This is an appeal from a jury verdict imposing civil liability for alleged predatory pricing in violation of the antitrust laws. More specifically, appellants Rule Industries, Inc. (“Rule”) and Tie Down Engineering, Inc. (“Tie Down”), defendants below, appeal the district court’s denial of their motions for judgment notwithstanding the verdict on claims by U.S. Anchor Manufacturing, Inc. (“U.S. Anchor”) that Rule and Tie Down attempted and conspired to monopolize the United States market for light weight fluke-style anchors for small boats by means of below-cost pricing intended to drive out competition. U.S. Anchor cross-appeals the district court’s order of a directed verdict on its state law claims arising from the same allegations. We reverse the denial of defendants’ motions concerning the federal claims. With respect to the state law claims, we certify the dispositive issues for authoritative resolution by the Supreme Court of Georgia.

I. FACTS

This case involves several manufacturers and suppliers of light weight anchors for ultimate retail purchase by owners of recreational boats and small commercial fishing craft. As the district court observed in denying cross-motions for summary judgment,

[ajnchors and other marine industry products are generally sold by suppliers to wholesale distributors, who in turn sell the anchors to boat dealers, marinas, and other retailers for ultimate resale to the consumer, the boat owner. The supplier may either manufacture its own anchors, as does U.S. Anchor, or purchase them from another domestic manufacturer, as [Rule] does from Tie Down, or import them from abroad.

U.S. Anchor Mfg. v. Rule. Indus., 717 F.Supp. 1565, 1568 (N.D.Ga. 1989).

Within the general category of fluke anchors are four distinct product groups recognized in the industry: (1) expensive premium anchors, (2) the “Danforth Standard” brand line of anchors sold only by Rule, (3) so-called “generic” versions of the Danforth Standard, and (4) inexpensive economy anchors used primarily for lake boating.

Rule is a diversified Massachusetts firm that sells an assortment of marine, hardware and automotive products to wholesale distributors. It entered the fluke anchor industry in 1983 when it obtained the rights to sell the Danforth brand line of anchors. Prior to 1985, Danforth anchors were manufactured for Rule exclusively by the Jacquith Company (“Jacquith”) in New York. Tie Down is a smaller manufacturing firm in Georgia that began selling generic and economy fluke anchors in the late 1970s under the “Hooker” brand name. In May 1985 Rule obtained the Hooker trademark and the exclusive right to purchase and distribute Tie Down’s anchor production in a transaction that U.S. Anchor has characterized as a “merger.” After it sold the right to market its own anchors, Tie *990Down agreed to manufacture both generic/economy and Danforth brand anchors for Rule. Tie Down’s only role in the fluke anchor industry since 1985 has been as one of Rule’s suppliers.

U.S. Anchor is a Georgia company founded in 1985 by William Chapman (“Chapman”), the immediate past president of Tie Down whose responsibilities there had recently ended. U.S. Anchor both manufactures and distributes generic and economy fluke anchors under the “Sentinel” brand name. Between August 1985 when it first sent out price lists and December 31, 1990, its market share increased to between 45 and 68%, depending on how the relevant product market is defined and measured.

Shortly after U.S. Anchor entered the market in August 1985, on the eve of the 1985-86 marine products season,1 Rule and U.S. Anchor engaged in a price war. Following publication of U.S. Anchor’s August price list, Rule published prices in September that were approximately 11 to 18% higher than U.S. Anchor’s. Thus, U.S. Anchor’s prices were 10 to 15% lower than Rule’s. (R30-27; compare USTX 343 with USTX 345.)2 In October, after U.S. Anchor had received substantial orders from Rule customers, Rule cut its prices by 20%, i.e., to levels 6 to 12% below U.S. Anchor’s August prices. (R49-28; RTX 584.) U.S. Anchor then matched Rule’s October prices. In a written report to Rule, USTX 683, Tie Down’s president Charles MacKarvich (“MaeKarvich”) estimated U.S. Anchor’s costs of production and hypothetical projected sales for a twelvemonth period. He theorized that if Rule lowered its prices further and offered extended credit terms to customers, U.S. Anchor would be forced to adopt even more attractive terms in order to compete. From his estimates of cash flow and net revenue derived from these cost and sales projections, he predicted that such terms would subject U.S. Anchor to a negative cash flow and an actual net loss over the 1985-86 marketing year. Rule implemented price reductions consistent with MacKarvich’s report in November 1985. In December, U.S. Anchor merely matched Rule’s prices and did not attempt to undercut them. (R30-39-40, USTX 351, 353, 543.) U.S. Anchor contends that Rule’s first price cut in October was predatory and that all subsequent sales at or below that level were also predatory.

After the pricing conduct at issue in this case began, distributors’ prices for generic brands in the smaller, popular sizes ranged between $3 and $14 depending on weight, and prices for Danforths were spread 50 to 96% higher.3 Among the more expensive, larger anchors the spread between Danforth and generic brands was even greater. Excluding premium anchors,4 annual unit sales of fluke anchors in the United States during the time relevant to this case has varied from 232,000 to 347,000. (USTX 479.)

At trial the parties noted differing possible measures of Rule’s share of the relevant product market after the acquisition of Tie Down’s anchor line in May 1985, four months before the close of the 1984-85 marine season at the end of August. This dispute encompassed two aspects of market share: whether to define the product market as including the high priced Danforth anchors or only the less expensive generic and economy models, and whether to measure market shares in terms of unit sales or dollar revenues. Including the Danforth line and measuring market shares in revenue, U.S. Anchor asserts that Rule and Tie Down togeth*991er controlled 90.5% of the fluke anchor market during the 1984-85 season, the last year before Rule’s alleged predation began and the last year before the merger with Tie Down, and that Rule possessed 60.0% of the market during the 1985-86 selling year. (USTX 467.) Using Rule’s most favorable calculation, which measures share in units and excludes Danforths from the market, Rule contends that the combined Rule/Tie Down market share in 1984-85 was only 61.5%, (RTX 674), and that Rule’s aggregate 1985-86 share was 30.1%, (id.; RTX 675 at l).5 Rule also submitted evidence that its 1985-86 unit market share, including Dan-forths, was 43.1%. (RTX 675 at 1.) Notably, all of these figures encompass an entire season and none of them attempts to pinpoint Rule’s share at the exact date when the alleged predation began in October 1985, several months after U.S. Anchor entered the market. The evidence shows, and the parties agree, that the Rule/Tie Down market share consistently decreased after August 1985 when U.S. Anchor first began to solicit orders. The parties also agree that the relevant geographic market was the United States.

II. PROCEDURAL HISTORY

In November 1985 Rule filed suit against U.S. Anchor for various violations of state and federal law not involving predatory pricing. The suit was settled on March 19,1986, when U.S. Anchor and Rule executed an agreement releasing each other from liability for all events occurring prior to the date of the release. Tie Down was a party to neither the litigation nor the ensuing release.

