In the

United States Court of Appeals For the Seventh Circuit ____________________ No. 22-3279 ARANDELL CORPORATION, et al., Plaintiffs-Appellees, v.

XCEL ENERGY INC., et al., Defendants-Appellants. ____________________

Appeal from the United States District Court for the Western District of Wisconsin. Nos. 3:07-cv-00076 & 3:09-cv-00240 — James D. Peterson, Chief Judge. ____________________

ARGUED APRIL 5, 2023 — DECIDED AUGUST 5, 2025 ____________________

Before SYKES, Chief Judge, and HAMILTON and BRENNAN, Circuit Judges. HAMILTON, Circuit Judge. This interlocutory appeal challenges certification of a statewide plaintiff class in a price- fixing case brought under Wisconsin law. The case arises from a broad conspiracy to manipulate natural gas prices from 2000 to 2002. The central issue is whether common issues predominate over individual issues under Federal Rule of Civil Procedure 23(b)(3) so that class certification is 2 No. 22-3279

appropriate. To be more precise, the question here is whether the district court had a sound basis for answering that question “yes” without addressing in more detail the conflicting expert testimony on the issue of antitrust impact. It is well recognized that price fixing causes antitrust injury within the meaning of Brunswick Corp. v. Pueblo Bowl-O-Mat, Inc., 429 U.S. 477, 4888 (1977) (plaintiffs could not recover damages for alleged antitrust violation on theory that violation enabled more competitors to survive in relevant market); see generally Robert H. Bork, The Antitrust Paradox 67 (1978) (“The per se rule against naked price fixing and similar agreements not to compete is the oldest and clearest of antitrust doctrines, and its existence can be explained only by a preference for consumer welfare as the exclusive goal of antitrust.”). Higher fixed prices are the result of the illegal conduct. In this case, therefore, the question of antitrust impact is primarily an issue of causation and the scope of the geographic market. As we explain below, while that issue overlaps substantially with the merits of plaintiffs’ claims, that overlap does not allow the district court to postpone dealing with the issue at the class certification stage. The existence of a conspiracy here has been proven in other proceedings. The existence and scope of the alleged con- spiracy are certainly subject to common proof. It is also well established that individual questions of damages should not defeat class certification. The pivotal question here is whether plaintiffs can use common proof to show antitrust impact from the conspiracy. That will depend on whether they can show the existence of a national market such that defendants’ conspiracy to manipulate prices affected the prices plaintiffs paid for gas in Wisconsin. As a general rule, class certification is common in price-fixing cases, see 7AA Wright & Miller, No. 22-3279 3

Federal Practice & Procedure § 1781 & n.22 (3d. ed. 2025), par- ticularly where the product involved is a fungible commodity like gas. See Messner v. Northshore University HealthSystem, 669 F.3d 802, 816 (7th Cir. 2012) (showing antitrust impact of price-fixing is relatively simple in market for generic com- modity). Nevertheless, defendants have offered evidence that they contend shows that natural gas price markets were so complex and unusual as to prevent such common proof. Plaintiffs’ economic experts have responded with detailed re- buttals. The district court deferred full engagement with those de- bates at the class certification stage, leaving further consider- ation for the merits of the case. In light of recent decisions by the Supreme Court and this court, however, we conclude that the district court needed to engage more fully with the con- flicting expert evidence to decide on class certification. While we would not be surprised if the case turns out to be proper for class certification—this is after all a price-fixing case in- volving a fungible commodity traded across a nationwide network of pipelines—we must vacate the certification and re- mand for further consideration of expert disputes. I. Factual Background A. The Market for Natural Gas Natural gas is a fungible commodity. One molecule is in- distinguishable from another. It is transported around the United States by a nationwide network of pipelines. Most consumers of natural gas in the United States are individuals and small businesses who buy gas from local utilities. Arandell Corp. v. Centerpoint Energy Services, Inc., 900 F.3d 623, 625–26 (9th Cir. 2018) (earlier appeal reversing denial of class 4 No. 22-3279

certification in larger multidistrict litigation). But larger con- sumers, including this case’s plaintiffs, contract to buy large volumes of gas directly from the gas companies themselves or through intermediaries. Different pricing arrangements are common, including fixed-price contracts and contracts with variable prices tied to an index. Contract durations also vary. At the relevant time, gas companies would report their sales to trade publications, such as Inside FERC, Gas Daily, and NYMEX. Those trade publications used that data to publish index prices that would directly affect some contract prices tied to those indices. Fixed-price contracts were also affected, as buyers and sellers negotiated with the index price as a benchmark. See Arandell, 900 F.3d at 626 & n.1. B. The Manipulation of the Market In the early 2000s, natural gas prices climbed “to extraor- dinary levels.” Oneok, Inc. v. Learjet, Inc., 575 U.S. 373, 382 (2015) (internal quotations omitted), quoting FERC, Final Re- port on Price Manipulation in Western Markets: Fact-Finding In- vestigation of Potential Manipulation of Electric and Natural Gas Prices (hereafter “FERC Report”) at ES-1 (Mar. 2003) (https://perma.cc/55SP-F8ST). The Federal Energy Regulatory Commission (FERC) investigated and uncovered tell-tale signs of market manipulation: churning, wash trading, and false reporting. See FERC Report at ES-5, ES-11, & I-18. “Churning” refers to buying and selling gas during the same trading interval such that the trades offset each other. Arandell, 900 F.3d at 626 n.3. “Wash trading” refers to arranging a “pair of trades of the same good between the same parties, involving no economic risk and no net change in beneficial ownership.” Id. at n.2 (internal quotation marks No. 22-3279 5

