dissenting from the denial of rehearing in banc:
I respectfully dissent from the denial of rehearing in banc.
In 1968, it became law in this Court that, for public policy reasons, federal antitrust claims could not be arbitrated. See Am. Safety Equip. Corp. v. J.P. Maguire & Co., 391 F.2d 821, 827-28 (2d Cir.1968). The Supreme Court rejected that public policy approach in Mitsubishi Motors Corp. v. Soler Chrysler-Plymouth, Inc., 473 U.S. 614, 636, 105 S.Ct. 3346, 87 L.Ed.2d 444 (1985). And in 1991, it reiterated that federal statutory claims can be subject to valid arbitration agreements. See Gilmer v. Interstate/Johnson Lane Corp., 500 U.S. 20, 25, 111 S.Ct. 1647, 114 L.Ed.2d 26 (1991).
Now the panel opinion in this case uses public policy to hold that arbitration agreements containing class-action waivers are unenforceable when applied to federal statutory claims if (as is always so easy to assert) a claim would not be “economically rational” to pursue individually. In re Am. Express Merchs. ’ Litig., 667 F.3d 204, 214 (2d Cir.2012) (Amex III). The panel opinion thus impairs the Federal Arbitration Act’s strong federal policy favoring the enforcement of arbitration agreements, and frustrates the goals of arbitration by multiplying claims, lawsuits, and attorneys’ fees. “[T]he longstanding judicial hostility to arbitration agreements,” Gilmer, 500 U.S. at 24, 111 S.Ct. 1647, is undiminished.
At issue is a provision, of a kind commonly used in arbitration agreements, that bars class actions. The underlying arbitration involves an antitrust claim. In In *143re American Express Merchants’ Litigation, 554 F.3d 300 (2d Cir.2009) (Amex I), the panel held that such a bar ran afoul of the federal substantive law of arbitration because the litigation expense of the antitrust suit — expert testimony, in particular — would render separate arbitrations too expensive. So the panel ruled that a class action may proceed in court notwithstanding the agreement to arbitrate. Id. at 320. The Supreme Court granted certiorari and vacated Amex I in light of Stolt-Nielsen S.A. v. AnimalFeeds Int’l Corp., — U.S. -, 130 S.Ct. 1758, 176 L.Ed.2d 605 (2010). Am. Express Co. v. Italian Colors Rest., — U.S. -, 130 S.Ct. 2401, 176 L.Ed.2d 920 (2010).
Stolt-Nielsen holds that a party to an arbitration agreement cannot be compelled to submit to class arbitration absent a “contractual basis for concluding that the party agreed to do so ... because class-action arbitration changes the nature of arbitration to such a degree that it cannot be presumed the parties consented to it by simply agreeing to submit their disputes to an arbitrator.” 130 S.Ct. at 1775. On remand the (by then) two-judge panel reached the same conclusion as in Amex I. See In re Am. Express Merchs. ’ Litig., 634 F.3d 187, 199 (2d Cir.2011) (Amex II).
Shortly after Amex II was published but before the mandate issued, the Supreme Court decided AT&T Mobility LLC v. Concepcion, — U.S.-, 131 S.Ct. 1740, 179 L.Ed.2d 742 (2011), which holds that state law may not be used to invalidate a class-action waiver in an arbitration agreement on the ground that the only economical way to litigate the claim is through a class action. Id. at 1748. After soliciting briefing on the impact of Concepcion, the panel issued its third opinion. In Amex III, the panel yet again concludes that the class-action waiver is unenforceable on the ground that the only effective way to litigate the antitrust claims was by a class action in court. Amex III, 667 F.3d at 218-19.
As I undertake to show, the public policy rationale which Amex III relies upon is wrong because: [1] it runs counter to the public policy that the Supreme Court has made paramount in the context of the Federal Arbitration Act (“FAA”); [2] it employs a dubious ground of distinction to overcome Concepcion, which teaches that the FAA does not allow courts to invalidate class-action waivers even if “class proceedings are necessary to prosecute small-dollar claims that might otherwise slip through the legal system,” Concepcion, 131 S.Ct. at 1753; and [3] the dicta on which the panel precariously relies — that large “arbitration costs” cannot be allowed to prevent a plaintiff from “effectively vindicating” a statutory right — is pulled out of context and distorted.
I
Amex III cannot be squared with the FAA, as it has been applied and explained by the Supreme Court. In banc review is needed because [A] the panel opinion is unbounded and can be employed to defeat class-action waivers altogether; [B] it makes the district court the initial theater of arbitral conflict on the merits (how else does a district court estimate the cost of a litigation?); and [C] it is already working mischief in the district courts.
