Full opinion text
NELSON, Circuit Judge: The States of Arizona, California, Oregon, and Washington appeal from the district court’s grant of summary judgment to the defendants in these consolidated antitrust actions. For the reasons stated below, we reverse the judgment of the district court and remand for further proceedings. I. INTRODUCTION Between June 1975 and August 1977, the plaintiffs filed their complaints in these actions, alleging several violations of the Sherman Act, 15 U.S.C. § 1 et seq. As developed during the subsequent pretrial proceedings, the plaintiffs’ allegations fall into three categories. First, the plaintiffs allege that the defendant oil companies conspired to raise or stabilize prices for refined oil products in violation of § 1 of the Sherman Act, 15 U.S.C. § 1. The plaintiffs assert that, in furtherance of this conspiracy, the defendants continually engaged in the mutual exchange of pricing and price-related information. Second, the plaintiffs allege that the defendants conspired to create, by various means, an artificial scarcity of crude oil and refined oil products in the western United States, in violation of §§ 1 & 2 of the Sherman Act, 15 U.S.C. §§ 1 & 2. Third, the plaintiffs allege that the defendants conspired not to compete in bidding on the plaintiffs’ annual bulk sale petroleum supply contracts, in violation of § 1 of the Sherman Act. After several years of extensive discovery, the plaintiffs filed in January 1983 a three volume pretrial brief (“Plaintiffs’ Initial Pretrial Brief” or “PIPB”), setting out their analysis of what the evidence would prove. The PIPB was supplemented on several occasions. In July 1983, the defendants moved for summary judgment, asserting that the evidence as summarized in the PIPB failed to raise a triable issue of antitrust conspiracy. After three days of oral argument on the summary judgment motions, the district court took the matter under submission. On November 25, 1986, the court filed an opinion and order granting the defendants’ summary judgment motion in its entirety. In re Coordinated, Pretrial Proceedings in Petroleum Prods. Antitrust Litig., 656 F.Supp. 1296 (C.D.Cal.1986) [hereinafter Petroleum Prods.]. The plaintiffs have timely appealed. Before turning to an analysis of the proper summary judgment standards and their application in this case, we think it appropriate and useful first to outline certain background facts concerning the industry’s structure as well as the nature of appellants’ theory concerning the operation of the alleged conspiracy. The appellees are major oil companies which, among other activities, produce crude oil, refine it into gasoline, and sell the gasoline to various distributors. During the time periods relevant to this appeal, these distributors fell into roughly four classes: (1) independent service station owners who operated franchises selling one particular brand of gasoline; (2) company-owned service stations run by company employees; (3) independent “jobbers” or brokers who resold gasoline to various service stations and other purchasers; and (4) governmental entities and others who purchased under bulk sales contracts. All parties agree that the lion’s share of appellees' gasoline that was sold at retail was sold by independent franchised service stations. As franchisees, these “independent” dealers were not free to purchase their supply of gasoline from any oil company at any time; as long as they remained franchisees they could only purchase from their particular franchisor. Each company sold gasoline to its franchised dealers at a price known as the “dealer tankwagon price.” In actuality, the official tankwagon prices were only occasionally changed; fluctuations in the cost of gasoline to franchised dealers were more frequently reflected in changes to the applicable discounts from the tankwagon price. These discounts were variously known as “temporary dealer assistance,” “dealer aid,” or simply “discounts.” The appellants argue that, as a consequence of this market structure, each oil company was effectively able to control the retail price at which its gasoline was sold. That is, the appellants claim that, although individual dealers “showed varying degrees of independence,” an oil company could essentially determine the retail price by setting the applicable discount from the tank-wagon price at which it sold gasoline to its franchised dealers. In the present actions, the appellants claim that the appellees have engaged in a conspiracy to raise and stabilize the retail price of gasoline at the pump. They do not claim, however, that the appellees engaged in a resale price maintenance scheme whereby each dealer was required to charge a predetermined price; indeed, they have expressly disavowed such a theory. Rather, the appellants claim that the appel-lees conspired to fix retail prices by coordinating dealer discounts from the tankwag-on price. The parties hotly contest on appeal whether the application of this theory is limited by the Supreme Court’s decision in Illinois Brick v. Illinois, 431 U.S. 720, 97 S.Ct. 2061, 52 L.Ed.2d 707 (1977), which held that indirect purchasers of goods whose price was fixed earlier in the stream of commerce may not maintain an antitrust damages action for overcharges passed on to them by those who purchase directly from the pricefixers. Although the district court had earlier ruled that Illinois Brick limited the relief available to the plaintiffs, see In re Coordinated Pretrial Proceedings in Petroleum Prods. Antitrust Litig., 497 F.Supp. 218, 225-27 (C.D.Cal.1980), aff'd on other grounds, 691 F.2d 1335 (9th Cir.1982), cert. denied, 464 U.S. 1068, 104 S.Ct. 972, 79 L.Ed.2d 211 (1984), it nonetheless recognized that Illinois Brick did not completely bar recovery. Accordingly, in ruling on the summary judgment motion, the district court recognized that, despite the ruling that Illinois Brick limited the relief that was available, the plaintiffs would still be able to receive some relief if they could establish that the defendants conspired to fix retail prices by fixing wholesale prices. See Petroleum Prods., 656 F.Supp. at 1299 (noting that plaintiffs’ theory was that the defendants conspirato-rily eliminated dealer discounts, thus “restoring” the official tankwagon prices, in order to fix retail prices). For purposes of this appeal, we operate with this same premise. Accordingly, we have no occasion to consider whether the district court was correct in its earlier ruling concerning the applicability of Illinois Brick. II. SUMMARY JUDGMENT STANDARDS We review de novo the district court’s grant of summary judgment to the defendants. Richards v. Neilsen Freight Lines, 810 F.2d 898, 902 (9th Cir.1987). Determining whether the grant of summary judgment was proper in this ease involves a careful application of the standards set down by the Supreme Court in Matsushita Elec. Indus. Co. v. Zenith Radio Corp., 475 U.S. 574, 106 S.Ct. 1348, 89 L.Ed.2d 538 (1986). Because the parties sharply disagree over the proper interpretation of Matsushita, and because the issue is crucial to a proper resolution of this case, we think it appropriate to review in some detail the nature of the standards established in Matsushita. A. The Matsushita Decision , In Matsushita, plaintiffs Zenith and National Union Electric Corp. (“Zenith”) sued several Japanese companies that manufacture consumer electronic products. Zenith alleged that