omitted). “False reporting” refers to reporting such churning, wash trading, and other fabricated trading data to the trade publications that set the index price. Id. at 626–27. Unknown to outside observers, this conduct masqueraded as legitimate market activity. But as known by many inside the industry, according to plaintiffs, these tactics among com- petitors really amounted to mechanisms to fix prices. FERC’s investigation concluded that these price increases were driven by some of the nation’s largest natural gas conglomerates, in- cluding Enron and defendants in this case. Id. at 626. These practices can function as covert mechanisms for price-fixing. By engaging in churning and wash trading, market participants can report what amount to fictional prices, creating the illusion of supply and demand dynamics that do not actually exist. A higher trading volume signals a higher demand for natural gas and vice versa. When this trading data was reported to price index publishers, it artificially inflated—or at times deflated—the benchmark prices used to set rates in many natural gas contracts. In doing so, competitors were not simply engaging in misleading conduct; they were collectively manipulating the market to produce prices that would not have applied under competitive conditions. C. Penalties and Remedies for Manipulating the Market The defendants’ conduct attracted attention from FERC, the Securities and Exchange Commission, the Commodities Futures Trading Commission, and the Department of Justice, leading to government investigations, administrative orders, and criminal indictments. Those proceedings resulted in civil penalties, criminal convictions, prison sentences, and heavy 6 No. 22-3279

financial sanctions. This case is a slice of the private civil liti- gation over the market manipulation. The point is that the price-fixing conspiracy here was not merely alleged. That does not mean the extent of the conspiracy has been deter- mined, but its existence has been proven again and again. Horizontal price-fixing violates both federal and Wisconsin antitrust law under the Sherman Act, 15 U.S.C. § 1, and Wis. Stat. § 133.03. The Wisconsin Supreme Court has long made clear that federal interpretations of the Sherman Act should guide interpretation of the Wisconsin Act. E.g., Conley Publishing Group Ltd. v. Journal Communications, Inc., 2003 WI 119, ¶ 17, 265 Wis. 2d 128, 140–41, 665 N.W.2d 879, 885–86, abrogated on other grounds, Olstad v. Microsoft Corp., 2005 WI 121, ¶ 74, 284 Wis. 2d 224, 259, 700 N.W.2d 139, 156; State v. Waste Management of Wisconsin, Inc., 261 N.W.2d 147, 153 n.12 (Wis. 1978); City of Madison v. Hyland, Hall & Co., 243 N.W.2d 422, 428–29 (Wis. 1976). The state law closely tracks the language of the federal law as to the conduct prohibited. There are important differences, however, in the available civil remedies. Under Wisconsin law, if a contract has been made by a party while it is a member of a combination or conspiracy out- lawed under § 133.03 and the contract is connected with the violation, the contract “shall be void.” Wis. Stat. § 133.14. A party who made a payment under the contract may recover “any payment” made under the contract from any person who received or benefited from such payment. Id. This “full consideration” remedy is not available under federal antitrust law. It appears to have been designed by the Wisconsin legis- lature to offer a streamlined and relatively simple remedy. No. 22-3279 7

Federal law offers a treble-damage civil remedy for anti- trust violations, 15 U.S.C. § 15(a), and Wisconsin also offers a treble-damage remedy as a supplement or alternative to the “full consideration” remedy, Wis. Stat. § 133.18(1)(a). Federal law limits that remedy to direct purchasers from violators; in- direct purchasers may not recover damages. Illinois Brick Co. v. Illinois, 431 U.S. 720, 746–47 (1977). But Wisconsin wrote its antitrust law to reject the federal Illinois Brick doctrine, ex- pressly permitting suit by “any person injured, directly or in- directly.” Olstad, 700 N.W.2d 139, ¶61 (alteration omitted), cit- ing Wis. Stat. § 133.18. II. Procedural Background A. Allegations of an Antitrust Conspiracy Numerous plaintiffs—industrial and commercial pur- chasers of natural gas—sued the defendant gas companies, al- leging that the plaintiffs paid higher gas prices because each defendant participated in a price-fixing conspiracy. Under 28 U.S.C. § 1407, various actions around the country were trans- ferred to the District of Nevada under a multidistrict litigation docket for consolidated pretrial proceedings. The cases in- cluded actions on behalf of class members in Kansas, Mis- souri, Colorado, and Wisconsin. In re Western States Wholesale Natural Gas Antitrust Litig., MDL No. 1566, 2017 WL 1243135, at *10–14 (D. Nev. Mar. 30, 2017) (denying class certification in several cases, including this Wisconsin case), vacated and remanded, 743 F. App’x 825, 828–30 (9th Cir. 2018). After re- mand from the Ninth Circuit, the MDL transferee court re- turned this Wisconsin case to the originating court, the West- ern District of Wisconsin. After nearly twenty years of litiga- tion, we are still dealing with the threshold issue of class cer- tification. 8 No. 22-3279

The Wisconsin plaintiffs in this case assert antitrust claims under only Wisconsin antitrust law, not federal antitrust law. See Wis. Stat. § 133.01 et seq. They pursue both the “full consideration” remedy under § 133.14 and treble damages under § 133.18(1)(a). B. The Expert Disputes After transfer back to the Western District of Wisconsin, the Wisconsin plaintiffs moved again to certify the same Rule 23(b)(3) class of industrial and commercial purchasers of natural gas for their own use or consumption in Wisconsin from January 1, 2000, to October 21, 2002, excluding certain other purchasers. The plaintiffs proposed common questions of law and fact, including whether the defendants conspired to manipulate the price of natural gas through wash trading, churning, and false reporting; whether the class paid higher prices for natural gas as a result; the method to determine the determine recovery for full consideration; and the method to determine recovery for treble damages. In support, the plaintiffs relied on the same expert evidence providing economic models to calculate antitrust impact and actual damages for the treble damage remedy. This expert evidence is the focus of this appeal regarding whether the expert models established predominance. The plaintiffs’ experts, Dr. Michael Harris and Dr. Mark Dwyer, began by describing the nature of the natural gas market. Natural gas, they explained, is a fungible commodity sold at standardized prices and transported through a nationwide network of pipelines. R.208-18 at 28, ¶ 59. The natural gas market is nationwide, as shown by tightly correlated prices from region to region and index to index. R.208-12 at 26, ¶ 59; R.208-13 at 14 ¶ 34. Given these market No. 22-3279 9