On November 13, 1986, U.S. Anchor sued Rule and Tie Down, alleging that Rule had attempted to monopolize the fluke anchor market in violation of section 2 of the Sherman Act6 beginning in October 1985 by engaging in predatory pricing. U.S. Anchor also alleged that Rule and Tie Down had conspired to restrain trade in violation of section 1 of the Sherman Act7 and conspired to attain a monopoly in violation of section 2, by agreeing to charge predatory prices, also beginning in October 1985. U.S. Anchor also asserted an illegal tying arrangement by Rule, whereby its newly patented and supposedly revolutionary “Deepset” anchors allegedly were sold only to distributors who abstained from buying generic fluke anchors from suppliers other than Rule, in violation of section 1 of the Sherman Act and section 3 of the Clayton Act, as amended by the Robinson-Patman Act.8 U.S. Anchor further alleged that Rule and Tie Down had conspired to restrain trade in violation of Georgia law.

*992The district court denied the parties’ cross-motions for summary judgment. 717 F.Supp. 1565. A jury trial followed during which the defendants moved for directed verdicts9 as to all claims. The court granted their motions on the state law claims because it concluded that Georgia law did not permit damages to be recovered for a conspiracy in restraint of trade. The jury found Rule solely liable for attempted monopolization and jointly liable with Tie Down on both conspiracy counts. The verdict exonerated Rule of illegal tying. (RIO-321.) The jury set damages for each of the three violations at $1,638,028, which the court trebled to $4,914,-084. Tie Down and Rule both moved for judgment notwithstanding the verdict on liability and for a new trial on the issue of damages. (Rll-348, 349.) The district court denied these motions, awarded U.S. Anchor statutory attorney fees in the stipulated amount of $800,000 and entered judgment accordingly. (R14-382.) Rule and Tie Down appealed, and U.S. Anchor cross-appealed with respect to the state law tort claims. U.S. Anchor does not appeal the judgment on the tying claim.

III. CONTENTIONS OF THE PARTIES

Rule contends that it engaged in no predatory conduct and disputes U.S. Anchor’s showing of Rule’s and Tie Down’s costs of producing the anchors. Since a predatory pricing claim requires proof that defendants attempted or conspired to drive a competitor out of the relevant market by “pricing below some appropriate measure of cost,” the issue of which costs to count may be vital. Matsushita Elec. Indus. Co. v. Zenith Radio Corp., 475 U.S. 574, 585 n. 8, 106 S.Ct. 1348, 1355 n. 8, 89 L.Ed.2d 538 (1986) (noting but not resolving debate over which costs are “relevant”), on remand, In re Japanese Elec. Prods. Antitrust Litig., 807 F.2d 44 (3d Cir. 1986), cert. denied, 481 U.S. 1029, 107 S.Ct. 1955, 95 L.Ed.2d 527 (1987). Rule advances numerous criticisms of U.S. Anchor’s expert testimony on this point. Tie Down contends that U.S. Anchor failed to adduce any evidence that Tie Down’s prices to Rule were below Tie Down’s cost, that Tie Down had any knowledge of (or control over) Rule’s other costs, or that it had any control over Rule’s prices.

Rule also contends that it had no dangerous probability of successfully achieving a monopoly. The parties first dispute the existence of barriers to entry in the relevant market. Rule and Tie Down contend that without high barriers, a successful monopolist would not have been able to recoup the foregone profits inherent in below-cost pricing by charging supra-competitive prices following the end of the victim’s competitive presence.10 U.S. Anchor contends that there was sufficient evidence of entry barriers to permit the jury to find them and that in any case actual recoupment is not required as a matter of law before the jury may find an attempt or conspiracy to monopolize. Second, Rule points to U.S. Anchor’s own success and Rule’s declining fortunes in the anchor market as evidence that it could not have monopolized.

Rule and Tie Down also challenge the sufficiency of the evidence of unlawful conspiracy. The parties dispute the inference to be drawn from plaintiffs exhibit 683, the Mac-Karvich market report. U.S. Anchor contends that MacKarvich was proposing to drive the new entrant from the marketplace. Defendants offered expert testimony, corroborated by MacKarvich himself, that studies *993of competitors’ costs and revenues are common in competitive industries and that a projected loss after the first year of operation is ordinarily not enough to drive any new entrant from the market, since start-up companies must generally expect early losses. In its cross-appeal U.S. Anchor challenges the district court’s exclusion of certain evidence that allegedly supports the existence of a conspiracy.

Rule and Tie Down challenge the sufficiency of U.S. Anchor’s proof concerning damages. They argue that at least some of their price cuts were instituted to meet competition from foreign fluke anchor manufacturers and any loss of sales by U.S. Anchor resulting from such reductions is not antitrust injury. Moreover, they contend, the base price from which U.S. Anchor’s revenue losses were calculated should have reflected competitive levels as shown by Rule’s and U.S. Anchor’s early, allegedly non-predatory reductions rather than prices prevailing before U.S. Anchor’s entry into the market.

Rule and U.S. Anchor dispute the scope and effect of their settlement agreement in the prior litigation. Rule contends that liability for all predatory sales before the date of the release was discharged. Moreover, Rule maintains that the alleged predatory scheme was ongoing at the time the contract was executed and therefore all post-release liability was discharged as well. U.S. Anchor contends that a general release is ineffective to discharge undiscovered antitrust liability as a matter of law and, moreover, that post-release damages were not waived. We do not reach this dispute as it applies to the federal antitrust claims.11 As applied to the state law claims, we certify the question, along with the substantive issues of Georgia law, for resolution by the Supreme Court of Georgia.

In its cross-appeal U.S. Anchor also argues that the district court should not have granted a directed verdict on its state law claims because Georgia law allows private damage actions for conspiracies in restraint of trade. Rule and Tie Down disagree with U.S. Anchor’s interpretation of Georgia law.

IV. STANDARD OF REVIEW

We review rulings on motions for judgment as a matter of law by applying de novo the same legal standards used by the district court. Miles v. Tennessee River Pulp & Paper Co., 862 F.2d 1526, 1528 (11th Cir. 1989). Both courts consider all the evidence, but all reasonable inferences must be drawn in the nonmovant’s favor. If the jury verdict is supported by substantial evidence — that is, enough evidence that reasonable minds could differ concerning material facts — the motion should be denied. A mere scintilla of evidence in the entire record, however, is insufficient to support a verdict. See Hessen ex rel. Allstate Ins. Co. v. Jaguar Cars, Inc., 915 F.2d 641, 644 (11th Cir. 1990). Denial of a motion for a new trial is reviewed for clear abuse of discretion. Id. at 644-45. A district court’s evidentiary rulings are not disturbed unless there is a clear showing of abuse of discretion. Id. at 645.