the Japanese companies had engaged in a predatory pricing conspiracy aimed at driving American firms from the U.S. consumer electronic products market. After several years of discovery, the defendants moved for summary judgment. The district court granted the defendants’ motion, concluding that any inference of conspiracy was unreasonable. Zenith Radio Corp. v. Matsushita Elec. Indus. Co., 513 F.Supp. 1100 (E.D.Pa.1981). The Third Circuit reversed as to 21 of the 24 defendants, concluding that the plaintiffs had presented sufficient evidence to permit a reasonable factfinder to infer that the defendants had engaged in a predatory pricing conspiracy. In re Japanese Elec. Prods. Antitrust Litig., 723 F.2d 238, 304-11 (3d Cir.1983). The Supreme Court reversed this decision because it concluded that the evidence proffered by the plaintiffs did not provide any basis for concluding that the alleged predatory pricing was rational; in the Court’s words, the defendants simply “had no rational economic motive to conspire.” Matsushita, 475 U.S. at 596, 106 S.Ct. at 1361. This fact, combined with the evidence that the defendants’ conduct was “consistent with other, equally plausible [innocent] explanations” implied that the defendants’ conduct could “not give rise to an inference of conspiracy.” Id. at 596-97, 106 S.Ct. at 1361. Moreover, the Court stated that, even if the defendants had had a rational economic motive to conspire, summary judgment would still have been appropriate in light of the fact that “conduct that is as consistent with permissible competition as with illegal conspiracy does not, without more, support even an inference of conspiracy.” Id. at 597 n. 21, 106 S.Ct. at 1362 n. 21 (citing Monsanto Co. v. Spray-Rite Serv. Corp., 465 U.S. 752, 763-64, 104 S.Ct. 1464, 1470-71, 79 L.Ed.2d 775 (1984)). In such circumstances, the plaintiff must come forward with “sufficiently unambiguous” evidence “ ‘that tends to exclude the possibility’ ” that the defendants were acting lawfully. Id. 475 U.S. at 588, 597, 106 S.Ct. at 1356, 1362 (quoting Monsanto, 465 U.S. at 764, 104 S.Ct. at 1471). We do not take these latter comments as suggesting that a district court may grant summary judgment to antitrust defendants whenever the court concludes that inferences of conspiracy and inferences of innocent conduct are equally plausible. Allowing the district court to make that decision would lead to a dramatic judicial encroachment on the province of the jury. To read Matsushita as requiring judges to ask whether the circumstantial evidence is more “consistent” with the defendants’ theory than with the plaintiff’s theory would imply that the jury should be permitted to choose an inference of conspiracy only if the judge has first decided that he would himself draw that inference. This approach would essentially convert the judge into a thirteenth juror, who must be persuaded before an antitrust violation may be found. Indeed, under this interpretation of Mat-sushita, a trial judge would not even need to evaluate whether the inference of conspiracy meets a threshold standard of reasonableness, which is the usual standard for judging inferences on summary judgment. See Batchelor v. Oak Hill Medical Group, 870 F.2d 1446, 1447 (9th Cir.1989). Since any inference that the trial judge thinks is more plausible than its alternatives must necessarily be a reasonable one, the trial judge would only need to ask whether he thinks the inference of conspiracy is the more plausible one. This cannot be what the Supreme Court meant when it stated that “antitrust law limits the range of permissible inferences from ambiguous evidence in a § 1 case.” Matsushita, 475 U.S. at 588, 106 S.Ct. at 1356. The Court purported to limit the application of the traditional summary judgment rules in the antitrust context; it did not intend to abolish them and replace them with an entirely different set, one which raises troubling seventh amendment concerns. Cf. Standard Oil v. Arizona, 738 F.2d 1021 (9th Cir.1984) (holding, in a prior appeal in this case, that the seventh amendment guarantees a right to jury trial in antitrust cases), cert. denied, 469 U.S. 1132, 105 S.Ct. 815, 83 L.Ed.2d 807 (1985). Indeed, the Court noted that, within the limits imposed by antitrust law, it remained true that “ ‘[o]n summary judgment the inferences to be drawn from the underlying facts ... must be viewed in the light most favorable to the party opposing the motion.’ ” Matsushita, 475 U.S. at 587-88, 106 S.Ct. at 1356 (quoting United States v. Diebold, Inc., 369 U.S. 654, 655, 82 S.Ct. 993, 994, 8 L.Ed.2d 176 (1962)). Nor do we think that Matsushita and Monsanto can be read as authorizing a court to award summary judgment to antitrust defendants whenever the evidence is plausibly consistent with both inferences of conspiracy and inferences of innocent conduct. Such an approach would imply that circumstantial evidence alone would rarely be sufficient to withstand summary judgment in an antitrust conspiracy case. After all, circumstantial evidence is nearly always evidence that is plausibly consistent with competing inferences. See United States v. Henderson, 693 F.2d 1028, 1031 (11th Cir.1982) (“[CJircumstantial evidence is not testimony to the specific fact being asserted, but testimony to other facts and circumstances from which the jury may infer that the fact being asserted does or does not exist.”). Thus, such an interpretation of Matsushita would seem to be tantamount to requiring direct evidence of conspiracy. This cannot be what the Court meant in Matsushita. Since direct evidence will rarely be available, such a reading would seriously undercut the effectiveness of the antitrust laws. Furthermore, Monsanto itself clearly indicates that circumstantial evidence may be sufficiently unambiguous to survive summary judgment. Monsanto, 465 U.S. at 764, 104 S.Ct. at 1471 (plaintiff must “present direct or circumstantial evidence” that reasonably tends to prove that the defendants were engaged in a conspiracy rather than acting independently) (emphasis added); see also Harkins Amusement Enterprises, Inc. v. General Cinema Corp., 850 F.2d 477, 485 (9th Cir.1988) (circumstantial evidence found to be sufficiently unambiguous under Monsanto), cert. denied, 488 U.S. 1019, 109 S.Ct. 817, 102 L.Ed.2d 806 (1989). We think that the key to the proper interpretation of Matsushita lies in the Court’s emphasis on the dangers of permitting inferences from certain types of ambiguous evidence. In Matsushita, the Court was unwilling to permit an inference of predatory pricing in part because the Court was concerned about the inference’s possible anticompetitive side-effects. The Court noted that the plaintiffs attempted to prove an antitrust conspiracy “through evidence of rebates and other price-cutting activities.” 475 U.S. at 594, 106 S.Ct. at 1360. However, as the Court went on to observe, “[Cjutting prices in order to increase business often is the very essence of competition. Thus, mistaken inferences in cases such as this one are especially costly, because they chill the very conduct the anti-trust laws are designed to protect. See Monsanto, [465 U.S.] at 763-764 [104 S.Ct. at 1470-1471], ‘[W]e must be concerned lest a rule or precedent that authorizes a search for a particular type of undesirable pricing behavior end up by discouraging legitimate price competition.’ ” Id. (quoting Barry Wright Corp. v. ITT Grinnell Corp., 724 F.2d 227, 234 (1st Cir.1983)). Thus, Matsushita establishes that a trial judge should not permit an inference of antitrust conspiracy from circumstantial evidence where to do so would have the effect of deterring significant procompetitive conduct. We emphasize, however, that an antitrust defendant is not entitled to summary judgment simply by virtue of the fact that the plaintiff’s inference of conspiracy would have some deterrent effects; the effects must be significant. As the Court stated in Matsushita, the concern with avoiding deterrent effects “must be balanced against the desire that illegal conspiracies be identified and punished.” 475 U.S. at 594, 106 S.Ct. at 1360. In short, the trial court must consider whether, on the evidence presented, the protection of innocent independent conduct outweighs the costs associated with the potential decrease in strict antitrust enforcement. If it does, then the plaintiff must come forward with additional, “sufficiently unambiguous” evidence that does not have these undesirable deterrent effects. Id. at 597-98 & n. 21, 106 S.Ct. at 1361-62 & n. 21. This reading of Matsushita is further supported by examining the Court’s decision in Monsanto Co. v. Spray-Rite Serv. Corp., 465 U.S. 752, 104 S.Ct. 1464, 79 L.Ed.2d 775 (1984), upon which the Matsu-shita, Court relied heavily. In Monsanto, plaintiff Spray-Rite alleged that Monsanto had illegally conspired with its other distributors in deciding to terminate Spray-Rite’s distributorship. Among the items of evidence introduced in support of this claim was the fact that Monsanto terminated Spray-Rite after receiving complaints from its other distributors about Spray-Rite’s price cutting practices. Monsanto, on the other hand, contended that it had acted independently in deciding to terminate Spray-Rite. In analyzing the proper standard of proof for such a claim, the Supreme Court began by noting that, while it is per se illegal for a manufacturer to get together with its distributors and agree on prices, see Dr. Miles Medical Co. v. John D. Park & Sons Co., 220 U.S. 373, 404-09, 31 S.Ct. 376, 383-85, 55 L.Ed. 502 (1911), it is nonetheless permissible for a manufacturer to announce its retail prices in advance and then to terminate those distributors who fail to comply, see United States v. Colgate & Co., 250 U.S. 300, 307, 39 S.Ct. 465, 468, 63 L.Ed. 992 (1919). See Monsanto, 465 U.S. at 761, 104 S.Ct. at 1469. In light of these cases, the Court concluded that it could not permit an inference of conspiracy to be drawn from the mere fact that a distributor had been terminated after the manufacturer had received complaints from other distributors about price cutting. The Court reasoned that permitting such an inference would effectively preclude a manufacturer from exercising its independent termination rights under Colgate once it received any complaints from its distributors. In the Court’s words: Permitting an agreement to be inferred merely from the existence of complaints, or even from the fact that termination came about “in response to” complaints, could deter or penalize perfectly legitimate conduct_ To bar a manufacturer from acting solely because the information upon which it acts originated as a price complaint would create an irrational dislocation in the market. In sum, “[t]o permit the inference of concerted action on the basis of receiving complaints alone and thus to expose the defendant to treble damage liability would ... inhibit management’s exercise of its independent business judgment....” Monsanto, 465 U.S. at 763-64, 104 S.Ct. at 1470-71 (quoting Edward J. Sweeney & Sons, Inc. v. Texaco, Inc., 637 F.2d 105, 111 n. 2 (3d Cir.1980), cert. denied, 451 U.S. 911, 101 S.Ct. 1981, 68 L.Ed.2d 300 (1981)) (other citations omitted). Monsanto thus clearly illustrates that courts should be careful not to permit inferences of antitrust conspiracy when to do so would create a significant irrational dislocation in the market or would result in significant anticompetitive effects. In summary, we conclude that where an antitrust plaintiff relies entirely upon circumstantial evidence of conspiracy, a defendant will be entitled to summary judgment if it can be shown that (1) the defendant’s conduct is consistent with other plausible explanations, and (2) permitting an inference of conspiracy would pose a significant deterrent to beneficial procom-petitive behavior. Once the defendant has made such a showing, the plaintiff must come forward with other evidence that is sufficiently unambiguous and tends to exclude the possibility that the defendant acted lawfully. B. Direct v. Circumstantial Evidence Before turning to the specific issues of this case, we must add one further comment with regard to the applicability of Matsushita. As the above analysis demonstrates, the concerns highlighted in Matsu-shita and Monsanto arise only in the context of whether to permit inferences from circumstantial evidence. Accordingly, the Matsushita standards do not apply when the plaintiff has offered direct evidence of conspiracy. This important limitation on the applicability of Matsushita was emphasized in our decision in McLaughlin v. Liu, 849 F.2d 1205 (9th Cir.1988). In Liu, the Secretary of Labor argued that summary judgment was properly granted, under Matsushita, on the grounds that the defendant’s sworn affidavit—which, if believed, would have provided a complete defense as to part of the Secretary’s claim—was “implausible.” Id. at 1207. We emphatically rejected this argument, noting that in Matsushita, “the Court was not speaking of direct evidence, but of circumstantial evidence.” Id. We noted the important difference between circumstantial evidence, in which a party asks that certain “inferences be drawn in his favor” and direct evidence, where, in order to defeat a request for summary judgment, the nonmovant need only ask that his evidence “be taken as true.” Id. at 1208. We emphasized that the cases of both the Supreme Court and this court “have honored the difference between weighing direct evidence and refusing to draw unreasonable inferences from circumstantial evidence.” Id. We summarized our cases as indicating that Matsushita only applies “where the non-movant reliefs] on inferences from circumstantial evidence.” Id. We then went on to quote extensively from T.W. Electrical Services, Inc. v. Pacific Electrical Contractors Assoc., 809 F.2d 626 (9th Cir.1987), highlighting that the Matsushita inquiry was appropriate only “ ‘[w]here there is no direct evidence of a conspiracy.’ ” Id. at 1209 (quoting T.W. Elec., 809 F.2d at 632) (emphasis added by Liu). See also Christofferson Dairy, Inc. v. MMM Sales, Inc., 849 F.2d 1168, 1172 n. 4 (9th Cir.1988) (standards spelled out in T.W. Elec, apply only “[i]f there is no direct evidence of a conspiracy”). Thus, in applying Matsushita, a court must consider the nature of the evidence that the plaintiffs have offered with respect to each element of the cause of action. If the plaintiffs rely exclusively on circumstantial evidence in order to establish at least one element of their cause of action, then the court must proceed to analyze, under Matsushita, whether the inferences which the plaintiffs seek to draw from the indirect evidence are reasonable and permissible under the governing substantive law. T. W. Elec., 809 F.2d at 631; see also United Steelworkers of America v. Phelps Dodge Corp., 865 F.2d 1539, 1542 (9th Cir.) (en banc), cert. denied, — U.S. -, 110 S.Ct. 51, 107 L.Ed.2d 20 (1989). In the antitrust field, this inquiry will be guided by the principles outlined earlier. If, however, the plaintiffs offer direct evidence to support each element, then summary judgment must be denied. III. APPELLANTS’ PRICE-FIXING CLAIMS In their briefs to this court, the appellants have identified three classes of evidence that they claim create a genuine issue of material fact as to whether the appellees were engaged in a conspiracy to fix or stabilize prices. For purposes of analysis, we will first examine these classes of evidence successively; having done so, we will then be in a better position to evaluate the probative value of the entirety of the appellants’ evidence. A. Pricing Pattern Evidence The first set of evidence that the appellants claim indicates the existence of a conspiracy is a set of analyses of the appellees’ pricing patterns. This analysis was performed by appellants’ expert Keith Leffler. The raw data which Mr. Leffler analyzed were derived, in large measure, from pricing information collected by Lundberg Surveys, Inc. During the period from 1968 to 1978, the Lundberg firm collected, on a weekly basis, retail price information from individual stations in 14 different cities in Arizona, California, Oregon, and Washington. Mr. Leffler analyzed this data for the largest city in each of the four states. His analysis of this weekly price data indicates that, between 1968 and August 1972, the average retail prices of stations selling ap-pellees’ gasoline in the Los Angeles, Phoenix, Portland, and Seattle markets generally followed a “sawtooth” pattern. That is, the average prices would generally decline over a period of time ranging from approximately three to eleven weeks, and then the prices would return to roughly their original, higher levels over approximately a one to four week period. Both the price increases and decreases from week to week were often quite sharp; but the increases were often especially sharp. The parties do not seriously dispute that the sharp increases in retail prices were generally the consequence of the withdrawal of dealer discounts, thus resulting in the “restoration” of the “normal” higher dealer tankwagon prices. Accordingly, we cannot accept the district court’s conclusion that the cyclical pattern of retail prices does not suggest that wholesale prices followed a similar pattern. See Petroleum Prods., 656 F.Supp. at 1300 (“the cycle of retail price changes reflected by the charts does not establish that the defendants’ wholesale prices followed the same pattern”). Indeed, we note that the data presented by the appellees themselves concerning their wholesale prices indicates that the pattern of these prices was a cyclical one of sharp increases followed by gradual declines. The key disagreement between the parties concerns the significance of this pattern. The appellees argue that the restorations and the sawtooth pattern of prices do not indicate conspiracy because the same pattern would be produced by individual companies each deciding to follow the others’ price lead. In short, the appellees essentially assert that their pricing behavior was interdependent rather than conspiratorial. The appellants disagree, arguing that these price cycles were not the result of any sort of independent action, but are rather an indication of collusion. The appellants assert that a conspiracy may be inferred from this pricing data because the price increases were too large and too risky to have occurred in a competitive market. According to the appellants, any company that raised its prices this sharply would so risk losing business that it would never take such a step without some advance assurance that others would follow. Therefore, argue the appellants, the fact that such price increases were made is an indication of a prior agreement to raise prices. In applying Matsushita, our first task is to decide whether the appellees’ claim of interdependent pricing constitutes a sufficiently plausible independent explanation for their pricing behavior. The appellants’ key argument on this score is that the price increases were so risky that it is implausible to believe that any firm would have undertaken them without some advance agreement from competitors. We note that such an argument may often be valid when the relevant market is highly uncon-centrated or where the increase cannot be reversed easily or readily without substantial loss of goodwill. See VI P. Areeda, Antitrust Law ¶ 1425d at 150-53, H 1431a at 183-84 (1986). In a highly unconcentrat-ed market, any firm that attempted a unilateral supracompetitive price increase would quickly lose business to its many other competitors. Interdependent pricing is not likely to occur because each firm realizes that the possibility of undetected “shading” of the interdependent price by a single firm would quickly lead to a return to the original market price. Turner, The Definition of Agreement Under the Sherman Act: Conscious Parallelism and Refusals to Deal, 75 Harv.L.Rev. 655, 659-60 (1962); see also VI P. Areeda, supra, at ¶11429a. In such a market, any single firm can increase its output without substantially affecting the market price or the market shares of its competitors. This fact substantially increases the odds that shading by an individual firm would either go undetected or be ignored, and the consequent spread of such discounting would rapidly bring the overall price back down. Turner, 75 Harv.L.Rev. at 660; VI P. Areeda, supra, at ¶ 1429a. Therefore, without some sort of prior agreement prohibiting such shading, the temptation for any individual firm would be almost irresistible. Turner, 75 Harv.L.Rev. at 660. Accordingly, in highly unconcentrated markets, a claim that price increases were the result of sequential interdependent price increases is often likely to prove implausible. However, as the number of firms in a market declines, the possibilities for interdependent pricing increase substantially. In determining whether to follow a unilateral price increase by a competitor, a firm in a relatively concentrated market will recognize that, because its pricing and output decisions have an effect on market conditions and will generally be watched by its competitors, there is less likelihood that any shading would go undetected or would be ignored. The firm thus knows that if it fails to follow the price lead, the leading firm will quickly reduce its prices back to their earlier level. VI P. Areeda, supra, at 111429b. On the other hand, the firm may recognize that the higher price is one that would produce higher profits. It may therefore decide to follow the price increase, knowing that the other firms will likely see things the same way and that, at any rate, any subsequent downward movement in prices would likely be detected before there was any substantial loss of market share. Id.; Turner, 75 Harv.L.Rev. at 661-62. There has been a considerable debate in the literature over whether interdependent pricing is an inevitable or inherent feature of certain types of concentrated markets. Compare VI P. Areeda, supra, at 11¶ 1430-32 (inherent); Sullivan, The Viability of the Current Law on Horizontal Restraints, 75 Cal.L.Rev. 835, 858-59 (1987); Turner, 75 Harv.L.Rev. at 665-66 with R. Posner, Antitrust Law: An Economic Perspective 42-47 (1976) (not inherent); Posner, Oligopoly and the Antitrust Laws: A Suggested