fundamentals, they explained that “it can reasonably be concluded every area of the country was impacted by the manipulation.” R.208-12 at 9, ¶ 17. In that regard, the plaintiffs sought to measure impact and damages with regression models and a “but-for price index.” R.208-12 at 28. That is, the experts calculated what they believe the index prices would have been without the conduct of the defendants and non-defendant co-conspirators. After controlling for location and price type, the plaintiffs’ experts concluded that “all, or nearly all of the class members in Wisconsin were impacted by Defendants’ conduct.” R.208-12 at 39, ¶ 79. The defendants’ experts, Dr. Michelle Burtis and Dr. Ran- dall Heeb, raised numerous objections to plaintiffs’ experts’ opinion. See Dkt. 213-1 & 213-3 (Burtis); Dkt. 213-4 (Heeb). They opined that the plaintiffs’ models fail to prove a single, interrelated market for natural gas, fail to account for varieties in contract terms and data indicating manipulated prices sometimes fell, and in other ways do not evaluate or sample the correct data. R.210 at 30. Plaintiffs’ experts, Dr. Harris and Dr. Dwyer, responded to these criticisms with detailed rebuttal reports. Dkt. 225-6, 225-9, & 225-12. They argued that the defense experts had, in essence, raised a number of red herrings to try to complicate and obscure what is at bottom a fairly straightforward price- fixing case involving a fungible commodity in a national mar- ket. Indeed, among price-fixing defendants in fungible com- modity cases, such efforts to defeat class certification by fo- cusing on varieties of related products and contracts appear to be quite common. See, e.g., Kleen Products LLC v. Int’l Paper Co., 831 F.3d 919, 927–29 (7th Cir. 2016) (affirming certification 10 No. 22-3279

of nationwide plaintiff class in containerboard price-fixing case relying on opinions of Dr. Harris and Dr. Dwyer to show antitrust impact); In re Ready-Mix Concrete Antitrust Litig., 261 F.R.D. 154, 170–72 (S.D. Ind. 2009) (variations in concrete products did not defeat class certification in regional price- fixing class action); In re Bromine Antitrust Litig., 203 F.R.D. 403, 414–15 (S.D. Ind. 2001) (same in nationwide price-fixing case for bromine products); In re Auction Houses Antitrust Litig., 193 F.R.D. 162, 166–67 & n.13 (S.D.N.Y. 2000) (same in nationwide price-fixing case for auction services: “Price fixing conspiracies, at least to the extent they succeed in fixing prices, almost invariably injure everyone who purchases the relevant goods or services. Defendants appear greatly to ex- aggerate the extent to which individualized proof of impact may be required.”) (footnote omitted); In re Industrial Dia- monds Antitrust Litig., 167 F.R.D. 374, 383 (S.D.N.Y. 1996) (same in nationwide price-fixing case for industrial diamond products); In re Linerboard Antitrust Litig., 203 F.R.D. 197, 216– 20 (E.D. Pa. 2001) (certifying nationwide classes in case alleg- ing conspiracy to restrict output of linerboard), aff’d, 305 F.3d 145 (3d Cir. 2002). C. The Class Certification Order Judge Conley, to whom the case was originally assigned, agreed with the defendants that although “these experts’ opinions are admissible, [that] does not mean plaintiffs have established the existence of a common method of proof.” Arandell Corp. v. Xcel Energy, Inc., No. 07-CV-076-WMC, 2022 WL 2314717, at *9 (W.D. Wis. June 28, 2022). The court continued: “the finding that these experts are qualified to render an opinion that is sufficiently reliable and will aid in the jury’s determinations may well support a finding that No. 22-3279 11

plaintiffs have presented a credible theory for pursuing their claims on behalf of a larger class.” Id. Judge Conley found that whether a conspiracy to raise prices existed and whether there was a nationwide market susceptible to such a conspiracy were common questions subject to common proof. This finding supported the court’s finding of predominance: “Given the significant issues that can be resolved at the class level based on plaintiffs’ theory of liability and the evidence of a single, nationwide natural gas market due to defendants’ allegedly illicit actions, the court concludes that the predominance requirement is satisfied.” Id. at *12. Judge Conley acknowledged that the defendants “point[ed] to differences between both the plaintiffs and defendants,” but he did not expressly address or make findings about the defense arguments that the plaintiffs’ models were inadequate so as to defeat predominance. Id. at *10. Judge Conley certified the proposed plaintiff class but soon after that recused from further proceedings in the case. Chief Judge Peterson, to whom the case was reassigned, quickly denied the defendants’ motion to vacate the class cer- tification but encouraged this interlocutory appeal. We granted leave to appeal class certification under Federal Rule of Civil Procedure 23(f), limiting the interlocutory appeal to “whether the district court properly assessed the parties’ ex- pert opinions when deciding the motion to certify.” We re- view a class certification order for an abuse of discretion, which can include reviewing whether a district court commit- ted a legal error by conducting an insufficiently “rigorous analysis.” Santiago v. City of Chicago, 19 F.4th 1010, 1016 (7th Cir. 2021). 12 No. 22-3279