A
Amex III is a broad ruling that, in the hands of class action lawyers, can be used to challenge virtually every consumer arbitration agreement that contains a class-action waiver — and other arbitration agreements with such a clause. While it purports to require a case-by-case approach, its wording is categorical: “Supreme Court precedent recognizes that the *144class action device is the only economically rational alternative when a large group of individuals or entities has suffered an alleged wrong, but the damages due to any single individual or entity are too small to justify bringing an individual action.” Amex III, 667 F.3d at 214. Thus every class counsel and every class representative who suffers small damages can avoid arbitration by hiring a consultant (of which there is no shortage) to opine that expert costs would outweigh a plaintiffs individual loss.
The breadth of the holding is illustrated in the opinion. Amex III uncritically adopts the affidavit of a paid consultant to find that expert costs would be so high relative to potential damages, that “the only economically feasible means for plaintiffs enforcing their statutory rights is via a class action [in court].” 667 F.3d at 218.
However, Section 4 of the Clayton Act provides for the recovery of costs, including expert costs, and attorneys’ fees. See 15 U.S.C. § 15(a) (“[A] ny person who shall be injured in his business or property by reason of anything forbidden in the anti-trust laws ... shall recover threefold the damages by him sustained, and the cost of suit, including a reasonable attorney’s fee.”). The Amex panel is evidently of the view that the incentivizing fees and cost afforded by the statute would not fully compensate plaintiffs for the costs of pursuing their claims. See Amex III, 667 F.3d at 218. But Congress deems these incentives sufficient to encourage private suits. The judgment of Congress in such a matter is entitled to deference, not the panel opinion’s dismissive treatment.
Amex III does not vouchsafe what is meant for a suit to be “economically feasible,” or when a hypothetical “economically rational” plaintiff might be willing to pursue a claim. Id. at 218. It cannot mean that a potential plaintiff must have the opportunity to be made whole and happy by recovery of damages, costs, attorneys’ fees, expert charges, etc., because such a result is rarely achieved by even the most successful litigants. Moreover, Amex III demands more than such complete victory; it demands a “risk-of-losing” premium. Id. at 218 (“Even with respect to reasonable attorney’s fees[,] ... the plaintiffs must include the risk of losing, and thereby not recovering any fees, in their evaluation of their suit’s potential costs.”). This formulation betrays a dominant consideration — that, without the class-action vehicle, no lawyer will be incentivized to pursue these claims. That may be; but Concepcion rejected this very policy rationale. See Concepcion, 131 S.Ct. at 1753 (rejecting argument that “class proceedings are necessary to prosecute small-dollar claims that might otherwise slip through the legal system,” because rules inconsistent with the FAA cannot be imposed “even if desirable for unrelated reasons”); see also Coneff v. AT & T Corp., 673 F.3d 1155, 1159 (9th Cir.2012) (rejecting argument that plaintiffs had insufficient incentive to pursue individual claims as “primarily a policy rationale” that “cannot undermine the FAA”).
B
Under the panel opinion, arbitration must now begin in federal court — and be litigated there on the merits in many critical respects. The courtroom inquiry that the panel requires to be undertaken before any class arbitration can in fact take place is searching. Whether a dispute may require expert testimony is a question inseparable from the merits (and raises Daubert and other vexed questions). Without a close inquiry into the merits, no court can decide what expert testimony would be required, or how much discovery is needed. And it cannot be decided whether any *145discovery or testimony is needed at all without deciding if the claim is dismissible — or such prior questions as the statute of limitations and laches, controlling law, res judicata, etc., etc., not to mention little things like whether the putative class is duly constituted and properly represented, without which there is no class claim.
Under the FAA, however, all those questions are for the arbitrator to decide. See, e.g., Prima Paint Corp. v. Flood & Conklin Mfg. Co., 388 U.S. 395, 403-404, 87 S.Ct. 1801, 18 L.Ed.2d 1270 (1967). By requiring the district court to consider this at the threshold, Amex III effectively displaces arbitration with a trial court proceeding whenever lawyers assert a class claim. (And they will, often.) Even if arbitration is given a green light at the end of the judicial proceeding, the party seeking to arbitrate may have already spent many times the cost of an arbitral proceeding just enforcing the arbitration clause. And the partial list of issues above will create fertile ground for appeal, adding yet more delay, expense, and uncertainty. The predictable upshot is that Amex III will render arbitration too expensive and too slow to serve any of its purposes.