Approach, 21 Stan.L. Rev. 1562, 1566-75 (1969). We need not take any side in this debate. Regardless of whether such conduct is an inevitable consequence of a more concentrated market structure, we conclude that it is at least plausible to assert that such pricing behavior can occur when the number of significant rivals in a market is sufficiently small. We do not think that the number of significant firms in the gasoline markets involved in this case is so large as to render the possibility of sequential interdependent price increases wholly implausible. Although these markets would probably not be considered highly concentrated in Her-findahl-Hirschmann terms, that does not imply that the appellees’ claim of interdependent pricing is implausible. VI P. Aree-da, supra, ¶ 1431a at 183 (noting that the Herfindahl-Hirschmann Index “cannot capture the full range of factors bearing upon the existence or strength of recognized interdependence”). As Professor Areeda has noted, “All one can say with assurance is that such interdependence and price coordination are implausible when there are ‘many’ ‘significant’ rivals (more than a dozen?) but more likely as the number of firms decreases.” Id. at 184. The gasoline markets under discussion here appear to lie in that gray area where, although not altogether inevitable, interdependence is distinctly possible. See F. Scherer, Industrial Market Structure and Economic Performance 288 (2d ed. 1980) (discussing empirical evidence of interdependent pricing in the gasoline industry). Indeed, the appellants have not pointed to any evidence indicating that the number of firms in these markets is too large for interdependence to occur. Nonetheless, we noted earlier that, even in highly concentrated markets, a unilateral price hike might be too risky to make without advance agreement if the increase could not be readily reversed without a significant loss of goodwill. On the other hand, where a unilateral price increase can be withdrawn quickly, a firm may very well decide to test the waters without any sort of advance commitment from its competitors as to whether they will follow. In this regard, we note that the uncontested record evidence indicates that unilateral price increases in these markets were quickly reversible. Furthermore, the evidence indicates that, although unilateral restorations could often entail significant costs if unsuccessful, the costs do not appear to have been so prohibitive as to suggest that the possibility of such unilateral action was implausible. We recognize that such interdependent pricing may often produce economic consequences that are comparable to those of classic cartels. Nonetheless, proof of such pricing, standing alone, is generally considered insufficient to establish a violation of the Sherman Act. See Wilcox v. First Interstate Bank of Oregon, N.A., 815 F.2d 522, 526 (9th Cir.1977) (“[T]he fact that competitors may see proper, in the exercise of their own judgment, to follow the prices of another manufacturer, does not establish any suppression of competition or show any sinister domination.”) (quoting United States v. International Harvester Co., 274 U.S. 693, 708-09, 47 S.Ct. 748, 754, 71 L.Ed. 1302 (1927)); see also Clamp-All Corp. v. Cast Iron Soil Pipe Inst., 851 F.2d 478, 484 (1st Cir.1988), cert. denied, 488 U.S. 1007, 109 S.Ct. 789, 102 L.Ed.2d 780 (1989); Apex Oil Co. v. DiMauro, 822 F.2d 246, 253-54 (2d Cir.), cert. denied, 484 U.S. 977, 108 S.Ct. 489, 98 L.Ed.2d 487 (1987); E.I. DuPont de Nemours & Co. v. F.T.C., 729 F.2d 128, 139 (2d Cir.1984) (“The mere existence of an oligop-olistic market structure in which a small group of manufacturers engage in consciously parallel pricing of an identical product does not violate the antitrust laws.”); VI P. Areeda, supra, at ¶ 1432; Turner, 75 Harv.L.Rev. at 669-72. Additional proof beyond mere parallel pricing usually is required. Our discussion of Matsushita highlights the wisdom of this rule. In applying the second prong of the Matsushita test, we must ask whether, given the possibly interdependent nature of the pricing behavior observed in this case, permitting a § 1 violation to be found solely on such evidence would have the effect of deterring important legitimate conduct. We conclude that, if we were required to decide based on the price cycle evidence standing alone, we would find that it would not be enough to survive summary judgment. To permit an antitrust violation to be based on the sawtooth price pattern in this case, without more, would require a company making wholly independent pricing decisions to consider that the possible responses of its competitors might render it liable for treble damages. Similarly, following another company’s price increase might very well provide the evidence that a disgruntled customer would need to get to a jury in a treble damage antitrust suit. It thus appears that permitting an inference of conspiracy from the parallel pricing evidence alone would result in an anticompetitive dislocation by distorting independent pricing decisions. We also cannot agree that, without more, the conspiratorial nature of the price restorations is indicated by the fact that the price hikes were “too high.” Were we to permit such a theory, few pricing decisions would be immune from antitrust scrutiny. Any unhappy buyer in a market experiencing a general increase in prices could bring an antitrust action against the sellers, arguing that the conspiratorial nature of the price increase is indicated by the fact that the increase was “too high.” The court would then be required to receive evidence as to the wisdom of, and necessity for, the price increase, and to judge what constitutes a “fair” price. The federal courts generally are unsuited to act as rate-setting commissions. See VI P. Areeda ¶ 1432 at 203. Accordingly, the appellants must rely on additional evidence beyond mere price parallelism in order to avoid summary judgment. B. Price Data Dissemination Evidence The appellants point to two additional classes of evidence which they claim indicate the existence of a conspiracy to fix or stabilize prices. The first of these two sets consists of evidence concerning various practices whereby the appellees disseminated information concerning their wholesale and retail prices. The appellants contend that, not only were these price dissemination practices themselves conspiratorily and unlawfully adopted, but that they were adopted pursuant to an agreement or understanding to engage in concerted restorations. The appellees argue that these practices were both lawful and independently adopted and that they are not probative of a conspiracy to stabilize prices. 