III. Class Certification, Rigorous Analysis, and Resolving Expert Disputes A. Requirements for Class Certification The basics of class certification are not disputed here. To be certified, a proposed class must first meet “the Rule 23(a) requirements of numerosity, typicality, commonality, and ad- equacy of representation.” Messner, 669 F.3d at 811. The dis- trict court found all four requirements under Rule 23(a) were met here, and there is no dispute about them on appeal. The class must then fit into one of three categories under Rule 23(b). To certify a plaintiff class seeking damages under Rule 23(b)(3), as plaintiffs seek here, “a court must find ‘that the questions of law or fact common to class members predomi- nate over any questions affecting only individual members ….” Fed. R. Civ. P. 23(b)(3). A common question is one where “the same evidence will suffice for each member to make a prima facie showing [or] the issue is susceptible to generalized, class-wide proof.” Ty- son Foods, Inc. v. Bouaphakeo, 577 U.S. 442, 453 (2016) (internal quotations omitted), quoting 2 W. Rubenstein, Newberg on Class Actions § 4:50, p. 196 (5th ed. 2012). An individual ques- tion is one where “’members of a proposed class will need to present evidence that varies from member to member.’” Id., quoting Rubenstein, supra, § 4:50, p. 196. To determine which questions are common versus individual, the district court must circumscribe the claims and break them down into their constituent elements. Santiago, 19 F.4th at 1018. In so doing, the district court must separately analyze the claims “to better understand the relationship between each claim’s common and individual questions.” Eddlemon v. Bradley University, 65 F.4th 335, 339–40 (7th Cir. 2023). No. 22-3279 13

With common and individual questions identified, the district court must determine whether common questions predominate. Predominance is a “qualitative rather than quantitative concept.” Parko v. Shell Oil Co., 739 F.3d 1083, 1085 (7th Cir. 2014). “There is no mathematical or mechanical test for evaluating predominance.” Messner, 559 F.3d at 814. It is not determined “simply by counting noses: that is, deter- mining whether there are more common issues or more indi- vidual issues….” Parko, 739 F.3d at 1085. More than a “tally of common questions,” the district court must “give careful scrutiny to the relation between common and individual questions in a case,” and “consider their relative importance.” Tyson, 577 U.S. at 453 (second quotation); Santiago, 19 F.4th at 1016 (first and third quotations); accord, 2 Rubenstein, New- berg on Class Actions § 4:51, pp. 198–200. B. “Rigorous Analysis” and the Delicate Balance The trial court can certify a class only if it is “’satisfied, af- ter a rigorous analysis, that the prerequisites’ for class certifi- cation have been met.” Bell v. PNC Bank, N.A., 800 F.3d 360, 373 (7th Cir. 2015), quoting CE Design Ltd. v. King Architectural Metals, Inc., 637 F.3d 721, 723 (7th Cir. 2011). The Supreme Court has made it clear that class certification imposes more than a pleading standard. Wal-Mart Stores, Inc. v. Dukes, 564 U.S. 338, 350 (2011). “The party seeking certification bears the burden of demonstrating that certification is proper by a pre- ponderance of the evidence.” Bell, 800 F.3d at 373. At the same time, the Supreme Court has long cautioned lower courts to focus on the requirements of Rule 23 and not to “conduct a preliminary inquiry into the merits of a suit in order to determine whether it may be maintained as a class action.” Eisen v. Carlisle & Jacquelin, 417 U.S. 156, 177 (1974). 14 No. 22-3279

The Eisen principle does not mean, however, that there is a solid wall between class certification and the merits, as more recent cases have made clear. The Court has recognized that there will often be some overlap between class issues and merits. The overlap does not allow a district court to skip over an issue critical to certification just because it overlaps with the merits. See Goldman Sachs Group, Inc. v. Arkansas Teacher Retirement System, 594 U.S. 113, 122 (2021) (issue of price impact in securities fraud cases), citing In re Allstate Corp. Securities Litig., 966 F.3d 595, 613 n.6 (7th Cir. 2020); see generally Amgen Inc. v. Connecticut Retirement Plans and Trust Funds, 568 U.S. 455, 466 (2013) (securities fraud: “Rule 23 grants courts no license to engage in free-ranging merits inquiries at the certification stage. Merits questions may be considered to the extent—but only to the extent—that they are relevant to determining whether the Rule 23 prerequisites for class certification are satisfied.”); Halliburton Co. v. Erica P. John Fund, Inc., 573 U.S. 258, 283 (2014) (securities fraud: price impact “has everything to do with the issue of predominance at the class certification stage”); Dukes, 564 U.S. at 351–52 (noting necessity of touching on merits issues in employment discrimination class certification: “The necessity of touching aspects of the merits in order to resolve preliminary matters, e.g., jurisdiction and venue, is a familiar feature of litigation.”); General Telephone Co. of Southwest v. Falcon, 457 U.S. 147, 161 (1982) (“rigorous analysis” required under Rule 23). This general principle applies to antitrust cases. Comcast Corp. v. Behrend, 569 U.S. 27, 34–35 (2013) (reversing class cer- tification where lower courts refused to entertain challenges to plaintiffs’ damages model as undermining predominance of common issues; overlap with merits did not justify refusal No. 22-3279 15