Amex III is incompatible with the FAA. The FAA “establishes that, as a matter of federal law, any doubts concerning the scope of arbitrable issues should be resolved in favor of arbitration.” Moses H. Cone Mem’l Hosp. v. Mercury Constr. Corp., 460 U.S. 1, 24-25 & n. 32, 103 S.Ct. 927, 74 L.Ed.2d 765 (1983). The federal substantive law of arbitration “is a congressional declaration of a liberal federal policy favoring arbitration agreements.” Id. at 24, 103 S.Ct. 927. This is particularly true in light of Concepcion’s reaffir-mance of the “overarching purpose” of the FAA:
The overarching purpose of the FAA, evident in the text of §§ 2, 3, and 4, is to ensure the enforcement of arbitration agreements according to their terms so as to facilitate streamlined proceedings. Requiring the availability of classwide arbitration interferes with fundamental attributes of arbitration and thus creates a scheme inconsistent with the FAA.
Concepcion, 131 S.Ct. 1740.
C
In the six years Amex has been pending in this Court, its several iterations have been relied upon no fewer than three times in the Southern District of New York alone. See Raniere v. Citigroup, Inc., 827 F.Supp.2d 294, 309-10 (S.D.N.Y.2011); Chen-Oster v. Goldman, Sachs & Co., No. 10 Civ. 6950, 2011 WL 2671813, at *2-5 (S.D.N.Y. July 7, 2011); Sutherland v. Ernst & Young, LLP, 768 F.Supp.2d 547, 550-54 (S.D.N.Y.2011).1 Given the recurrent influence of Amex, this Court should subject it to in banc review.
That responsibility is even more compelling because the panel opinion now splits with a recent holding of the Ninth Circuit Court of Appeals. See Coneff v. AT & T Corp., 673 F.3d 1155, 1158 n. 2, 1159 n. 3 (9th Cir.2012). In Coneff, a putative class of AT & T wireless customers sued AT & T on a variety of claims, including a violation of the Federal Communications Act. Id. at 1157. The Ninth Circuit held that Green Tree Financial Corp. v. Randolph, 531 U.S. 79, 121 S.Ct. 513, 148 L.Ed.2d 373 (2000), was no obstacle to the enforcement of the arbitration agreement containing a *146class-action waiver because under the FAA it is irrelevant whether customers “have insufficient incentive” “to vindicate their rights.” Id. at 1159. (citing Concepcion, 131 S.Ct. at 1753).
II
Amex III is thus incompatible with the longstanding principle of federal law, embodied in the FAA and numerous Supreme Court precedents, favoring the validity and enforceability of arbitration agreements. It should come as no surprise, then, that the panel opinion finds no support in the Supreme Court’s case law. Instead, Amex III proceeds by selective quotation from Supreme Court dicta, and by aggressive measures to distinguish away the Supreme Court’s recent holding in Concepcion.
A
Concepcion, decided after the second iteration of Amex, vindicated the FAA against an unconscionability challenge that was materially indistinguishable from the challenge upheld in Amex. In Concepcion, the Supreme Court rejected a common-law rule, developed by the California Supreme Court, that was applied to void class-action waivers in contracts of all types. This is what the discredited California opinion had said:
[Bjeeause ... damages in consumer cases are often small and because a company which wrongfully exacts a dollar from each of millions of customers will reap a handsome profit, the class action is often the only effective way to halt and redress such exploitation.... Such one-sided, exculpatory contracts in a contract of adhesion, at least to the extent they operate to insulate a party from liability that otherwise would be imposed under California law, are generally unconscionable.
Discover Bank v. Superior Court, 36 Cal.4th 148, 161, 30 Cal.Rptr.3d 76, 113 P.3d 1100 (2005) (internal quotation marks, citations, and alterations omitted).
The Supreme Court ruled that this attempt by California to police arbitration agreements was inconsistent with the FAA. Concepcion, 131 S.Ct. at 1748. Refuting the dissent’s argument that “class proceedings are necessary to prosecute small-dollar claims that might otherwise slip through the legal system,” the majority affirmed that rules inconsistent with the FAA cannot be imposed “even if desirable for unrelated reasons.” Id. at 1753.
After the Amex panel solicited briefing from the parties on the effect of Concepcion, the panel reissued Amex (in the form of Amex III), evading the broad language and clear import of Concepcion. Again in Amex III, the panel found that a class-waiver provision in an arbitration agreement is unenforceable if “the only economically feasible means for plaintiffs enforcing their statutory rights is via a class action.” Amex III, 667 F.3d at 218.
