The opinion
EASTERBROOK, Circuit Judge,
dissenting.
An employer, fearing a rival with a reputation for low wages (more to the point, low prices) demands a most-favored-nations clause, under which it would receive the benefit of any concessions the union makes to the rival. Unions prefer not to make such bargains. A strike looms: the employer will not sign a contract without a most-favored-nations clause, and the union will not sign a contract with one.
At the last minute they agree to replace the contested clause with a reopener. Although the contract is to run for three years, if the dreaded competitor enters the market sooner either side may renew negotiations with the usual privileges: after *900 good faith bargaining to impasse, the employer may implement its final offer, the union may strike, or both. Electrical Workers Local 47 v. NLRB, 927 F.2d 635 (D.C.Cir.1991). A reopener can work in either side’s favor depending on economic conditions. It does not modify or contradict any other term of the contract; it is part of the terms, giving each side an option.
The rival entered; the employer exercised the right to reopen; union and employer bargained to impasse, and the employer implemented its final offer. When the dust settled the union retained the fundamental terms of the contract (including wages) and won back pay for its members. But it did not seek back pay for employees who left the firm before that settlement. These employees tried to tax the union with breach of its duty of fair representation but lost. Merit v. Jewel Companies, Inc., 848 F.2d 761 (7th Cir.1988). So also the ex-employees lost their contract suit against the employer. A jury determined that the deal was as I have described it, that the plaintiffs received their contractual due.
The reason this case went to a jury at all is that the reopener was oral, while the bulk of the pact was written. But as the union conceded that there had been an oral side agreement, disputing only the employer’s description of its content, the parol evidence rule did not prevent the jury from determining whose version is correct. My colleagues agree with the trial judge, 734 F.Supp. 330 (N.D.111.1990), and 1990 WL 71027 , 1990 U.S.Dist. Lexis 5486, that the doctrines applied to commercial cases enforce that oral agreement. Had this been a long-term contract for coal with an oral agreement to renegotiate the price if a low-cost strip mine should open, or a contract to build a skyscraper with an oral agreement to renegotiate the number of stories if the real estate market should collapse, the agreement could be enforced.
But not when the commodity is labor, the majority says. If my colleagues thought there was some problem with the feeble enforcement of the parol evidence rule in modern litigation, I would share their concern. The parol evidence rule and the statute of frauds reduce the costs of fabrication, litigation, and error — at the expense of making the bargaining process less flexible and withholding enforcement of some bargains. Most courts today think the costs of strict insistence on written instruments excessive. Whether they are right in believing this is an interesting question. What the majority does, however, is not signal a revival of the writing requirement; it maintains, instead, that labor law is different. What, then, is the source of the difference?
My colleagues point to “national labor policy”, which they say mandates that “collective bargaining agreements must be more secure than garden variety contracts.” 945 F.2d at 894. Labor policy may be found in labor statutes. It turns out that the National Labor Relations Act does not forbid oral agreements. Quite the contrary, the Act requires only one sort of agreement to be in writing. Pension plans must be written down, and the writings will be strictly enforced on command of both labor, 29 U.S.C. § 186 (c)(5)(B), and pension, 29 U.S.C. § 1145 , laws. See Central States Pension Fund v. Gerber Truck Service, Inc., 870 F.2d 1148 (7th Cir.1989) (in banc). When Congress singles out one kind of agreement for a strict writing requirement (as it did to protect the interests of the pension plans, which are not present at the bargaining table), it makes sense to infer that labor and management may decide for themselves how formal the rest of their dealings shall be. If labor policy really emphasized stability (and protection of workers from their own leaders’ “backroom deals”, 945 F.2d at 899), the NLRA would require the commitment of all agreements to writing, would require leaders to submit them to vote by members, and would require the written pact to be circulated far enough in advance of the vote that it could be digested and intelligently debated. Labor law requires none of these things, as my colleagues concede, id. at 895. By any reckoning, ordinary commercial contracts — to which §§ 2-201, 2-202, and 2-209 of the Uniform Commercial Code *901 apply — are more secure against disputes about oral colloquies than are labor agreements governed by § 8(d) of the NLRA, 29 U.S.C. § 158 (d). Labor law emphasizes flexibility, not certainty.