to engage with defense evidence and arguments); In re Hydro- gen Peroxide Antitrust Litig., 552 F.3d 305, 322–25 (3d Cir. 2008) (vacating class certification where district court relied on ad- missibility of plaintiffs’ expert evidence but did not engage with evidence to resolve relevant disputes with defense ex- perts). We have noted before that the Supreme Court’s teachings call for district judges to perform a careful balancing act. See Allstate, 966 F.3d at 603 (securities fraud). In conducting its rigorous analysis, the court is not permitted to turn the class certification proceedings into a “dress rehearsal for the trial on the merits.” Messner, 669 F.3d at 811. “[C]ertification is largely independent of the merits … and a certified class can go down in flames on the merits.” Schleicher v. Wendt, 618 F.3d 679, 685 (7th Cir. 2010). The trial court is not permitted to re- solve disputed issues that affect only the merits. A court is permitted to “‘take a peek at the merits before certifying a class,’ but the peek must be ‘limited to those aspects of the merits that affect the decisions essential under Rule 23.’” Messner, 669 F.3d at 823–24, quoting Schleicher, 618 F.3d at 685. As we will see, this appeal involves much more than a “peek,” but the work must still be done. In the face of material disputes bearing on class certifica- tion, the trial court must receive evidence, whether by affida- vit, evidentiary hearings, or otherwise, and then resolve the disputes or choose between competing perspectives. West v. Prudential Securities, Inc., 282 F.3d 935, 938 (7th Cir. 2002) (re- versing class certification when court did not address dispos- itive merits issue at certification stage); Szabo v. Bridgeport Ma- chines, Inc., 249 F.3d 672, 676 (7th Cir. 2001) (vacating and re- manding class certification; district court must make 16 No. 22-3279

“whatever factual and legal inquiries are necessary under Rule 23). The trial court must do its best to maintain a delicate balance “between evaluating evidence to determine whether a common question exists and predominates, without weigh- ing that evidence to determine whether the plaintiff class will ultimately prevail on the merits.” Bell, 800 F.3d at 377. But the trial court is not permitted to “duck hard questions by observ- ing that each side has some support, or that considerations relevant to class certification also may affect the decision on the merits. Tough questions must be faced and squarely de- cided …,” even if the dispute regards expert evidence. West, 282 F.3d at 938. C. Resolving Expert Disputes As the district court recognized here, in the context of ex- pert evidence, Rule 23’s “rigorous analysis” requirement for class certification is not the same as the test for admissibility under Federal Rule of Evidence 702 and Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993), and its progeny. Expert evidence can be admissible under Rule 702 and Daub- ert but still fall short of proving the Rule 23 requirements for class certification. Comcast, 569 U.S. at 32–33, n.4 (distinguish- ing between admissibility and adequacy of evidence). In other words, just because expert evidence is admissible does not mean that it will justify class certification. At class certification, a district court is not required to re- solve every expert dispute. A court is neither required nor permitted at the class stage to resolve disputes on the merits unrelated to the decisions essential to Rule 23. However, where the defendants have offered admissible evidence that, if credited, would mean individual questions would predom- inate over common questions, then the district court must No. 22-3279 17

“investigate[ ] the realism” of the expert evidence “in light of the defendants’ counterarguments,” and take evidence to that end. Parko, 739 F.3d at 1086. Otherwise, any party would be able to obtain (or defeat) class certification “just by hiring a competent expert.” West, 282 F.3d at 938. Plaintiffs defend the district court’s decision by arguing that at trial they will have to prove a nationwide market as part of their merits case. Their core evidence on that issue will come from Dr. Harris and Dr. Dryer, and defendants’ core ev- idence on that issue will come from Dr. Burtis and Dr. Heeb. Plaintiffs say that, win or lose at trial, the relevant evidence will be common to all class members. We have pointed out that common evidence does not mean a class will win; the class can lose on common evidence. E.g., Schleicher, 618 F.3d at 685. We appreciate the point, but the issue of causation in a national market is also critical to plaintiffs’ case for class certification. The fact that it also overlaps with their case on the merits does not allow the district court to defer resolving the factual issue. IV. Expert Disputes Material to the Class Certification A. The National Market Issue We now come to the crux of the experts’ debates as they affect class certification. Plaintiffs’ case on antitrust impact is straightforward. Natural gas is a fungible commodity that is transported relatively easily across the nation in a network of pipelines. Dr. Harris and Dr. Dwyer studied the natural gas market and vast quantities of data on transactions from the class period. They concluded that the manipulated prices tended to move together across the country, so that price 18 No. 22-3279

manipulations by defendants in, say, Louisiana or Texas af- fected the prices plaintiffs paid for natural gas in Wisconsin. This theory for antitrust impact fits comfortably into eco- nomic theory for this sort of commodity and with case law, which treats horizontal price-fixing cases like this one as ap- propriate for class action status. The defendants argue that the antitrust impact/national market issue is not as straightforward as plaintiffs contend. The defense experts raise a host of issues with the methods and conclusions of the plaintiffs’ experts. They point to the variety of ways in which natural gas is sold and priced, in- cluding spot prices, prices fixed for a month or for several months, or prices indexed to various published prices. They point to the differing lengths of time for which parties would agree on pricing. And they point to what they contend are anomalies in the data indicating that prices did not actually move as predicted by plaintiffs’ experts’ model. They also point to the role of brokers through whom many plaintiffs purchased natural gas. Dkt. 213-1; 213-3; 213-4. These flaws, the defendants argue, undermine the foundation for the Har- ris Dwyer model, namely, a national market in natural gas. Defendants also contend these flaws mean the case will re- quire so much individualized evidence that the district court should not have certified the class. As noted, plaintiffs’ experts have responded to these criti- cisms with detailed rebuttals. In the view of Dr. Harris and Dr. Dwyer, the defense experts have distorted their work and cherry-picked data to find a few anomalies that are easily ex- plained within plaintiffs’ larger economic model. See Dkt. 225-6 & 225-9. No. 22-3279 19