V. ATTEMPTED MONOPOLIZATION

There are three essential elements of a claim alleging attempted monopolization under section 2 of the Sherman Act. First, the plaintiff must show that the defendant possessed the specific intent to achieve monopoly power by predatory or exclusionary conduct. Second, the defendant must in fact commit such anticompetitive conduct. Third, there must have existed a dangerous probability that the defendant might have succeeded in its attempt to achieve monopoly power. Spectrum Sports, Inc. v. McQuillan, — U.S. -, -, 113 S.Ct. 884, 890, 122 L.Ed.2d 247 (1993); see McGahee v. Northern Propane Gas Co., 858 F.2d 1487, 1493 (11th Cir. 1988), cert. denied, 490 U.S. 1084, 109 S.Ct. 2110, 104 L.Ed.2d 670 (1989); 3 Phillip Areeda & Donald F. Turner, Antitrust Law ¶ 820 at 312 (1978) [hereinafter Areeda & Turner', Antitrust Law ]. We address these elements in reverse order.

*994A. Dangerous Probability of Success

To have a dangerous probability of successfully monopolizing a market the defendant must be close to achieving monopoly power.12 Monopoly power is "the power to raise prices to supra-competitive levels or the power to exclude competition in the relevant market either by restricting entry of new competitors or by driving existing competitors out of the market." American Key Corp. v. Cole Nat'l Corp., 762 F.2d 1569, 1581 (11th Cir. 1985). Most attempts to measure monopoly power involve quanti~iing the degree of concentration in a relevant market and/or the extent of a particular firm's ability to control productive capacity in that market. In analyzing attempted monopolization's dangerous probability of success element, the estimate of market power is necessarily speculative to some extent because it requires an evaluation of future behavior by market par~ ticipants, viewed at the time the alleged attempt began. We are not without guideposts, however.

Relevant determinants of the market power of a prospective predator in this regard include its absolute and relative market shares, and those of competing firms; the strength and capacity of current competitors; the potential for entry; the historic intensity of competition; and the impact of the legal or natural environment.

International Tel. & Tel. Corp., 104 F.T.C. 208, 412 (1984) (citation and footnotes omitted). Despite the seemingly broad array of factors employed by the Federal Trade Commission, the principal judicial device for measuring actual or potential market power remains market share, typically measured in terms of a percentage of total market sales. Thus, at the outset the appropriate market must be defined or identified.13

Defining the market is a necessary step in any analysis of market power and thus an indispensable element in the consideration of any monopolization or attempt case arising under section 2. Walker Process Equip., Inc. v. Food Mach. & Chem. Corp., 382 U.s. 172, 177, 86 S.Ct. 347, 350, 15 L.Ed.2d 247 (1965); American Key, 762 F.2d at 1579. Although the issue is fully developed in the fact section of Rule's brief, the argument section does not address the precise question of market definition. U.S. Anchor, in the fact section of its brief, contends that the question of market definition is not appropriately before us because Rule does not argue the point. (U.S. Anchor's Br. at 3 n. 1.) We must considei the question nevertheless before passing on the legal significance of evidence concerning Rule's potential market power. As the issue of Rule's dangerous probability of success has been preserved through argument, the subsidiary question of market definition is also preserved because it is set forth fully in the fact section of Rule's brief.14 The issue was fully argued before the district court.15

The definition of the relevant market is essentially a factual question, so the precise issue we first must address is whether U.S. Anchor introduced sufficient evidence to raise a jury question on the inclusion of Danforths. See, e.g., Yoder Bros. v. Califor*995nia-Florida Plant Corp., 537 F.2d 1347, 1366 (5th Cir. 1976), cert. denied, 429 U.S. 1094, 97 S.Ct. 1108, 51 L.Ed.2d 540 (1977).16

1. Defining the Market

“Defining a relevant product market is primarily ‘a process of describing those groups of producers which, because of the similarity of their products, have the ability — actual or potential — to take significant amounts of business away from each other.’ ” General Indus. Corp. v. Hartz Mountain Corp., 810 F.2d 795, 805 (8th Cir. 1987) (quoting SmithKline Corp. v. Eli Lilly & Co., 575 F.2d 1056, 1063 (3d Cir.), cert. denied, 439 U.S. 838, 99 S.Ct. 123, 58 L.Ed.2d 134 (1978)). The reasonable interchangeability of use or the cross-elasticity of demand17 between a product and its substitutes constitutes the outer boundaries of a product market for antitrust purposes. Brown Shoe Co. v. United States, 370 U.S. 294, 325, 82 S.Ct. 1502, 1523, 8 L.Ed.2d 510 (1962).

[W]ithin this broad market, well-defined submarkets may exist which, in themselves, constitute product markets for antitrust purposes. The boundaries of such a submarket may be determined by examining such practical indicia as industry or public recognition of the submarket as a separate economic entity, the product’s peculiar characteristics and uses, unique production facilities, distinct customers, distinct prices, sensitivity to price changes, and specialized vendors.... The cross-elasticity of production facilities may also be an important factor in defining a product market....

Id. at 325 & n. 42, 82 S.Ct. at 1523-24 & n. 42 (citations and footnotes omitted). As the Supreme Court’s language itself suggests, defining a “submarket” is the equivalent of defining a relevant product market for antitrust purposes. International Telephone & Telegraph adequately summarizes our view of the relevant proof:

Reliable measures of supply and demand elasticities provide the most accurate estimates of relevant markets. However, it is ordinarily quite difficult to measure cross-elasticities of supply and demand accurately. Therefore, it is usually necessary to consider other factors that can serve as useful surrogates for cross-elasticity data.... In the case of product market definition, these factors may include

whether the products and services have sufficiently distinctive uses and characteristics; whether industry firms routinely monitor each other’s actions and calculate and adjust their own prices (at least in part) on the basis of other firms’ prices; the extent to which consumers consider various categories of sellers ... as substitutes; and whether a sizeable price disparity between different types of ... sellers ... persists over time for equivalent amounts of comparable goods and services.

104 F.T.C. at 409 (quoting Grand Union Co., 102 F.T.C. 812, 1041 (1983)) (footnotes omitted).

We note that Danforth brand anchors are functionally interchangeable with their equivalent counterparts among the generic brands. Indeed, among smaller sized anchors the Hooker and Danforth anchors have always been virtually identical. (R30-131-33; R33-129-31.) This interchangeability suggests a likelihood that consumers of generic brands would willingly switch to Dan-forths in the event of significant price increases among generics. Similarly, Danforth customers might switch to generic brands if Rule implemented a significant increase in the price of Danforths. The likelihood of demand substitution, if proven, weighs strongly in favor of including the two categories of product within a single market for antitrust analysis. This is so because the very purpose of defining the relevant market under section 2 is to determine whether a monopolist, cartel or oligopoly in that market would be able to reduce marketwide output simply by cutting its own output, and thereby *996raise marketwide prices above competitive levels. United States v. E.I. du Pont de Nemours & Co. (The Cellophane Case), 351 U.S. 377, 395, 76 S.Ct. 994, 1007, 100 L.Ed. 1264 (1956); Satellite Television & Associated Resources, Inc. v. Continental Cablevision, Inc., 714 F.2d 351, 356 (4th Cir. 1983), cert. denied, 465 U.S. 1027, 104 S.Ct. 1285, 79 L.Ed.2d 688 (1984).18

We hold, however, that the relevant market in this case constituted light weight generic and economy fluke anchors. Four of the Brown Shoe factors weigh strongly in favor of excluding Danforths from the relevant market: distinctly higher prices, a distinct group of customers, strongly inelastic demand and limited substitution of supply. Moreover, the higher prices charged for Danforths are evidence that a distinct group of customers was unwilling to switch away from the prestigious branded product in response to price increases above competitive levels. The fact that this group remained loyal to Danforths despite prices 50 to 96% and more above prices for functionally interchangeable alternative products shows inelastic demand and limited demand interdependence. More importantly, U.S. Anchor showed no reasonable possibility that a significant number of consumers would have switched to Danforths, many of which were offered at nearly double the price of their generic substitutes, in response to more modest increases in generic prices. And as more fully discussed below, there is no evidence that Rule had (or would have) varied its output of Danforths in response to price changes in the broader market. We hold, therefore, that the record provides no support for finding significant cross-elasticity of demand or supply between Danforths and generic anchors.