Flexibility is a commodity negotiators value, and not only because it is impossible to write down all of the understandings and practices (even all of the “important” understandings and practices) that make up the terms and conditions of employment. Collective bargaining agreements are relational contracts. They do not settle all important terms; rather they establish a framework within which the parties compose their differences.
It is perverse to say that the contracting process in labor must be more formal than the contracting process in shipping or construction or natural resources. You can define how much coal to sell and where to deliver it; you can set the duration of a charter party. Labor agreements govern the ongoing relations among thousands of persons and affect matters not so easy to specify. Rigidity backfires. Competitive conditions and technology change. Labor relations must change too, even over so short a period as three years. Failure to adjust to new developments means failure in product markets (look at the automobile and railroad industries), and when the employer slips in product markets labor takes the fall. If a low-price rival enters and siphons business, a firm bound to pay high wages will lay off many workers. Layoffs hurt the employees with the shortest tenure — that is, the persons who are plaintiffs in this case. Greater flexibility promotes responses in the marketplace that may benefit both sides.
Ongoing accommodation is a hallmark of labor relations. Representatives of labor and management meet continually to discuss both individual grievances and more general questions. Changes may ensue, with or without written agreements. “[T]he parties to a CBA [collective bargaining agreement] may tacitly acquiesce to an amendment of the agreement through their course of dealing.” Matuszak v. Torrington Co., 927 F.2d 320, 324 (7th Cir.1991). Because labor and management recognize that even this process cannot be comprehensive, they deputize arbitrators to resolve still more issues. Arbitrators sometimes are limited to grievances, but more commonly their powers extend to all interpretive debates. Courts regularly say that arbitrators are not bound by the parol evidence rule, that because collective bargaining agreements are incomplete, and so much of labor relations consists in oral understandings, arbitrators may use these oral exchanges to supplement or even contradict the written terms. E.g., SEPTA v. Railroad Signalmen, 882 F.2d 778, 784 (3d Cir.1989); Loveless v. Eastern Air Lines, 681 F.2d 1272, 1279-80 (11th Cir.1982); NLRB v. L.B. Priester & Sons, Inc., 669 F.2d 355, 365 (5th Cir.1982); Piggly Wiggly Operators’ Warehouse, Inc. v. Piggly Wiggly Operators’ Warehouse Independent Truck Drivers Union, 611 F.2d 580 (5th Cir.1980); Machinists v. Sargent Industries, 522 F.2d 280 (6th Cir.1975); Tolbert v. Union Carbide Corp., 495 F.2d 719, 721 (4th Cir.1974). If my colleagues are right, these decisions must be wrong.
The Supreme Court has said repeatedly that labor relations require more flexibility than the common law allows — not, as my colleagues hold, more formality. E.g., Transportation—Communication Emplo yees v. Union Pacific R.R., 385 U.S. 157, 160-61 , 87 S.Ct. 369, 371 , 17 L.Ed.2d 264 (1966) (“A collective bargaining agreement is not ... governed by the same old common-law concepts which control such private contracts. ... In order to interpret such an agreement it is necessary to consider the scope of other related collective bargaining agreements, as well as the practice, usage and custom pertaining to all such agreements.”); United Steelworkers v. Warrior & Gulf Navigation Co., 363 U.S. 574, 578-81 , 80 S.Ct. 1347, 1350-52 , 4 L.Ed.2d 1409 (1960); McKinney v. Missouri-Kansas-Texas R.R., 357 U.S. 265, 273-74 , 78 S.Ct. 1222, 1227-28 , 2 L.Ed.2d 1305 (1958). See also Consolidated Rail Corp. v. Railway Labor Executives’Association, 491 U.S. 299, 308-09 , 109 S.Ct. 2477, 2483 , 105 L.Ed.2d 250 (1989) (describ *902 ing “flexibility” as a hallmark of labor relations under both the NLRA and the Railway Labor Act and adjuring courts not to disfavor informal arrangements).