1. Press releases The appellants presented evidence indicating that at various times each of the defendants except ARCO, Union, and Gulf engaged in the practice of publicly announcing, in press releases, their decisions to withdraw dealer assistance and to restore tankwagon prices. Indeed, the evidence indicates that such price increases were sometimes announced in advance of their effective date. Thus, for example, Mobil announced on March 26, 1970 that it would withdraw dealer aid effective March 30. The appellees argue that no inference of conspiracy may be drawn from such evidence because they claim that the publication of this information was “perfectly lawful” and an “ordinary business practice.” We conclude, however, that such an inference is both reasonable and permissible under Matsushita. In their depositions in this case, several officers of the appellee oil companies were questioned concerning the business reasons for publicly announcing changes in tank-wagon prices and in the levels of dealer assistance. Their virtually uniform response was that it was done for the purpose of quickly informing competitors of the price change, in the express hope that these competitors would follow the move and restore their prices. Thus, the appellees’ officers’ own testimony indicates that there was essentially no purpose for publicly announcing tank-wagon prices and dealer discount information other than to facilitate either interdependent or plainly collusive price coordination. Announcing such price increases publicly reduced the likelihood that the increase would fail to be detected or that it would be detected only after the price leader had been “hung out to dry” in the market for several days. Without a press release, a withdrawal of dealer support might not be readily detected because the retail prices of individual branded gas stations varied considerably. As one Standard Oil official put it: “[S]treet pricing [was] all over the lot; so to us it was important that we set forth clearly and exactly what we had done so there would be no misunderstanding of it.” Indeed, ARCO’s failure publicly to announce its restoration attempts to the trade press appears to have been responsible for the failure of its January 1970 restoration. An internal ARCO memorandum explained that “[t]he reason this partial restoration did not work is that our competitors obviously thought our dealers were merely overpricing and our constructive efforts went completely unnoticed at the market place.” To avoid this problem, the memorandum recommended that if ARCO decided to lead a restoration in the future, it should be “telegraphed to all news media as far as possible in advance as Legal will approve.” It was important for the appellees to reduce the uncertainties concerning restorations because the costs associated with leading a restoration were significant, and they increased with each day that a move was not detected or was not followed. An internal ARCO memorandum noted that its unsuccessful restoration attempts had been “costly to us, not only in terms of volume, but more importantly in expense dollars required to recoup and serious adverse reaction with our customers and in our dealer organization.” Another ARCO document noted that “[p]ast experience has taught us how disastrous it can be for us to be the leader in an upward move.” Even with public announcement, the costs of leading a restoration were substantial; Standard Oil’s internal analysis of the effect of three restorations it led indicated an average drop in sales volume of over 19% within one week. The same Standard Oil document blamed, in part, the company’s practice of leading restorations for causing a significant decline in Standard’s overall market share. An ARCO document echoed this view, noting that Standard’s and Shell’s leading of restorations had led to their “disastrous” loss in market share. The evidence presented by the appellants thus indicates that the publication of wholesale price increases was intended to make, and had the effect of making, restorations more effective by ensuring that competitors could quickly learn of, and respond to, any withdrawal of dealer aid. The appellees’ actions in announcing such information made the market more receptive to price coordination than it otherwise would have been. Although we concluded earlier that mere proof of interdependent pricing, standing alone, may not serve as proof of an antitrust violation, we believe that the evidence concerning the purpose and effect of price announcements, when considered together with the evidence concerning the parallel pattern of price restorations, is sufficient to support a reasonable and permissible inference of an agreement, whether express or tacit, to raise or stabilize prices. The inference is reasonable because at the very least, a jury could conclude that the appellees implicitly agreed to engage in practices that they knew and hoped would lead to greater price coordination, with the result that when a price leader wished to initiate a restoration, it could effectively solicit tacit agreement. See United States v. Container Corp., 393 U.S. 333, 336-37, 89 S.Ct. 510, 512, 21 L.Ed.2d 526 (1969) (informal agreement to provide price information may, under appropriate market conditions, constitute circumstantial evidence of an agreement to stabilize prices); King & King Enterprises v. Champlin Petroleum Co., 657 F.2d 1147, 1152 (10th Cir.1981) (Container established that exchange of price information may serve as basis for inferring price fixing where effect of such exchange is to stabilize prices), cert. denied, 454 U.S. 1164, 102 S.Ct. 1038, 71 L.Ed.2d 320 (1982); Penne v. Greater Minneapolis Area Bd. of Realtors, 604 F.2d 1143, 1148-49 (8th Cir.1979) (summary judgment was erroneously granted where evidence indicated possible connection between price information exchanges and alleged conspiracy to fix brokerage fees); see also Posner, Information and Antitrust: Reflections on the Gypsum and Engineers Decisions, 67 Geo. L.J. 1187, 1199 (1979) (“[I]f the effect of the information exchange were to raise the level [of] prices, one could infer that the motive was price fixing.”); id. at 1203 (a jury should be allowed “to consider exchanges of information, other communications among the parties to an alleged conspiracy, and such other relevant circumstances as the effect on the price level ..., as circumstantial evidence of alleged price fixing.”). We reject the appellees’ circular argument that the publication of dealer discount and price information cannot support an inference of conspiracy because the information was “publicly available.” The information was publicly available only because the appellees chose to make it so; without such dissemination, price moves were not always readily and easily detected. Furthermore, we agree with Professor (now Judge) Posner that “the form of the exchange — whether through a trade association, through private exchange as in Container, or through public announcements of price changes — should not be determinative of its legality.” R. Posner, Antitrust Law: An Economic Perspective 146 (1976). The fact that it is feasible for the appellees to communicate the necessary price information through press releases does not “immunize the exchange of price information from legal sanction [where] the conditions of the market suggest that the exchange promotes collusive rather than competitive pricing.” Id. at 147. We do not believe that permitting such an inference poses any problem under Mat-sushita. The district court concluded that the inference could not be permitted, reasoning that the hope of imitation was a proper business purpose for the announcements, and that the publication of such information was in the legitimate individual interest of each company. Petroleum Prods., 656 F.Supp. at 1304-05. We agree that it is plausible to contend that independent considerations of self-interest would have led a company to choose to publish a price increase that it had decided to make. That does not settle the inquiry, however; the question remains whether permitting an inference of conspiracy from the fact of such publication would significantly deter important