First, U.S. Anchor’s evidence was insufficient for a reasonable juror to conclude that there was a significant cross-elasticity of demand. U.S. Anchor’s evidence demonstrated that an increase in the spread between prices for Danforths and other anchors had coincided with lower sales of Danforths. During the period from September 1985 until August 1990, sales of Danforths fell by 61.5% while the spread between the prices of Danforths and other anchors increased by 9.1%. (USTX 638; R40-106.) (According to the exhibit, Danforth prices rose while Sentinel and Hooker prices fell). Although we recognize that correlation is often relied upon to infer causation, see, e.g., Cellophane, 351 U.S. at 400, 76 S.Ct. at 1010, we do not believe that this aggregation of sales data over five years provided a sufficiently close correlation between changes in demand and price to justify the inference that consumers were willing and able to switch away from Dan-forths because of increasing price differences. The exhibit wholly fails to take account of factors other than price (or quality) which may have affected demand for Danforths. If changes in relative prices had been more closely correlated in time with shifting purchases then it might have been reasonable to infer that the demand shifts were caused by the price differences. As the evidence stands, however, the datum aggregating demand behavior from 1985 to 1990 fails to provide any basis from which the jury could have inferred that the demand shifts were caused by prices instead of other factors. Those non-price, non-quality factors might well have included consumers’ increased awareness of the similarities between Dan-forths and other brands (perhaps caused by U.S. Anchor’s successful promotion of its own products), changing attitudes concerning thrift and the value of money, the decline in demand for fluke anchors generally after the 1987-88 season, (see USTX 479), or competition from Rule’s own more expensive premium Deepset line. Over time the shape of a demand curve changes independently of variations in the pricing and quality of particular substitute products. Aggregate (or average) evidence of demand over too long a period of time provides no support for inferring that changes apparently correlated with substitute price movements represent shifts in the curve caused by those variations in prices. Given the changes in the behavior of competitors that occurred over the five years in *997question, namely the development of fierce price competition between Rule and U.S. Anchor in the generic and economy market and the introduction of Deepsets, we conclude that the average Danforth sales statistic was insufficient evidence from which the jury could have inferred demand cross-elasticity in October 1985 or thereafter. Cf. Yoder Bros., 537 F.2d at 1367-68. This conclusion is buttressed by the more precise sales data provided by USTX 508. Comparing the 1985-86 and 1986-87 seasons, which are the two closest in time to the date when the alleged predation began in October 1985 for which data were offered, the exhibit shows that unit sales of Danforths fell 5.4%19 despite a price reduction of 0.5% and a simultaneous increase in the prices of generic anchors of 0.8%. Id. at 2. Danforths suffered this decline while the overall demand for fluke anchors jumped 19%, from 273,000 to 325,000 in annual unit sales. (USTX 479.)20

Just as an increase in Danforth prices might have been expected to drive customers away from Rule and into the arms of generic manufacturers, an increase in prices for generic brands would likely cause some otherwise price-sensitive customers to prefer the more expensive Danforths. Nonetheless, the present record provides no basis other than guesswork for concluding that a shift away from generics would have been significant in magnitude;21 the large spread in prices between generic anchors and Danforths tends to suggest that the shift would not have been great. Thus, we conclude that the record provides no support for finding significant cross-elasticity of demand between Danforths and generics.

Second, the evidence was insufficient for a reasonable juror to find a significant cross-elasticity of supply. The jury could not reasonably have found that the manufacturing capacity used to make Danforths likely would have been switched to making generic anchors in response to moderate price increases by a sole seller of the lower priced products. To be sure, the productive processes employed in manufacturing Danforths were virtually identical to those used for generics. (R33-145-50.) Yet it defies logic to suggest that a rational supplier22 would switch from selling branded products at high prices to selling equally costly equivalent products at lower prices, even assuming that the lower prices would yield significant supranormal profits. Put another way, it would be unreasonable to expect Rule to lower the price of Danforths and abandon its ability to discriminate against brand-conscious boaters solely to earn smaller profits. There was insufficient evidence of likely supply substitution from which to conclude that any portion of Danforth output would have served to constrain price increases among the generic anchors.23

Moreover, the record demonstrates that the Danforth line, although functionally equivalent to their counterparts, may have constituted its own market based on consumer brand loyalty. The fluke anchor industry presented the unusual circumstance of severe price discrimination against a distinct group of consumers based solely on brand preference. U.S. Anchor’s expert, Dr. Willard F. *998Mueller, testified on direct examination that “people have gotten an attachment to the Danforth Standard in this case, it had kind of a mystique about it at one time, ... what happens in one year, simply a price difference, doesn’t result in an immediate kind of shift.” (R40-106.) Although interbrand competition generally restrains the pricing behavior of individual brand sellers, Continental T.V., Inc. v. GTE Sylvania Inc., 433 U.S. 36, 52 n. 19, 97 S.Ct. 2549, 2558 n. 19, 53 L.Ed.2d 568 (1977), on remand, 461 F.Supp. 1046 (N.D.Cal. 1978), aff'd, 694 F.2d 1132 (9th Cir. 1982), it is settled that customer brand loyalty may constitute an impediment to competition and thus an aid in the exercise of market power. See, e.g., United States v. Pabst Brewing Co., 384 U.S. 546, 559-61, 86 S.Ct. 1665, 1672, 16 L.Ed.2d 765 (1966) (Harlan, J., concurring), on remand, 296 F.Supp. 994 (E.D.Wis. 1969).24 A single branded product may, in rare cases, constitute its own relevant market. Los Angeles Mem. Coliseum Comm’n v. National Football League, 726 F.2d 1381, 1393 (9th Cir.), cert. denied, 469 U.S. 990, 105 S.Ct. 397, 83 L.Ed.2d 331 (1984).

The understanding that brand loyalty may facilitate monopolization is consistent with the general proposition that the ability to discriminate against a distinct group of customers by charging higher prices for otherwise similar products demonstrates the existence of market power with respect to that group. See United States v. Grinnell Corp., 384 U.S. 563, 574, 86 S.Ct. 1698, 1706, 16 L.Ed.2d 778 (1966).25 The existence of such market power may, as a practical matter, remove the higher priced product from the broader market composed of its functional substitutes. See C.E. Services, Inc. v. Control Data Corp., 759 F.2d 1241, 1246 (5th Cir.), cert. denied, 474 U.S. 1037, 106 S.Ct. 604, 88 L.Ed.2d 583 (1985) (holding that “a ubiquitous price differential of some 20-25%” between branded and unbranded services, combined with other Brown Shoe factors, could justify finding a separate market for the unbranded services and thus precluded summary judgment on the issue of market definition).