Flexibility entails giving labor and management the option of having a fully integrated agreement, of banning oral side agreements, as well as the option of coming to partially-written, partially-oral understandings. My colleagues concede that this agreement, which lacked an integration clause, did not ban oral understandings. Our question is whether labor law forbids such side agreements even when the parties decide against an integrated writing. A handful of cases say that it does; the majority relies on Gatliff Coal Co. v. Cox, 152 F.2d 52 (6th Cir.1945), and a smattering of unreviewed decisions by district courts, all more than a generation old. Gatliff rests in part on the premise that the NLRA “contemplates that a collective bargaining agreement be in writing”, 152 F.2d at 56 , which it obtained by misreading H.J. Heinz Co. v. NLRB, 311 U.S. 514, 523-26 , 61 S.Ct. 320, 324-26 , 85 L.Ed. 309 (1941). Heinz holds that it is an unfair labor practice for the employer, having reached an agreement with the union, to refuse to sign a document embodying its agreement. That labor is entitled to a written memorial of a bargain hardly shows that labor is forbidden to reach oral understandings with management. Gatliff dealt with a common problem: a member of a multi-employer bargaining unit refused to sign the pact the association reached with the union, pleading that local officials said that it would not have to. Today a belated effort to withdraw from a multi-employer unit would be dealt with as an unfair labor practice, see Charles D. Bonanno Linen Service, Inc. v. NLRB, 454 U.S. 404 , 102 S.Ct. 720 , 70 L.Ed.2d 656 (1982), and not as a problem of parol evidence. After the Labor Board resolved the multiemployer unit problem in 1958, Retail Associates, Inc., 120 N.L.R.B. 388 (1958), Gatliff vanished without a trace. 1 More recent cases in the Supreme Court (e.g., Conrail and Union Pacific), in the sixth circuit (e.g., Sargent), in our own circuit (e.g., Matus- zak), and in other circuits take a different approach to contractual formalities. For example, the fifth circuit believes that the parol evidence rule does not apply at all in labor law. Manville Forest Products, Inc. v. Paperworkers Union, 831 F.2d 72 , 75-76 (5th Cir.1987). Going to the extreme opposite from my colleagues, Manville said that “national labor policy” forbids labor and management to have a fully integrated contract no matter how strongly they prefer the benefits of certainty. 2 See also West- *903 ing house Elevators of Puerto Rico, Inc. v. S.I.U. de Puerto Rico, 583 F.2d 1184 (1st Cir.1978); cf. Bokunewicz v. Purolator Products, Inc., 907 F.2d 1396, 1401-02 (3d Cir.1990). I am no more enamored of an approach that dishonors attempts to put everything in writing than I am of an approach that dishonors oral agreements when the parties did not want their writings to be complete. There is a happy middle ground: Parties may choose for themselves.
This sets up the majority’s argument that the union did choose to exclude oral agreements, by requiring ratification of the pact by the membership. Technically, the argument is that the union’s negotiators lacked actual authority to agree to a reo-pener, because the union’s constitution requires ratification, and they lacked apparent authority because Jewel not only knew about this requirement but also understood that the union’s negotiating team was not going to tell the membership about the reopener. Notice that this argument does not supply any reason to differentiate labor law from other bodies of law. “Agreement” by an agent lacking both actual and apparent authority is empty in all corners of the law. Claims about how labor law is special melt away if, in the end, this case is about the negotiators’ lack of authority.