Amex III tries to narrow Concepcion to (in the words of Amex III) a “path for analyzing whether a state contract law is preempted by the FAA.” Amex III, 667 F.3d at 213. In so doing, Amex III conceives the following distinction: Concepcion decided only whether California’s doctrine of unconscionability was preserved by the FAA’s savings clause for “grounds as exist at law or in equity for the revocation of any contract,” 9 U.S.C. § 2, whereas Amex III invalidates the arbitration agreement (for the same reason of unconscionability) on the ground that the underlying antitrust claim was federal, a circumstance that the panel dresses up rhetorically as a “federal substantive law of arbitrability,” Amex III, 667 F.3d at 213 (quotation marks omitted). This labored analysis does not rise to a distinc*147tion, and treats the reasoning of Concepcion as an obstacle to be surmounted or evaded. Since, as the Supreme Court has held, the FAA preempts even state law that permits evasion of a class action waiver clause, it is hard for me to see any justification for a rule permitting precisely the same sort of evasion as part of the “federal substantive law of arbitrability.”
B
The panel opinion leans on the distortion of dicta from Green Tree Financial Corp. v. Randolph, 531 U.S. 79, 121 S.Ct. 513, 148 L.Ed.2d 373 (2000). In Green Tree, a lender sought to compel a borrower to arbitrate claims she had raised under certain federal statutes. Id. at 83, 121 S.Ct. 513. The question was “whether [her] agreement to arbitrate is unenforceable because it says nothing about the costs of arbitration, and thus fails to provide her protection from potentially substantial costs of pursuing her federal statutory claims in the arbitral forum.” Id. at 89, 121 S.Ct. 513. The Court reconfirmed “that federal statutory claims can be appropriately resolved through arbitration,” id. at 89, 121 S.Ct. 513, and “rejected generalized attacks on'arbitration that rest on a ‘suspicion of arbitration as a method of weakening the protections afforded in the substantive law to would be complainants,’ ” id. at 89-90, 121 S.Ct. 513 (quoting Rodriguez de Quijas v. Shearson/Ameri-can Express, Inc., 490 U.S. 477, 481, 109 S.Ct. 1917, 104 L.Ed.2d 526 (1989)). And the challenge failed for want of evidence of the “cost” of the arbitration. Id. at 90, 121 S.Ct. 513.
A passage in dicta (relied upon in Amex III) added that “the existence of large arbitration costs could preclude a litigant ... from effectively vindicating her federal statutory rights.” Id. at 90, 121 S.Ct. 513. However, “large arbitration costs” is not a reference to expense generally. Green Tree uses the phrase to reference the cost of access to an arbitral forum and is about the price of admission: “payment of filing fees, arbitrators’ costs, and other arbitration expenses.” Green Tree, 531 U.S. at 84, 121 S.Ct. 513. Only Amex III has suggested that a claim that may be expensive to litigate — whether in court or in arbitration — can for that reason be deemed to entail preclusive “arbitration costs.” In any event, even if the Green Tree dicta were to have the meaning the panel ascribes to it, it is nonetheless still dicta. And it loses any persuasive power it might once have had in light of the Supreme Court’s holding in Concepcion, which is more clear and more recent — and authoritative.
Similarly misleading is the panel’s quotation of Mitsubishi, for the proposition that “should clauses in a contract operate ‘as a prospective waiver of a party’s right to pursue statutory remedies for antitrust violations, we would have little hesitation in condemning the agreement as against public policy.’ ” Amex III, 667 F.3d at 214 (quoting Mitsubishi, 473 U.S. at 637 n. 19, 105 S.Ct. 3346). The Court was there concerned with a hypothetical arbitral panel that might, relying on provisions concerning choice of forum or choice of law, refuse to apply American law to a federal statutory claim. Mitsubishi, 473 U.S. at 637 n. 19,105 S.Ct. 3346.