legitimate conduct. Given the market conditions present in this case, we conclude that it would not. It is important to recognize that, given the system of branded franchising, the tankwagon prices or dealer discounts are not of immediate significance to anyone other than the oil companies and their franchised dealers. Retail purchasers do not care what the dealer paid for the oil; they are concerned only with the price at the pump. Moreover, the public dissemination of information concerning dealer tankwag-ons and discounts could hardly be described as fostering the efficient or rational operation of the market inasmuch as there simply was no wholesale market to rationalize; the branded dealers were not free to shop around for their oil. Accordingly, the district court’s reliance on Maple Flooring Mfrs. Ass’n v. United States, 268 U.S. 563, 582-84, 45 S.Ct. 578, 584-85, 69 L.Ed. 1093 (1925) (noting that certain types of price information exchanges may help to “avoid the waste which inevitably attends the unintelligent conduct of economic enterprise”), quoted in Petroleum Prods., 656 F.Supp. at 1304, is wholly inapposite. The uncontested record evidence indicates that the dealers were individually notified concerning any changes in the tank-wagon price or in the level of dealer discount. In light of this fact, it appears that the public dissemination of such information served little purpose other than to facilitate interdependent or collusive price coordination. Under these circumstances, there is no significant probability that permitting the proffered inference would deter important legitimate conduct. The Mat-sushita balance is particularly one-sided in this case. Nothing in our earlier discussion concerning mere proof of parallel pricing is to the contrary. Simple interdependent pricing does not violate the Sherman Act, not because it is desirable (it is not), but because permitting proof of conspiracy solely on the basis of price parallelism is undesirable. The appellants, however, have offered evidence indicating that the appellees in this case did more than simply price interdependently. A jury could conclude that the oil companies agreed, either implicitly or explicitly, to create market conditions that would facilitate tacit or express price coordination. The evidence would support a conclusion that the appellees strove to make such coordination possible where it otherwise might not have been. To paraphrase Professor Areeda, “One may reluctantly tolerate interdependent pricing behavior as such and still condemn [those agreements involving] practices which unjustifiably facilitate interdependent pricing and which can be readily identified and enjoined.” P. Areeda, Antitrust Analysis ¶ 325, at 381 (3d ed. 1981). We think that a jury should decide whether the appellees’ actions were undertaken pursuant to an agreement or understanding concerning prices or whether their activities were simply noncollusive and independent. 2. Posting of prices The appellants also presented evidence indicating that several of the appellees publicly posted their dealer tankwagon prices and any applicable dealer discounts. Specifically, the appellants produced evidence indicating that Mobil, Standard Oil, and Exxon posted such information, for public inspection, at either their company headquarters, division offices, or bulk plants. The appellants also presented testimony from an ARCO executive to the effect that ARCO “probably” posted its tankwagon prices at its bulk plants. In addition, there is evidence in the record indicating that ARCO, among others, checked these postings. For reasons comparable to those discussed above, we conclude that the practices of posting and checking this detailed information supports a reasonable and permissible inference of an agreement or understanding concerning prices. The ap-pellees’ assertedly independent explanation for such posting was that, as one Standard Oil official put it: It was our desire to be open and straightforward in the matter of our product pricing and to make available for any customer the ability to look and see what our posted price was.... I guess one might say that it was for the logic of being open and above board on what our prices were to our customers. We had nothing to conceal and it was just a mechanism to accomodate that openness. Another Standard official testified that the posting was done “[f]or the customer’s benefit, to inform the customer what the applicable prices were.” A Mobil official testified that the purpose of the posting was to allow any individual dealer to come in and check “that he wasn’t being discriminated against.” It is uncontested, however, that any changes in tankwagon prices or dealer discounts were directly reported to the dealers. Other testimony indicates that the purpose and effect of the posting was to allow competitors to learn quickly of any withdrawal of dealer support. The following testimony from a Standard Oil official is illustrative: Q. ... Was there any business reason why the temporary dealer assistance that was being granted in the trade zone ... should have been published or made available to anybody other than the dealers? A. Yes, I think a practical business reason. If we had raised our prices to our dealers, and our dealers had raised it on the street generally, I think particularly our dealers ... may have been as much as one, two, three, four cents — they were all over the lot. If a competitor saw this, he might wonder, “Is this a dealer movement, or has Standard Oil Company raised their prices to cause this reaction in the marketplace?” ... Q. ... The reason you are giving, then, is that you wanted the competition to know what the price move was. A. Exactly. Given that the evidence thus indicates that the purpose and effect of the public posting was the same as for the press releases, we similarly conclude that a reasonable inference of agreement may be drawn from this evidence. We also perceive no Matsushita problem in permitting such an inference. To the extent that the appellees assert business motivations comparable to those discussed in Section III — B—1, our analysis is the same. Furthermore, given that the appel-lees’ asserted purpose of assuring dealers that there was no discrimination could easily be achieved through other means, we have little difficulty concluding that permitting such an inference will not deter any significant or valuable independent conduct. Lastly, we perceive an additional reason why the posting of this information may serve as the basis for an inference of conspiracy. The record indicates that the information posted was unusually detailed, listing tankwagons and dealer discounts for each individual price zone. Such detailed information would have the effect of revealing any gradual shading of wholesale prices that might occur in a particular zone. The record indicates that the spread of such shading was one factor that contributed to the gradual collapse of price restorations. By posting and checking such information, several of the appellees created a mechanism whereby shading might be discouraged or detected. In light of the fact that disclosure of this sort of sensitive price information might be considered contrary to a firm’s self-interest, a jury may reasonably conclude that it was the intention and common understanding of these companies to discourage such shading, thus facilitating and possibly lengthening restoration periods. C. Evidence of Competitor Contacts The third set of evidence which the appellants claim supports an inference of an agreement consists of evidence concerning alleged contacts between the appellees. 