We do not suggest that the existence or hypothetical possibility of monopoly power over one product automatically excludes it from a broader market. “[SJubmarkets are not a basis for the disregard of a broader line of commerce that has economic significance.” United States v. Phillipsburg Nat’l Bank & Trust Co., 399 U.S. 350, 360, 90 S.Ct. 2035, 2041, 26 L.Ed.2d 658 (1970). We do hold, however, that regardless of which party in the case bears the ultimate burden of persuasion, the broader economic significance of a submarket must be supported by demonstrable empirical evidence. Although perhaps difficult to come by, evidence that the dominant firm within a submarket costs of production were insensitive to changes in the quantity of goods sold, suggesting that its only rational response would be to increase output to satisfy the higher demand in the event of price increases above competitive levels in the broader market, might show that submarket production in fact disciplined price levels in the broader market. Especially if the submarket represents a premium-priced segment of the broader market, the relevance of proof regarding elasticity of supply would depend on the validity of the assumption that significant numbers of consumers would switch in response to significant price increases in the broader market, an assumption that may or may not be supported by evidence or common experience. In the present case U.S. Anchor can rely upon neither evidence nor inference. Simpler evidence of supply and demand substitution, like proof that producers in the sub-market had actually increased or decreased their sales in response to corresponding price *999changes in the broader market, would also suffice. As we have pointed out, however, U.S. Anchor failed to meet its burden of proving interdependent market behavior by this method as well.

Considering all the evidence in light of the factors identified by Cellophane and Brown Shoe and explained in subsequent decisions, we conclude as a matter of law that the relevant product market was light weight generic and economy fluke anchors.

2. Measuring Power in the Market

The principal measure of actual monopoly power is market share, and the primary measure of the probability of acquiring monopoly power is the defendant’s proximity to acquiring a monopoly share of the market. Thus, a sufficiently large market share may alone create a genuine dispute over whether the defendant possessed a dangerous probability of successfully monopolizing a market despite the existence of other facts tending to make monopolization unlikely, thereby precluding summary judgment for the defendant. McGahee v. Northern Propane Gas Co., 858 F.2d at 1506. When assessing market shares for the purpose of ascertaining market power the appropriate measure of a firm’s share is the quantity of goods or services actually sold to consumers. Although revenues are often relied upon as a surrogate for quantity, actual unit sales must be used whenever a price spread between various products would make the revenue figure an inaccurate estimator of unit sales. Brown Shoe, 370 U.S. at 341 n. 69, 82 S.Ct. at 1533 n. 69.

In McGahee we noted in dicta that several factors may be relevant to whether a particular market share evidences a dangerous probability of success. 858 F.2d at 1505 (citing McGahee v. Northern Propane Gas Co., 658 F.Supp. 189, 196-97 (N.D.Ga. 1987), rev’d, 858 F.2d 1487 (11th Cir. 1988)). In finding no dangerous probability of success the district court had relied upon the ease of entry by new firms and expansion from adjacent geographic markets, the number and size of alleged victims of the predation and the defendant’s declining market share during the alleged attempt to monopolize. 658 F.Supp. at 196-97. Nevertheless, we held:

Without examining any factors to determine what market share would be necessary for Northern Propane’s alleged predatory pricing to present a dangerous probability of success, we can say that a sixty or sixty-five percent market share is a sufficiently large platform from which such a scheme could be launched to create a genuine issue of material fact as to whether there was a dangerous probability that Northern Propane would succeed in achieving a monopoly.

McGahee, 858 F.2d at 1506. Finding it “undisputed” that the defendant possessed such a share, we reversed the district court’s order of summary judgment for the defendant and remanded for further proceedings. Our holding in McGahee that market share estimated with reasonable confidence to fall between 60 and 65% suffices to raise a jury question concerning dangerous probability of success is binding circuit precedent. Sherry Mfg. Co. v. Towel King, Inc., 822 F.2d 1031, 1034 n. 3 (11th Cir. 1987). We do note, however, the tension between McGahee’s bright-line approach and Cliff Food Stores, Inc. v. Kroger, Inc., 417 F.2d 203 (5th Cir. 1969), in which the court noted that “one must be particularly wary of the numbers game of market percentage when considering an ‘attempt to monopolize’ suit” under the dangerous probability standard. 417 F.2d at 207 n. 2; cf. United States v. Columbia Steel Co., 334 U.S. 495, 528, 68 S.Ct. 1107, 1124, 92 L.Ed. 1533 (1948) (“the relative effect of percentage command of a market varies with the setting in which that factor is placed”) (actual monopolization case). We believe the cases may be reconciled by requiring a careful definition of the relevant market (as mandated by Walker Process and American Key)26 and an assessment of each firm’s *1000ability to vary its output in calculating the size of the market and attributing individual market shares. See, e.g., United States v. General Dynamics Corp., 415 U.S. 486, 499-504, 508-10, 94 S.Ct. 1186, 1194-97, 1199-1200, 39 L.Ed.2d 530 (1974) (measuring power in market for coal in terms of possession or likely near-term acquisition of uncommitted reserves instead of overall sales, because most sales represented fulfillment of existing long-term requirements contracts).

In Cliff Food Stores the former Fifth Circuit stated that something more than 50% market share would be required to show actual monopoly, at least in the absence of collusive price leadership or tacit coordination in an industry. 417 F.2d at 207 n. 2. The Second Circuit in Broadway Delivery Corp. v. United Parcel Service of America, Inc., 651 F.2d 122 (2d Cir.), cert. denied, 454 U.S. 968, 102 S.Ct. 512, 70 L.Ed.2d 384 (1981), similarly suggested that the absence of actual monopoly power could be found as a matter of law when the defendant supplies only 50% of the market, “or even somewhat above that figure, [when] the record contains no significant evidence concerning the market structure to show that the defendant’s share of that market gives it monopoly power.” 651 F.2d at 129. Despite these suggestions, we have discovered no eases in which a court found the existence of actual monopoly established by a bare majority share of the market. Nevertheless, a dangerous probability of achieving monopoly power may be established by a 50% share. For this reason, it is usually necessary to evaluate the prospects for monopolization as they existed when the alleged attempt began. As shown by the undisputed facts discussed infra, Rule never possessed a dangerous probability of success during the time for which U.S. Anchor seeks damages.

U.S. Anchor points to the combined market shares of Rule and Tie Down at the end of the 1984-85 season, immediately before the transaction that eliminated Tie Down as a supplier and transferred its production to Rule. Accepting arguendo the implicit contention that Tie Down’s pretransaction market share should be attributed to Rule, we conclude from the undisputed evidence that Rule’s market share on August 31, 1985, the eve of the 1985-86 season, was 61.5%, (RTX 674), and its aggregate (average) share over the entire season was 30.1%, (id.; RTX 675 at 1).