Other circuit cases have excised provisions from arbitration agreements for the precise reasons anticipated by Green Tree and Mitsubishi. See Kristian v. Comcast Corp., 446 F.3d 25, 47-48 (1st Cir.2006) (severing waiver of treble damages); Hadnot v. Bay, Ltd., 344 F.3d 474, 478 n. 14 (5th Cir.2003) (noting that waiver of exemplary and punitive damages is unenforceable); Paladino v. Avnet Computer Techs., Inc., 134 F.3d 1054, 1060 (11th Cir.1998) *148(holding that arbitration agreement cannot force a party to arbitrate a statutory right and at the same time bar it from being awarded damages in the arbitral forum). All of these three cases involved an arbitration agreement that entirely foreclosed a remedy to which one of the parties was otherwise entitled to seek at law. None of them invalidated an arbitration agreement on the ground that the claims were costly to litigate individually.2
In Amex, there is zero evidence that any “arbitration costs” — within the meaning of Green Tree — would hamper the plaintiffs’ ability to vindicate their statutory rights. None of the three panel opinions references the size of the filing fees, or any arbitrators’ fees that would befall the plaintiffs. In finding that claim-by-claim litigation would not be “economically feasible,” Amex III, 667 F.3d at 204, the panel relies solely on the affidavit of a paid consultant, Gary French, who opined that preparing an antitrust study would cost “at least several hundred thousand dollars, while a larger study can easily exceed $1 million.” Id. at 212.3 His preliminary review of the particular claim yielded a guess of nearly one million dollars. Id. However, that is beside the point: The ability to spread costs among a class is only a procedural right, the absence of which cannot render arbitration costs prohibitive. Gilmer v. Interstate/Johnson Lane Corp., 500 U.S. 20, 25, 111 S.Ct. 1647, 114 L.Ed.2d 26 (1991), is instructive: an employee sought to avoid arbitrating his ADEA claims on the ground that “arbitration is inconsistent with the ADEA.” Id. at 30, 111 S.Ct. 1647. The Supreme Court characterized that argument as “rest[ing] on suspicion of arbitration as a method of weakening the protections afforded in substantive law to would-be complainants, and as such, ... far out of step with our current strong endorsement of the federal statutes favoring this method of resolving disputes.” Id. (internal quotation marks omitted).
Gilmer’s argument about the unavailability of class actions was expressly rejected:
It is also argued that arbitration procedures cannot adequately further the purposes of the ADEA because they do not provide for broad equitable relief and class actions.... But even if the arbitration could not go forward as a class action or class relief could not be granted by the arbitrator, the fact that the [ADEA] provides for the possibility *149of bringing a collective action does not mean that individual attempts at conciliation were intended to be barred.
Id. at 32, 111 S.Ct. 1647 (internal quotation marks omitted). As the passage from Gil-mer reflects, the ADEA expressly provides for a collective action; a fortiori, the same result obtains under the antitrust laws, which do not. The only right to an antitrust class action is “merely a procedural one, arising under Fed.R.Civ.P. 23, that may be waived by agreeing to an arbitration clause.” Johnson v. W. Suburban Bank, 225 F.3d 366, 369 (3d Cir.2000) (enforcing, due to absence of congressional intent to the contrary, a bilateral arbitration clause “even though [such clauses] may render class actions to pursue statutory claims ... unavailable”).
. These three cases also happen to be the only citations in Amex III that support its “vindication of rights” analysis. See Amex III, 667 F.3d at 219. This is of course self-referential: the citation of Second Circuit opinions by the district courts of this Circuit is not a form of endorsement.
. Amex III asserts that "[o]ther Circuits permit plaintiffs to challenge class-action waivers on the grounds that prosecuting such claims on an individual basis would be a cost prohibitive method of enforcing a statutory right,” Amex III, 667 F.3d at 216-17 (citing In re Cotton Yam Antitrust Litig., 505 F.3d 274, 285 (4th Cir.2007)); Livingston v. Assocs. Fin., Inc., 339 F.3d 553, 555, 557 (7th Cir.2003); Adkins v. Labor Ready, Inc., 303 F.3d 496, 502-03 (4th Cir.2002). Each of those opinions quotes the "prohibitive costs” passage of Green Tree, but none uses the phrase as Amex III uses it — and all find in favor of the party seeking to enforce the arbitration clause. For one thing, the plaintiffs in each case failed to provide non-conclusory cost evidence. Notably, in Livingston and Adkins (upon which Cotton Yarn relies) the plaintiffs had raised the specter of prohibitive arbitration fees — not expenses incident to litigation. See Livingston, 339 F.3d at 557 ("Tellingly, [plaintiffs'] only 'evidence' of prohibitive arbitration costs is an unsubstantiated and vague assertion that discovery in an unrelated arbitration matter disclosed fees of nearly $2,000 per day.”); Adkins, 303 F.3d at 503 ("[Plaintiff] does not even provide any evidence about the most basic element of this challenge: the size of the allegedly 'prohibitive' arbitration fee itself.”). These cases were thus concerned about the price of admission.
. It evidently did not occur to French or the panel that the rules of evidence do not govern arbitration, and that an arbitrator can consult treatises and articles for relevant antitrust and economic principles, and should do so in some cases.