Four issues precede the question whether defects in the ratification process prevent enforcement of the reopener. First: do provisions of a union’s constitution affect the validity of labor agreements? The National Labor Relations Board thinks not. McLaughlin & Moran, Inc., 299 N.L.R.B. No. 7 (1990), holds that once the responsible officials of a union report to the firm that the union has approved an agreement, it is binding — even if the union did not follow prescribed procedures. The Board wrote that “it is none of the Employer’s business how (or even whether) the [Union] obtains the employees’ approval” of the agreement, and that “it is for the union, and not the employer, to construe the meaning of the union’s internal requirements for ratification.” Any other approach, the Board thought, would make it “difficult, if not impossible, for the parties to a collective-bargaining agreement to arrive at a final settlement without the fear of being forced into protracted litigation regarding the union’s compliance with its own procedures, clearly a collateral issue. The encouragement of such industrial instability could not have been with the in-tendment of the Act.” McLaughlin & Moran reiterates a doctrine the Board has applied for decades. E.g., NLRB v. M & M Oldsmobile, Inc., 377 F.2d 712 (2d Cir.1967); Houchens Market of Elizabethtown, Inc. v. NLRB, 375 F.2d 208, 211-12 (6th Cir.1967); NLRB v. Darlington Veneer Co., 236 F.2d 85, 88 (4th Cir.1956). Darlington holds that the union’s ratification procedures affect the validity of the agreement only to the extent the employer accepts them; otherwise labor law makes the word of the negotiators the union’s bond. To put this differently: Labor’s negotiators have authority, as a matter of federal law, no matter what the union’s internal procedures may be. This principle is conclusive against plaintiffs. I cannot find any contrary holding. Yet my colleagues, who extoll the importance of stability, also promote the sort of litigation regarding internal union procedures that, the Board and other circuits believe, undermines stability.
Second: if contrary to McLaughlin & Moran the union’s procedures matter, did Jewel know that the union’s rules required ratification term-by-term, and that the local’s officers would keep one term under *904 wraps? Unless Jewel knew, the union’s officers acted with apparent authority. Jewel denies having this knowledge, and the jury’s verdict hardly resolves this question against it. The majority distinguishes McLaughlin & Moran because that employer lacked knowledge of union’s internal rules while ignoring Jewel’s protest that it was in the same position. At all events this factor plays no role in the Board’s analysis or that of other circuits. 3 Third: does a defect in ratification apply exclusively to the oral portions of the agreement? Why should Jewel be bound by the provisions favorable to the union (those the union’s officers told the members about) but be unable to avail itself of the provisions favorable to itself? Ratification, one would think, is an all-or-nothing business. Fourth: why is Jewel, as opposed to the union, answerable to persons injured by defects in the ratification process? As the majority paints this picture, the union’s officers set out to hoodwink the members by keeping under wraps a reopener agreement that the members would oppose. Such a wrong sounds distinctly like a breach of the duty of fair representation. Yet three years ago this court held that these plaintiffs may not recover from this union on such a theory. 848 F.2d 761 . Suppose the majority could surmount these obstacles. More remain.
Nothing in the union’s constitution requires the officers to submit written documents to the members for perusal before a vote. This local does not circulate agreements to its members. Once the negotiating team reaches an agreement with management, the officers call a meeting at whieh they recount the terms they think significant. They are free under both the international union’s constitution and their own practices to trumpet the features they deem salient while overlooking concessions to management. Of course members may ask questions during debate, but there is no indication in this record that anyone asked whether the agreement included a reopener, let alone that the officers answered falsely. Treating a ratification requirement as if it were a writing requirement is a serious misstep.
Practice illuminates how the union understood the meaning of ratification. The jury found — as the parties agree — that union and management routinely left matters including the number of personal days off and details of the seniority system to oral agreements that were not described to the members before ratification. Judge Pos-ner asked the jury to decide whether under the parties’ practice an agreement had to be written in order to be valid. The jury determined that writing was unnecessary. My colleagues deem details such as seniority and days off to be trivial — questionable as an original matter, but no ground for upsetting the jury’s contrary decision on the same point.
Especially not when one of the examples presented to the jury was — another oral reopener agreement! The trial judge wrote:
Merk’s argument from triviality is contradicted by the evidence of a reopener agreed to in 1969. That agreement was premised upon a raid, not by Cub Food, but by the Teamsters, and allowed the contract to be reopened to renegotiate *905 higher wages. If the jury concluded that there was an agreement to reopen the contract in 1969, as it could have, and that that agreement was not ratified, then the agreement in this case was no more significant to Jewel than the 1969 reopener was to the Union, and could be valid even if unratified.