Rule has argued that we should not attribute all of Tie Down’s pre-transaction market share to it. After the transaction Tie Down had no need for its anchor sales representatives, many of whom found engagements with U.S. Anchor and employed their connections and reputation on behalf of the newcomer’s selling efforts. Moreover, U.S. Anchor’s Chapman was well known to customers from his days with Tie Down. Thus, according to Rule, U.S. Anchor stepped into Tie Down’s shoes and inherited at least some of Tie Down’s pre-transaction market share, presumably that portion which U.S. Anchor had the productive capacity to satisfy. This argument is persuasive, although it may be subject to rebuttal on at least two grounds. Cf. American Academic Suppliers, Inc. v. Beckley-Cardy, Inc., 922 F.2d 1317, 1321-22 (7th Cir. 1991). First, the depth of Rule’s product line and the expertise of its own sales force conferred competitive advantages which might have induced some of Tie Down’s former customers to stay with the Hooker line. Second, the anchor industry was highly concentrated and customers had few alternative sources of supply, a factor that is especially important in view of Rule’s effort to link purchase of the Deepset anchors to exclusive dealing arrangements with distributors. We need not reach the merits of Rule’s contention, however, because even if we consider Rule to have had 61.5% of the market on August 31,1985, there was insufficient evidence from which the jury could have found a dangerous probability of monopolization in October.

As we have outlined above, Rule’s average market share for the 1985-86 season was 30.1%, a fact which strongly indicates that Rule’s share declined sharply from 61.5% after U.S. Anchor’s entry into the market in August. For the month of October, U.S. Anchor’s sales of generic and economy anchors exceeded Rule’s by 5.7%. (RTX 675 at 15). Prior to October U.S. Anchor had no sales at all, but the firm was accepting orders *1001during this time and apparently possessed the capacity to fill them. Thus, Rule was never able to maintain a majority position in the market during the 1985-86 season. Cf. General Dynamics, 415 U.S. at 501-02, 94 S.Ct. at 1196. Accordingly, because Rule possessed less than 50% of the market at the time the alleged predation began and throughout the time when it was alleged to have continued, there was no dangerous probability of success in October 1985 as a matter of law.

3. Recoupment

Rule argues that the district court should have granted its motion for judgment notwithstanding the verdict based on its contention that there can be no dangerous probability of successful monopolization by predatory pricing unless it is shown that the defendant would have recouped the foregone revenues associated with its price-cutting strategy.27 Our disposition of this case, however, makes it unnecessary to address Rule’s recoupment argument.

B. Anticompetitive Conduct, Specific Intent and Damages

Our conclusion that U.S. Anchor failed to show a dangerous probability of success makes it unnecessary for purposes of resolving its attempt claim to evaluate the evidence of Rule’s and Tie Down’s costs, as would be required to classify its pricing conduct as anticompetitive. See Matsushita, 475 U.S. at 585 n. 8, 106 S.Ct. at 1355 n. 8; International Air Industries, Inc. v. American Excelsior Co., 517 F.2d 714, 723-25 (5th Cir. 1975). The same is true with respect to the evidence of specific intent to achieve monopoly power by unlawful conduct, although we note that such intent may sometimes be inferred from predatory conduct itself. Spectrum Sports, — U.S. at -, 113 S.Ct. at 892; International Tel. & Tel., 104 F.T.C. at 401-02; see also McGahee, 858 F.2d at 1503-04. Nor must we decide whether to parse this evidence for the precise level during each season at which Rule’s prices unlawfully dropped below its costs in order to assess U.S. Anchor’s proof of damages, as requested by Rule. See MCI Communications Corporation v. American Telephone and Telegraph Company, 708 F.2d 1081, 1162, 1165 (7th Cir. 1983).

VI. CONSPIRACY

U.S. Anchor’s conspiracy claims are distinct from its attempted monopolization claim. The elements of a conspiracy to monopolize under Section 2 are (1) an agreement to restrain trade, (2) deliberately entered into with the specific intent of achieving a monopoly rather than a legitimate business purpose, (3) which could have had an anticompetitive effect, and (4) the commission of at least one overt act in furtherance of the conspiracy. Seagood Trading Corp. v. Jerrico, Inc., 924 F.2d 1555, 1576 (11th Cir. 1991). The elements of a conspiracy to restrain trade under Section 1 are (1) an agreement to enter a conspiracy (2) designed to achieve an unlawful objective. Bolt v. Halifax Hosp. Medical Ctr., 891 F.2d 810, 820 (11th Cir.), cert. denied, 495 U.S. 924, 110 S.Ct. 1960, 109 L.Ed.2d 322 (1990), appeal after remand, 980 F.2d 1381 (11th Cir. 1993). The plaintiff must also prove (3) “actual unlawful effects [or] facts which radiate a potential for future harm” to competition. Times-Picayune Publishing Co. v. United States, 345 U.S. 594, 622, 73 S.Ct. 872, 888, 97 L.Ed. 1277 (1953).

There is no requirement, however, that a conspiracy-under either provision have a dangerous probability of successfully achieving its objectives. Copperweld Corp. v. Independence Tube Corp., 467 U.S. 752, 767-68, 104 S.Ct. 2731, 2740, 81 L.Ed.2d 628 (1984). Moreover, “[a] section 1 plaintiff ... need not prove an inteniMmJhe_pari_of_the co-conspirators to restrain trade or to build a monopoly. So long as the purported conspiracy has an Anticompetitive effect, the plaintiff has made out a ease under section 1.” Bolt, 891 F.2d at 819-20 (citations omitted). We have said, however, that “a section 1 claim and a section 2 conspiracy to monopolize *1002claim require the same threshold showing— the existence of an agreement to restrain trade.” Seagood, 924 F.2d at 1576.

U.S. Anchor points to evidence of the unlawful intent necessary to create such an agreement. We have reviewed this evidence and find it sufficient to show an intent to achieve an unlawful objective on Rule’s part, namely the use of predatory means to monopolize the fluke anchor market. Nevertheless, there is insufficient evidence linking Tie Down to Rule’s efforts to support a finding of conspiracy between them. Federal antitrust law requires a plaintiff to introduce evidence that tends to exclude the possibility that the defendants acted independently or legitimately. Bolt, 891 F.2d at 819; see also Monsanto Co. v. Spray-Rite Serv. Co., 465 U.S. 752, 764, 104 S.Ct. 1464, 1470, 79 L.Ed.2d 775 (1984). U.S. Anchor did not meet this heightened standard of proof. Cf. Boczar v. Manatee Hosps. & Health Sys., Inc., 993 F.2d 1514, 1518-19 (11th Cir. 1993) (finding sufficient evidence when defendant’s supposed legitimate reasons for acting were shown to be fabricated and contrived).28 The MacKarvich market report, USTX 683, for instance, does not show that Tie Down desired to employ predatory means to drive U.S. Anchor from the market. Rather, it merely shows the prices at which it would be possible to inflict losses on the newcomer. It says nothing about Rule’s costs, and U.S. Anchor does not dispute that Tie Down had no knowledge of Rule’s costs other than the price paid for anchors. Tie Down’s experts and MacKarvich himself testified that such studies are common in competitive industries and consistent with legitimate competition based on price. MacKarvich’s recommendation to set prices low enough to inflict losses on U.S. Anchor merely shows a desire to win on the basis of efficiently producing a product and selling it at a lower price than less efficient rivals. It is not unlawful to slash prices in an attempt to obtain more sales, even if the result is that a competitor happens to be driven out of business. Ball Mem. Hosp., Inc. v. Mutual Hosp. Ins., Inc., 784 F.2d 1325, 1338-39 (7th Cir. 1986). Moreover, to suffer a loss in the first year of operation is common in competitive industries, and for MacKarvich to anticipate that U.S. Anchor would be temporarily unprofitable does not necessarily show a desire or expectation that the firm would be driven from the marketplace. In short, we have examined the record closely and find there is insufficient evidence linking Tie Down with Rule’s scheme to constitute a conspiracy under the substantive proof requirements of federal antitrust law.