We have reviewed these many detailed reports. In this ap- peal, however, it is not our role to decide which experts have the better of their debates with one another. For the reasons explained above, the district court will need to engage with these debates and make findings of fact, ultimately, as to whether plaintiffs have shown the existence of a national mar- ket in which defendants’ manipulations of prices affected the prices plaintiff class members paid for the natural gas in Wis- consin. B. Guidance and Guardrails for Remand As these issues have been framed for the district court and for us, there is no shortcut that might let the district court avoid full engagement with the economic experts and their many and detailed point-counterpoint debates. That’s the point of Comcast, along with West, Szabo, and Hydrogen Perox- ide, among other cases. We can, however, offer some guide- lines and guardrails for the proceedings on remand. 1. Manipulated Prices Sometimes Declined? The defendants have argued that the evidence shows their conduct occasionally diverged, “consistent with their traders’ divergent interests in raising or lowering prices at different times based on their position in each transaction.” Dkt. 210 at 24. To the extent this evidence is aimed at refuting proof of the alleged conspiracy, the existence of the conspiracy need not be resolved at the class certification stage. To the extent the evidence is aimed at undermining the existence of a na- tional market, that is an issue that does need to be decided at class certification. Plaintiffs’ experts have responded that the evidence in question came from activities of defendants’ trading arms, 20 No. 22-3279

which were modest compared to the conspirators’ overall market activity, and that the market manipulation had the overall effect of driving prices dramatically higher during the class period. We note that if defendants were in fact manipu- lating the market prices to rise most of the relevant time, they would not be entitled to credit for occasionally manipulating the prices downward. E.g., Hawaii v. Standard Oil Co., 405 U.S. 251, 262 n.14 (1972) (where overcharge is proven, “courts will not go beyond the fact of this injury to determine whether the victim of the overcharge has partially recouped its loss in some other way”); In re Nexium Antitrust Litig., 777 F.3d 9, 27 (1st Cir. 2015) (affirming class certification in face of similar argument; collecting cases and saying “antitrust injury occurs the moment the purchaser incurs an overcharge, whether or not that injury is later offset”). Moreover, the FERC Report noted that even if a certain market participant were a net buyer of natural gas, unlawful trading activity could increase the appearance of volatility, raising prices and enabling a trader to work in agreement with others to offload its unneeded gas at higher prices. FERC Report at II-59. The MDL court in the District of Nevada ob- served correctly that “a conspiracy can exist even though not all members of the conspiracy participate in it in the exact same way at the exact time and that some members of the con- spiracy may in fact be at cross purposes at various times based on their individual interests.” In re Western States, 2017 WL 1243135, at *2. Notwithstanding occasionally different inter- ests, the plaintiffs contend, the defendants made many mil- lions of dollars from their speculative trading arms. Further, the plaintiffs have argued, ostensibly divergent price activity would increase the price of fixed-price contracts. No. 22-3279 21

Dr. Dwyer explained: “The more variable daily prices are, the more value there is in having a fixed price.…. To the extent that market manipulations increased the variability of gas prices, they would have increased” the price of fixed-price contracts. Dkt. 208-13 at 20, ¶ 59; 21–22, ¶ 67. 2. Uniformity and Perfection Not Required The plaintiffs’ theory is that the defendants’ conspiratorial wash trades, churning, and false reporting caused the plaintiffs to pay higher prices for natural gas in Wisconsin because natural gas is a fungible commodity in a nationwide market. In such a market, the plaintiffs explained, “prices move up and down together across locations and purchasing methods.” The defendants offered expert evidence showing, they contend, that prices did not consistently move up and down together; rather, depending on the location and price type, prices of natural gas can go in opposite directions. The defendants argue that “the existence of injury varied from transaction to transaction within individual months and among price types.” The defendants argue that the plaintiffs cannot win certification unless they can show “a single market that uniformly impacted all class members” or else “unique circumstances” of each plaintiff’s purchase would defeat predominance. Plaintiffs’ experts have offered rebuttals to the examples cited by the defendants. More fundamental, however, defendants overstate what is needed to show that antitrust impact can be proven by evidence common to all class members. The defendants describe the various ways in which the plaintiffs purchased gas. Even among fixed-price contracts and index-price contracts, the plaintiffs’ purchasing methods retain distinctions, including “trigger” contracts that would 22 No. 22-3279

purchase gas if prices fell to a certain level “managed supply” contracts, which would depend on managing supply for a pool of customers, and various individualized hybrids. This variety of arrangements means at least that price changes would affect different class members at slightly different times. To the extent that the plaintiffs purchased gas through intermediaries, the defendants argue that the expert models fail to reflect that the plaintiffs were paying for not only gas but also other services. Consequently, the defendants argue, these individual differences cause individual questions to overwhelm common questions regarding causation of anti- trust injury and the measurement of damages. To the extent defendants are arguing this evidence defeats the existence of a national market, we leave that question to the plaintiffs’ experts’ rebuttals and ultimately to the district court. But defendants also seem to argue that plaintiffs are re- quired to prove that defendants’ antitrust violations affected all class members to exactly the same degree at exactly the same time. We rejected a similar challenge to class certifica- tion in Messner v. Northshore University Health System. Plain- tiffs there argued that defendants’ merger of hospital net- works had raised prices for health care in hospitals. 669 F.3d at 808. That case involved more extensive pricing problems for complex packages or “baskets” of many different hospital services, and there were similar complications from long-term contracts, which could delay implementation of anticompeti- tive pricing. Id. at 816. The district court had denied class cer- tification on the ground that plaintiffs’ expert had not shown uniform price increases for all class members. Id. at 817–18. We reversed the denial of class certification, concluding that the district court had demanded too much in terms of No. 22-3279 23