1990 WL 71027 at *5, 1990 U.S. Dist. Lexis 5486 at *13-14 (emphasis in original). Given the protection the seventh amendment accords to jury verdicts, you would think this evidence conclusive. But no; the majority says that the 1969 reopener is different because “the 1969 provision furthered the interests of all parties: Jewel benefited by keeping the Teamsters out of its stores, the Union benefited by retaining its membership and the members benefited by gaining a possible raise in salary.” 945 F.2d at 897 (emphasis in original). This sounds more like a lawyer’s closing argument to a jury than like a reason why the jury’s verdict must be set aside. After page 893 of the majority’s opinion (where it concludes that the jury could enforce the oral reopener under the rules usually applied in civil litigation), you do not encounter the word “jury” again. My colleagues make de novo findings on all factual disputes and draw their own inferences.
Disposing of a raft of oral agreements on the ground that they are trivial, and of a vital oral agreement on the ground that it could propel wages upward, supports a decision for these plaintiffs only at substantial damage to other values. Recall that the majority started from the premise that “collective bargaining agreements must be more secure than garden variety contracts.” 945 F.2d at 894. How is the agreement “more secure” if labor and management cannot know whether it is binding until they obtain the views of three appellate judges on how “important” it is? Importance lies in the eye of the beholder (I think that seniority provisions do not belong in the “trivial” class to which my colleagues consign them). Worse, the majority makes it clear that enforceability depends on who wins and who loses ex post. Negotiators need to know while they are dickering which agreements will be enforced and which not; people at the bargaining table in January 1983 cannot wait for a court in September 1991 to tell them which deals are sufficiently unimportant, and which turn out sufficiently favorable to the workers, to be enforced.
Consider again the position of the parties when Jewel was insisting on the most-favored-nations clause. What options had they?
• They could have reached impasse, followed by Jewel’s implementation of its final offer (including the most-favored-nations clause), which might have precipitated a strike. All workers either would have been subject to the most-favored-nations clause (if they continued to work on Jewel’s terms) or would have lost wages during the struggle. Jewel might or might not have been able to “break” the union during a strike.
• They could have signed an agreement with a most-favored-nations clause. The membership might have ratified this (with results less favorable to the employees than what actually happened) or might have rejected it (with the same results as impasse).
• They could have signed an agreement with a written reopener. The membership might have ratified this (with results less favorable to the employees than what actually occurred — recall that uncertain enforceability of the oral reopener led Jewel to compromise with the union) or might have rejected it (with the same results as impasse).
• They could have signed a one-year rather than a three-year agreement, with effects fundamentally the same as a three-year agreement plus a reopener.
• They could have signed a three-year agreement without a reopener, but with lower salaries and benefits to compensate Jewel for the risk it was taking.
• They could have signed precisely the agreement they did, without a reopen-er. After the low-price rival entered, Jewel could have laid off workers on account of lower sales.
*906 None of these is a clear winner for the employees, whose pickle is attributable not to faithless agents but to impending competition in the market. All options are potentially inferior for the employees to the solution the union negotiated. Bargains usually yield benefits for both sides. The union’s agents believed that this deal with Jewel was the best one available. It was at the union’s behest, not Jewel’s, that the reopener was oral rather than written. My colleagues concede all this but reply: “[I]t is not for us to decide whether the reopener would have ultimately benefited or harmed the Union membership.” 945 F.2d at 897. This is the wrong question, for we are not trying to determine who won or lost, but whether the 1983 reopener is sufficiently like the 1969 reopener that a jury could conclude that the parties’ practice did not require its presentation to the membership in writing. That question is indeed “not for us”; it is for the jury, and we do not learn why the jury could not find the two reopeners sufficiently similar. Suppose, though, that the distribution of benefits were the right question, and judges the right persons to answer it. Why may we not “decide whether the reopener would have ultimately benefited or harmed the union membership” when, on the same page, we make exactly that decision about the 1969 reopener? Either we may inquire into this subject or we may not. If we may, we should examine both reopeners; if we may not, neither. Finding benefits for the union in the 1969 reopener while disclaiming any power to search for benefits in the 1983 reopener is tellingly inconsistent.