Without Tie Down, there was no one with whom Rule could have conspired. Hence, its unilateral conduct was not actionable as a conspiracy under federal antitrust law. The district court erred in denying judgment as a matter of law for Rule and Tie Down on the Sherman Act conspiracy claims.

VII. CLAIMS UNDER GEORGIA LAW

U.S. Anchor’s complaint alleged violations of article III, § VI, ¶ 5 of the Georgia constitution and O.C.G.A. § 13-8-2(a)(2), which invalidate certain contracts in restraint of trade. (Rl-1, ¶¶ 60-62.) U.S. Anchor concedes that these provisions merely render such agreements unenforceable and provide no cause of action for damages to those who are parties thereto, see E.T. Barwick Indus. v. Walter E. Heller & Co., 692 F.Supp. 1331, 1349 (N.D.Ga. 1987), but argues that Georgia recognizes a common law tort action in favor of third parties who are injured by a conspiracy in restraint of trade. We agree with U.S. Anchor that its complaint stated a valid claim for damages as a result of a conspiracy in restraint of trade. See Blackmon v. Gulf Life Ins. Co., 179 Ga. 343, 175 S.E. 798, 802-03 (1934) (holding that allegations of predato*1003ry pricing conspiracy with intent to monopolize stated a cause of action); Atlanta Association of Fire Ins. Agents v. McDonald, 181 Ga. 105, 181 S.E. 822, 828 (1935) (awarding nominal damages and injunction for group boycott); see also Harrison Co. v. Code Revision Comm'n, 244 Ga. 325, 260 S.E.2d 30, 34 (1979). The district court erred in failing to perceive "the distinction between a contract or agreement merely in restraint of trade as between the parties, and a combination or contract to stifle competition, or a conspiracy to ruin a competitor." Brown v. Jacobs Pharmacy Co., 115 Ga. 429, 41 S.E. 553, 556 (1902) (suit for damages and injunction). Although we have found insufficient evidence of a conspiracy under federal law standards, this does not answer the question of whether Georgia courts would find sufficient evidence of conspiracy under their substantive law. Cf. Sachdeva v. Smith, 167 Ga.App. 80, 306 S.E.2d 19, 20 (1983).

We have previously held that Georgia law provides a cause of action for tortious interference with the business relationships between a plaintiff and its customers, suppliers or representatives. To be held liable the defendant "must have (1) acted improperly and without privilege, (2) purposely and with malice with the intent to injure, (3) induced a third party or parties not to enter into or continue a business relationship with the plaintiff, and (4) [caused] plaintiff [to] suffer[] some financial injury." DeLong Equip. Co. v. Washington Mills Abrasive Co., 887 F.2d 1499, 1518 (11th Cir. 1989) (quotation omitted), cert. denied, 494 U.S. 1081, 110 S.Ct. 1813, 108 L.Ed.2d 943 (1990), appeal after remand, 990 F.2d 1186 (11th Cir. 1993), amended, 997 F.2d 1340 (11th Cir. 1993) (per curiam); see also NAACP v. Overstreet, 221 Ga. 16, 142 S.E.2d 816, 822 (1965), cert. dismissed, 384 U.S. 118, 86 S.Ct. 1306, 16 L.Ed.2d 409 (1966). The defendant may show that competitive conduct is privileged by establishing that it used no improper means. Integrated Micro Sys., Inc. v. NEC Home Elecs. (USA), Inc., 174 Ga.App. 197, 329 S.E.2d 554, 559 (1985), cart. denied, No. 69405 (Ga. Apr. 24, 1985).

U.S. Anchor's complaint adequately pleads a claim for relief under this theory to present it for adjudication by the district court. Count V gave full notice to the defendants that U.S. Anchor sought recovery under Georgia law for "Unfair Methods of Competition and Unfair Acts and Practices," including conduct which was "inequitable, unfair, unscrupulous, in violation of public policy and unconscionable and tend[ing] to defeat or lessen competition. ." (Ri-i ¶11 59-60.) The fact that paragraph 60 of the complaint also refers to the constitutional and statutory provisions which U.S. Anchor concedes confer no independent damages remedy does not by itself deprive the defendants of "fair notice of what the plaintiffs claim is and the grounds upon which it rests." Quality Foods de Centro Am., S.A. v. Latin Am. Agribusiness Dev. Corp., 711 F.2d 989, 995 (11th Cir. 1983) (quoting Conley v. Gibson, 355 U.S. 41, 47, 78 S.Ct. 99, 103, 2 L.Ed.2d 80 (1957)); see Fed. R.Civ.P. 8(a)(2). The issue of whether the tort theory is applicable to the facts of this case was adequately argued to the district court in connection with U.S. Anchor's requested jury instructions, (R28-136-45), and thus preserved for appellate review. Cf Weaver v. Casa Gallardo, Inc., 922 F.2d 1515, 1519 (11th Cir. 1991).

The novel questions presented are whether below-cost pricing can satisfy the improper action element of the tort and whether low prices, standing alone, can constitute a prohibited inducement of the plaintiffs customers. Cf. Parks v. Atlanta News Agency, Inc., 115 Ga.App. 842, 156 S.E.2d 137, 140 (1967) (holding that solicitation of competitor's customers is not itself tortious, even when combined with "preferential" prices), cert. denied, No. 42624 (Ga. July 14, 1967). We regard it as unclear whether tortious interference with business relations under Georgia law may be established by a showing of predatory pricing and, if so, what sort of pricing conduct would be deemed predatory. We also have some doubt as to whether intentional interference with business relations is a distinct cause of action from the tort of conspiracy in restraint of trade, or whether there is only a single theory of relief, so that proof of a conspiracy to interfere with the plaintiffs business relations *1004would be actionable as U.S. Anchor’s sole remedy for the alleged joint conduct of Rule and Tie Down. Compare Cook v. Robinson, 216 Ga. 328, 116 S.E.2d 742 (1960) with Jacobs Pharmacy, 41 S.E. at 554-57 (quoting Doremus v. Hennessy, 176 Ill. 608, 52 N.E. 924 (1898)); see also Overstreet, 142 S.E.2d at 822. This is not a matter of mere semantics, for while it appears settled that predatory pricing by a group or conspiracy is actionable, we have found no Georgia authority addressing predation by a single defendant acting unilaterally.