1. Pre-Container contacts The appellants assert that, prior to the Supreme Court’s decision in Container, the appellees all regularly provided one another with information concerning dealer tank-wagons and discounts. The record contains evidence indicating that the appellees did in fact engage in direct communication concerning price levels and dealer support. For example, Agnar Nerheim, a marketing official with Standard Oil, testified that between 1956 and 1958 he telephoned competitors to determine what retail price level they were supporting in the market, and that he sometimes asked what was the specific amount of dealer assistance being given. He also testified that he received similar calls from competitors. When asked about which specific companies were involved, Nerheim testified that he had such conversations with individuals from Union, Richfield (now ARCO), Texaco, General (now Mobil) and Shell. Nerheim testified that such price verifications continued after he was transfered to Standard’s Los Angeles office. Nerheim stated that he was transfered to Los Ange-les because most of the major oil companies were headquartered there, and it would therefore be easier for him to contact competitors. In addition to verifying prices, Nerheim stated that he also spoke with competitors in order to obtain “market intelligence,” including information concerning wholesale and retail matters. Among the subjects discussed were the price wars that were taking place in Los Angeles. Nerheim testified that, during this time period, he had discussions concerning such market intelligence with officers from ARCO, Humble (now Exxon), Texaco, Mobil, and Gulf. Most of these contacts were face-to-face visits rather than phone calls. Nerheim stated that he kept no records of these competitor contacts, and that he had been specifically instructed by his superior not to put down on his expense reports the names of any competitors whom he took to lunch. Ner-heim stated that such contacts continued until 1967. Moreover, Robert Erhard, an official with the Carter Oil Company (later acquired by Humble, now Exxon) testified that, sometime between 1959 and 1961, Ag-nar Nerheim came to his office in Seattle and explained to him how they could converse about pricing without there being any telephone record. According to Erhard, Nerheim explained that the local Standard office in Seattle could reach him on a company WATS line, and that calls could therefore be made through the local office without appearing on company long-distance records. In light of Erhard’s and Nerheim’s testimony, it seems clear that the record supports an inference that Ner-heim engaged in secret conversations with competitors concerning price levels during the 1960s. Other testimony indicates that other Standard employees engaged in similar conversations with competitors during this time period. The appellees do not appear to contest the appellants’ suggestion that such contacts took place prior to the decision in Container. Instead, they raise two arguments as to why the evidence should not be considered as supporting an inference of conspiracy. The appellees first argue that the testimony is “ancient” and is too “stale” to have any probative value. For several reasons, we must reject this contention. First, although all of this pre-CW-tainer evidence falls outside the limitations period applicable in this case, the district court dismissed this action without ruling on the plaintiffs’ claim that the statute of limitations had been equitably tolled by the defendants’ alleged fraudulent concealment of their activities. If the statute was tolled, then we would have little difficulty concluding that the appellants’ evidence of secret direct price exchanges among the appellees supports an inference of an agreement to fix or stabilize prices. Accordingly, on remand the district court should consider whether the statute of limitations has been tolled; if it has, then the plaintiffs may use this evidence of competitor contacts in order to establish the existence of a conspiracy during the pre-Ccm-tainer period. Second, we think that the pre-Container evidence is relevant to establishing the ap-pellees’ intent and motive in publicizing, through press releases and posting, their dealer tankwagons, dealer discounts, and supported retail prices. The appellants have produced evidence indicating that these methods of publication were generally meant to serve the same purposes as the direct competitor contacts that had preceded Container, and, indeed, that they were used as a substitute for such contacts in the post-Container period. Thus, for example, Nerheim testified that, by 1967 (two years before Container) Standard had largely abandoned the practice of verifying its prices directly to its competitors, relying instead on its system of posting such information publicly, a practice which it had begun earlier in the decade. Nerheim stated that, before he left Los Angeles in 1967, such information was publicly posted, and competitors therefore no longer needed to verify prices with him verbally; they would just check the listings and then drop by his office to say hello. Occasionally during this period, however, Nerheim still responded to phone calls requesting information on specific wholesale dealer allowances. Similarly, C.R. Jones, an official with Mobil, testified that Mobil used several different methods to obtain information from competitors concerning dealer aid, but that the use of direct horizontal communications was dropped after the decision in Container was announced. Thereafter, Mobil relied largely on trade publications, contacts with competitive service station dealers, and checking postings in order to obtain information concerning competitors’ levels of dealer aid. Although various Mobil bulk plants had been posting in various degrees for some time, after Container Mobil instituted a uniform practice of posting at all of its plants across the United States. As noted earlier, this uniform system of posting included publicizing Mobil’s tankwag-ons and dealer discounts for specific trade zones. Mr. Jones specifically testified that this shift in methods did not represent a change in overall business practices: ... The decisions in the Container case changed our communications. But we did not intend that it change any business procedures or practices that we had historically been operating under, that we felt were not necessary to change. We were attempting to maintain our conduct on a — in a legal manner, and certainly not attempting to say we have to change our way of life because of a decision in a court case. We think that this evidence concerning the shift to indirect methods of informing competitors of wholesale price levels provides some support for the appellants’ claim that the post-Container price publications were made pursuant to a common understanding or agreement. The pre-Container activities, which involved verifications of dealer support levels upon personal request, fit squarely within the scope of Container. The appellants’ evidence suggesting that the appellees shifted methods of communication, without abandoning the goal of giving and receiving such information, supports an inference that the p