Neither the parol evidence rule nor the union’s ratification requirement vitiates an oral reopener agreement. My colleagues fuse two unsatisfactory theories: “Taken individually ... none of plaintiffs’ arguments may be dispositive. Read together, however, they clearly illuminate the grave dangers posed by a backroom deal that is secretly negotiated between union officials and company management without the knowledge or consent of the union rank and file.” 945 F.2d at 899. Here the majority announces that the 1983 reopener was bad for the employees — despite its disclaimer at page 897 of authority to inquire into this question. This is special pleading. Reopeners are OK when they help the employees (page 896-97) but not when they help the employer (page 899). Such a disposition bears no resemblance to a rule of law. It is a thumb on the scales of justice.
. My colleagues think it significant that I have not found a case disapproving Gatliff by name. Guilty as charged. Plea in mitigation: As other courts take no notice of Gatliff, there are no citations to collect. The Supreme Court has never cited Gatliff. No court of appeals has cited it since 1964, when Lewis v. Owens, 338 F.2d 740, 742 (6th Cir.1964), pointedly declined to rely on it. Gatliff has been cited by courts of appeals a total of three times: Clothing Workers v. NLRB, 324 F.2d 228, 231 (2d Cir.1963); Lewis v. Lowry, 322 F.2d 453, 456 (4th Cir.1963) (in banc), and Owens. Lowry, like Owens , cites Gatliff without endorsing its conclusion that labor policy bars oral agreements that would be permitted by the parol evidence rule. Clothing Workers includes Gatliff in a string citation for the proposition established in Heinz: that the employer must sign a formal document if the union demands one. Gatliff’s conclusion that an employer may not withdraw from a unit after agreement has been reached is unimpeachable, given Bonanno and Retail Associates. The reason Gatliff gave has been abandoned, and the one that replaced it in Bonanno and Retail Associates does not assist the plaintiffs. A natural person missing for seven years is presumed dead. Justice Holmes remarked that "the reports of a given jurisdiction in the course of a generation take up pretty much the whole body of the law, and restate it from the present point of view.” The Path of the Law, 10 Harv.L.Rev. 457, 458 (1897), reprinted in Collected Legal Papers 167, 169 (1920). Gatliff has been missing for seven years plus a generation, which speaks more eloquently than express disapproval.
. The majority insists, 945 F.2d at 899 n. 4, that cases such as Manville "hold only that parol evidence may be considered to resolve ambiguities or fill in gaps in a written collective bargaining agreement.” This is the argument from triviality. More to come in the text below. Two observations here: First, none of the other courts of appeals believes that this matters. Second, I am baffled by the refrain that a reo-pener is specially grave because it "conflicts" with “fundamental” provisions of the agreement. With what provision does it conflict? If *903 the reopener had been written down we would not say that the contract had an internal contradiction. A reopener is an option to bargain in mid-term, and options do not "conflict” with the provisions that control in the event they are not exercised. At all events, what I find significant about these cases is not that their subjects are grave or trivial, but that these courts hold that oral agreements trump written ones. The writing in Manville contained a zipper clause — language specifying that the pact was fully integrated and waiving all right to bargain during the term of the agreement. The court of appeals held that such a clause may be ignored. That is why these cases represent the polar opposite to the majority’s position. They say that the parties cannot achieve an integrated collective bargaining agreement.
. In Houchens, for example, the union initially told the employer that “any contract proposal or recommendation would be subject to approval by the employees." 375 F.2d at 209 . When the negotiators reached agreement, the company said it would await the employees' approval. At this point the union’s negotiators insisted that the company implement the agreement without the employees’ approval; the company balked; the Board found that the company had committed an unfair labor practice; the court of appeals enforced the order, explaining: "The purpose of collective bargaining is to fix wages, hours and conditions of work by a trade agreement between the employer and his employees.... This can be done satisfactorily only if a bargaining agent is selected to represent all the employees with full power to speak in their behalf. The purpose of the statute would be largely frustrated if the results of bargaining must be submitted to a vote of the employees, with all of the misunderstandings and cross currents that would inevitably arise in an election of that sort.” 375 F.2d at 212 , quoting from Darlington Veneer, 236 F.2d at 88 . The reasons given in Houchens and Darlington Veneer sound remarkably like the reasons in my colleagues' opinion — except that the outcomes are diametrically opposed.