December 31 calculation. 29 U.S.C. § 1391(c)(2)(C)®. An employer withdrawing in mid-year is entitled to amortize its liability over a period set by the Plan. Section 1399(c)(1)(A)® provides the rule un *1004 der which the lump sum is converted to installments: “[A]n employer shall pay the amount determined under [§ 1391] ... over the period of years necessary to amortize the amount in level annual payments ..., calculated as if the first payment were made on the first day of the plan year following the plan year in which the withdrawal occurs’’ (emphasis added). In our case the Plan year apparently coincides with the calendar year. A withdrawing employer’s schedule thus assumes a delay in payment — although the employer may pay sooner, and in our case it did so. Schlitz, which withdrew in August 1981, made its first payment on October 29, 1981. An assessment of interest between the date of withdrawal and the date on which payments begin, in order to make the amortization schedule actuarially correct, would not be troubling. Huber, 916 F.2d at 98-100. It may help to think of the employer as “borrowing” the amount of withdrawal liability from the plan on the date of withdrawal, and owing this sum plus interest under the schedule set per § 1399(e)(l)(A)(i). But the Plan’s actuary did not start the interest in August 1981, or on the date the Plan presented its demand, or on the date the first payment came due. Instead the actuary computed interest from December 31, 1980, while Schlitz was still a contributing employer. The theory is that if the end of the last calendar year is used to compute the Plan’s funding shortfall, it should be used for all purposes, including interest. Other courts refer to this as “pre-demand” or “gap year” interest. By instructing plans to use the amount of unfunded vested benefits at year-end, Congress has spared actuaries the need to revalue plan assets and liabilities every time an employer withdraws during the year. Cf. 29 U.S.C. § 1082(c)(9) (requiring every plan to calculate the plan’s liability yearly). Sometimes the underfunding will rise between December 31 and withdrawal; sometimes it will fall; the law of large numbers evens these things out for plans (although not for particular employers). Establishing a simple rule for calculating funding shortfalls has nothing to do with interest, and the MPPAA does not imply that withdrawing employers owe anything as of the end of the year before they drop out. Schlitz owed $23.3 million to the Plan as of August 1981, not as of December 1980. Schlitz was not liable for a penny until it withdrew from the Plan and received a notice of liability. The MPPAA specifically authorizes assessing the withdrawing employer with interest under several circumstances, 29 U.S.C. § 1399(c)(3), (c)(4), (e)(5), but says nothing about interest prior to withdrawal. Neither the word “amortize” (on which the district court’s decision turned) nor any other language of the statute supports an approach that calls for pre-withdrawal interest. We join the fourth circuit in concluding that the MPPAA does not authorize the assessment of interest before an employer’s obligation to pay arises. Borden, Inc. v. Bakery Union Pension, 974 F.2d 528, 536 (4th Cir.1992). To the extent Huber suggests that there are circumstances under which an employer must pay interest for the gap between the December 31 calculation date and the date of actual withdrawal, we disapprove that decision. (Whether Huber actually holds such a thing is doubtful.) The Plan has not argued that it is entitled to interest for the p'eriod between withdrawal in August 1981 and the commencement of the amortization schedule in January 1982. Hence there is no basis on which to question the arbitrator’s decision. The district court’s judgment is affirmed to the extent it holds that Schlitz’s withdrawal liability must be determined according to the presumptive method and reversed to the extent it requires Schlitz to pay interest for 1981. The case is remanded with instructions to enforce the arbitrator’s award in its entirety.