Another issue affecting the outcome of U.S. Anchor’s state law claims is the validity and effect of its settlement agreement with Rule, executed on March 19, 1986. The agreement provided that each party would release the other

from any and all actions, demands, claims or causes of action whatsoever, which now exist or which may arise in the future, as a result of events which occurred prior to the execution of the Settlement Agreement, including, without limitation, any claims which were or could have been presented by way of complaint or counterclaim in Civil Action Number C85^1466A.

(RTX 457.) Because the predatory pricing scheme allegedly began in October 1985, Rule contends that the settlement agreement operated as a release of U.S. Anchor’s cause of action. U.S. Anchor contends that its predatory pricing claims were undiscovered at the time the release was executed and therefore were not intended to be released. In addition, it contends that injuries caused by predatory conduct occurring after the release would not have been discharged even if they arose as a result of a scheme or conspiracy that was ongoing when the release was signed.29 The district court concluded that the federal predatory pricing claims were undischarged because the agreement unambiguously applied only to causes of action related to the prior litigation. It also relied on Chapman’s oral testimony concerning his intent at the time he signed the agreement and on Covington v. Brewer, 101 Ga.App. 724, 115 S.E.2d 368, 372-73 (1960), in which the court held that the scope of a release as intended by the parties could not be presumed to encompass rights respecting a subject matter not clearly referred to in the body of the agreement. But cf. Ingram Corp. v. J. Ray McDermott & Co., 698 F.2d 1295, 1311-12 (5th Cir. 1983).

The doctrine of pendent jurisdiction as outlined in United Mine Workers v. Gibbs, 383 U.S. 715, 86 S.Ct. 1130, 16 L.Ed.2d 218 (1966), gives the district court power to decide claims arising under state law as to which there was no independent basis for federal jurisdiction but which share a common nucleus of operative fact with federal claims. The court also has discretion not to hear such state law claims.

Under Gibbs, a federal court should consider and weigh, in each case, and at every stage of the litigation, the values of judicial economy, convenience, fairness, and comity in order to decide whether to exercise jurisdiction over a case brought in that court involving pendent state-law claims. When the balance of these factors indicates that a case properly belongs in state court, as when the federal-law claims have dropped out of the lawsuit in its early stages and only state-law claims remain, the federal court should decline to exercise its jurisdiction by dismissing the case without prejudice.

Carnegie-Mellon Univ. v. Cohill, 484 U.S. 343, 350, 108 S.Ct. 614, 619, 98 L.Ed.2d 720 (1988) (footnote omitted). While the doctrine is a flexible one according great leeway to the court, see id. at 350 n. 7, 108 S.Ct. at 619 n. 7, we have found an abuse of discretion in failing to dismiss a case when the federal claims were resolved early in the proceedings *1005and the state law claims posed issues of first impression. See Hardy v. Birmingham, Bd. of Educ., 954 F.2d 1546 (11th Cir. 1992).

In the present case, the federal claims have survived through trial and have only been resolved on appeal. Thus, the parties have already tried the state law claims in federal court, although the district court’s ruling prevented the jury from considering them. The legal issues have been decided by the district court and are now properly before us for review, so that judicial economy and convenience weigh in favor of retaining jurisdiction. On the other hand some of the state law issues are novel, and comity between federal and state judicial systems weighs in favor of determination by state courts. Moreover, a ruling by this court in favor of U.S. Anchor’s position would require a new federal trial in which only state law claims would be put in issue. Fairness to U.S. Anchor, however, prevents us from dismissing the state law claims. Dismissal would require the plaintiff to re-file its action in state court more than eight years after the allegedly tortious conduct began, thereby losing a substantial portion of its rights (if any) by application of Georgia’s four-year statute of limitations.30 We might have reached a different result under the Judicial Improvements Act- of 1990, Pub.L. No. 101-650, § 310, 104 Stat. 5089, 5113-14, codified at 28 U.S.C. § 1367. Under 28 U.S.C. § 1367(d), the statute of limitations would be tolled while the claims were pending until 30 days after an order of dismissal, thus allowing the plaintiff time for filing a new action in state court without a lapse of its rights. But the present case was commenced before the statute’s effective date on December 1, 1990, and § 1367 is not retroactive. Yanez v. United States, 989 F.2d 323, 327 n. 3 (9th Cir. 1993). In view of the fact that the case may be certified to the Supreme Court of Georgia for interlocutory resolution of the state law issues, we conclude that the balance of factors involved in the discretionary decision to retain pendent jurisdiction weighs clearly against dismissal.31

Accordingly, we respectfully certify the following questions of law to the Supreme Court of Georgia and the Honorable Justices of that Court.

Questions for Certification

1. DOES A GENERAL RELEASE UNDER GEORGIA LAW DISCHARGE LIABILITY FOR INJURY CAUSED BY SUBSEQUENT ACTS IN THE COURSE OF A SCHEME OR CONSPIRACY THAT WAS ONGOING AT THE TIME THE RELEASE WAS EXECUTED BUT UNKNOWN TO THE RELEASING PARTY?

2. DOES A GENERAL RELEASE UNDER GEORGIA LAW DISCHARGE LIABILITY FOR INJURY CAUSED BY TOR-TIOUS CONDUCT ALREADY COMMITTED THAT WAS UNKNOWN TO THE *1006RELEASING PARTY AT THE TIME THE RELEASE WAS EXECUTED?

3. DOES THE TORT OF INTENTIONAL INTERFERENCE WITH BUSINESS RELATIONS ENCOMPASS PREDATORY PRICING BELOW SOME MEASURE OF THE DEFENDANT’S COSTS?

4. IF THE ANSWER TO QUESTION 3 IS YES, THEN IN A CASE OF ACTIONABLE PREDATORY PRICING BELOW SOME MEASURE OF COST BY A CONSPIRACY OR A SINGLE DEFENDANT, WHAT IS THE APPROPRIATE MEASURE OF THE DEFENDANTS’ COSTS?

Our statement of the questions is not designed to limit the inquiry of the Supreme Court of Georgia. Instead, the Supreme Court has the widest possible latitude to consider the problems and issues involved in this case as it perceives them to be. Martinez v. Rodriquez, 394 F.2d 156, 159 n. 6 (5th Cir. 1968), conformed to certified answer, 410 F.2d 729 (5th Cir. 1969). To assist the Supreme Court, the entire record in this case and copies of the parties’ briefs are transmitted herewith.

VIII. CONCLUSION

The judgment of the district court is reversed with respect to all federal law causes of action and judgment is rendered in favor of the defendants thereon. Dispositive questions of law respecting the plaintiffs state law causes of action are certified to the Supreme Court of Georgia.

REVERSED and JUDGMENT RENDERED in part and QUESTIONS CERTIFIED.