uniformity. Id. at 818. The key was that the plaintiffs’ economic expert had offered a common method for showing impact on prices, even if it would produce different results for different individual class members in calculating damages. Id. at 818–19; see also Bogosian v. Gulf Oil Corp., 561 F.2d 434, 455 (3d Cir. 1977) (“If the price structure in industry is such that nationwide the conspiratorially affected prices at the wholesale level … though different in different regions, [were] higher … than the range which would have existed … under competitive conditions, it would be clear that all members of the class suffered some damage, notwithstanding that there would be variations among all dealers as to the extent of their damage.”), abrogation on other grounds recognized in In re Insurance Brokerage Antitrust Litig., 618 F.3d 300, 325 n.25 (3d Cir. 2010); In re Urethane Antitrust Litig., 768 F.3d 1245, 1255 (10th Cir. 2014) (affirming jury verdict for plaintiff class; “district judge could reasonably weigh the evidence and conclude that price-fixing would have affected the entire market, raising the baseline prices for all buyers. Based on the reasonableness of this finding, the judge had the discretion to treat impact as a common question that was capable of class-wide proof”). As we noted in Messner, real-world markets “are not as simple and elegant as the classic economic model….” 669 F.3d at 816. The vast amounts of natural gas pricing data reflect that complexity, and sound economic findings are not neces- sarily elegant. That points us to another important guideline on remand, well established in our antitrust law: perfect proof is not required. It sometimes can be tempting for courts to look so hard for perfect evidence that the search becomes the enemy of quite good evidence. That can be a mistake. 24 No. 22-3279

If the plaintiffs can show that the common method accounts for differences in methods of purchasing and pricing natural gas cited by defendants, it will be permissible to infer impact and to “make a reasonable estimation of actual damages through probability and inference.” Loeb Industries, Inc. v. Sumitomo Corp., 306 F.3d 469, 490 (7th Cir. 2002). “In essence, it is important not to let a quest for perfect evidence become the enemy of good evidence.” Messner, 669 F.3d at 808. The Supreme Court made the same point in Comcast: “Cal- culations need not be exact.” 569 U.S. at 35. In support, the Court cited Story Parchment Co. v. Paterson Parchment Paper Co., which also supports the principle: Where the tort itself is of such a nature as to preclude the ascertainment of the amount of damages with certainty, it would be a perversion of fundamental principles of justice to deny all relief to the injured person, and thereby relieve the wrongdoer from making any amend for his acts. In such case, while the damages may not be determined by mere speculation or guess, it will be enough if the evidence show the extent of the damages as a matter of just and reasonable inference, although the result be only approximate. The wrongdoer is not entitled to complain that they cannot be measured with the exactness and precision that would be possible if the case, which he alone is responsible for making, were otherwise. 282 U.S. 555, 563 (1931). No. 22-3279 25

3. Not Every Class Member and Not Every Transaction Two closely related points are that class certification does not require proof that every class member was injured or that every gas purchase was affected by the conspiracy. To the extent the defendants suggest that before class certification, the plaintiffs must show all class members suffered some injury, that is not correct. We have said this repeatedly. E.g., Kleen Products LLC v. International Paper Co., 831 F.3d 919, 927 (7th Cir. 2016); Suchanek v. Sturm Foods, Inc., 764 F.3d 750, 757 (7th Cir. 2014) (“If the court thought that no class can be certified until proof exists that every member has been harmed, it was wrong.”); Parko, 739 F.3d at 1084–85; McReynolds v. Merrill Lynch, Pierce, Fenner & Smith, Inc., 672 F.3d 482, 490–91 (7th Cir. 2012). We explained in Kohen v. Pacific Investment Management Co. LLC, 571 F.3d 672, 677 (7th Cir. 2009), that “a class should not be certified if it is apparent that a great many persons who have suffered no injury at the hands of the defendant….” We have also explained that, “[t]here is no precise measure for ‘a great many.’ Such determinations are a matter of degree, and will turn on the facts as they appear from case to case.” Messner, 669 F.3d at 825; see also Olean Wholesale Grocery Cooperative, Inc. v. Bumble Bee Foods LLC, 31 F.4th 651, 669 (9th Cir. 2022) (rejecting rule precluding class certification when class might include more than “de minimis” number of uninjured members). Also, if the class definition truly were too broad or heterogenous, the class definition could be adjusted or subdivided. Kohen, 571 F.3d at 680; see also Fed. R. Civ. P. 23(c)(5). Plaintiffs also do not need to prove that every gas purchase by a class member during the class period was affected by the conspiracy. A class member has a claim if any purchase 26 No. 22-3279

during the class period was affected by the conspiracy. For example, some named plaintiffs entered into fixed-price con- tracts before the class period began that lasted well into the class period. The prices they paid early in the class period un- der those contracts should not have been affected by any con- spiratorial price manipulation. But those plaintiffs all appear to have entered into later contracts that took effect during the class period, when, according to plaintiffs’ experts, the prices they paid had been raised by the conspiracy’s manipulations. 4. Actions of Non-defendant Co-conspirators Defendants also criticize plaintiffs’ experts for their reliance upon data showing sales and sales reports by non- defendants. It’s not clear how this criticism affects class certification, at least if plaintiffs can prove the existence of a national market. The extent of the conspiracy will be a matter for proof at trial, by evidence common to all class members. The actions of non-defendant co-conspirators will be fair game at trial. “[E]ach member of a conspiracy is liable for all damages caused by the conspiracy’s entire output.” Paper Systems Inc. v. Nippon Paper Industries Co., 281 F.3d 629, 632 (7th Cir. 2002), citing Texas Industries, Inc. v. Radcliff Materials, Inc., 451 U.S. 630 (1981). Not all co-conspirators need to be named defendants. See id. And a “co-conspirator who joins a conspiracy with knowledge of what has gone on before and with an intent to pursue the same objectives may, in the antitrust context, be charged with the preceding acts of its co- conspirators.” Kleen, 831 F.3d at 930 (internal quotation marks omitted), quoting Havoco of America, Ltd. v. Shell Oil Co., 626 F.2d 549, 554 (7th Cir. 1980). The defendants’ argument about non-defendants’ false reporting thus reduces to the merits of the conspiracy allegation itself. The existence of the No. 22-3279 27

conspiracy and whether non-defendant companies’ transactions were part of it are, as earlier described, common to the class. If the plaintiffs cannot show the non-defendant transactions were part of the conspiracy of the defendants, then that goes to the magnitude of impact and amount of damages, not predominance. 5. Full-Consideration Claim As noted above, Wisconsin law provides an unusual anti- trust remedy: “full consideration.” See Wis. Stat. § 133.14. The statute makes void all contracts that are a product of antitrust conspiracy: “All contracts or agreements made by any person while a member of any conspiracy prohibited by § 133.03, and which contract or agreement is founded upon, is the result of, or grows out of or is connected with any violation of such sec- tion, either directly or indirectly, shall be void ….” Id. Parties who have made payments under such contracts are entitled to full refunds: “Any payment made upon, under or pursuant to such contract or agreement to or for the benefit of any per- son may be recovered from any person who received or ben- efited from such payment .…” Id. On remand the district court may find it helpful to con- sider this remedy separately from the treble-damages remedy more familiar from federal antitrust law. For purposes of class certification, calculating damages should be much simpler for the “full consideration” remedy. The statutory language calls for a causal connection between the contract and the conspir- acy, but it is phrased quite broadly: “is founded upon, is the result of, or grows out of or is connected with any violation ….” Id. 28 No. 22-3279

V. Class Certification Available on This Record One last issue: We have explained why we must vacate the certification of the class and remand for further consideration and fact-finding on the disputed issues affecting antitrust im- pact and thus predominance of common issues. In their briefs on appeal, however, defendants reach further. They argue that the district court could not properly certify a plaintiff class on this record. Defendants would have us order the district court simply to deny class certification. We reject that pro- posal. Defendants seek support from Comcast Corp. v. Behrend, 569 U.S. 27 (2013), but a close look at that case shows critical differences. Most important, Comcast did not involve price- fixing, which makes the question of antitrust impact much simpler here than in Comcast. At its core, Comcast was a mo- nopolization case involving business tactics for which anti- trust impact was less well-recognized than it is for horizontal price-fixing alleged here. Plaintiffs in Comcast challenged a strategy called “clustering,” through which Comcast alleg- edly made “swap” deals with other cable television providers to let each build up large market shares in particular metro- politan or regional markets. Id. at 29–30. The plaintiff consum- ers offered four theories of antitrust impact from these “swaps” and concentrations of market power. Id. at 31. The district court had accepted only one of those theories and re- jected the other three. The problem for class certification was that the plaintiffs’ theory for a common method of proving damages assumed the validity of all four theories of antitrust impact. As a result, the plaintiffs’ damages theory was no longer consistent with their liability theory. Id. at 35. No. 22-3279 29

This case is fundamentally different. Horizontal price- fixing is the most basic, core antitrust violation. Bork, The Antitrust Paradox, supra, at 67. Its empirical effects—prices higher than would prevail in a competitive market—are the archetype, the paradigm, the Platonic form of antitrust impact. An empirical model that reliably estimates the effects of the price-fixing satisfies the needs of class certification. Defendants point out that plaintiffs here allege the con- spirators used different methods for fixing prices: “churning,” wash sales, and fraudulent reports. Defendants then try to equate those different methods with the Comcast plaintiffs’ different theories of antitrust impact. The argument tries to compare apples to oranges. The conspirators’ different mech- anisms to manipulate natural gas prices are all illegal. There is no need for plaintiffs and their experts here to try to distin- guish among those methods and any supposedly different ef- fects, or to distinguish among the conspirators depending on who used which methods, when, and where. Comcast calls for the district court to engage with the merits of the experts’ de- bate over causation and antitrust impact, but it does not show that certification would be inappropriate here. On the core issues—the nuts and bolts of the disputes over what the economic data show about the effects of price ma- nipulation in the natural gas market—plaintiffs’ experts have offered detailed rebuttals to the defense experts’ selections of data and arguments. We would not be surprised if a class can be properly certified here after sufficient “rigorous analysis” of the experts’ debates. After all, the case presents a conspir- acy to manipulate prices for a fungible commodity traded across a nationwide network of pipelines. 30 No. 22-3279

* * * We know that we are demanding a great deal from the district court on remand. The expert testimony relevant to the Wisconsin antitrust impact issue is laid out in around a dozen reports, not to mention the tables, charts, and deposition testimony. The parties’ experts have been debating each other in this litigation since at least 2009, and the key reports were all filed in class certification debates in Nevada and then refiled several years later in Wisconsin. Complicating this case still further, the four experts debating the Wisconsin antitrust impact issue were parts of larger teams of experts addressing a host of other issues. The relevant reports also contain a good deal of information on other issues and other states. On remand here, it will be necessary for the district court to dig into these debates among the experts. Given the number of reports and issues, the district court will be entitled to expect from counsel a guided tour of the expert evidence, perhaps with a guide comparable to detailed flow-charts familiar to competitive debaters. It might make sense for counsel for both parties to collaborate on making such a chart, and the court can determine what would be most helpful. The order certifying the plaintiff class is VACATED and the case is REMANDED to the district court for further consideration of class certification consistent with this opinion.