Reevaluating Representation in Class Action Law: Self-Interest vs. Members' Welfare, Study notes of Communication

Download Reevaluating Representation in Class Action Law: Self-Interest vs. Members' Welfare and more Study notes Communication in PDF only on Docsity! Rethinking Adequacy of Representation Jay Tidmarsh* Class representatives and class counsel must adequately represent the members of a class. This principle forms the foundation for the modern American class action, and it determines the structure of Rule 23 of the Federal Rules of Civil Procedure and every analogous state class-action rule.1 The absence of adequate representation dooms the certification of a class. The gnawing fear that class representation is inadequate—manifested through such phrases as “collusion,”2 “conflicts of interest,”3 “selling out the class,”4 and “sweetheart deals”5—is an enduring criticism of class actions. Indeed, the demand of adequate representation is so irresistible that in recent years the principle has spread beyond class actions to other forms of aggre- gate litigation.6 Despite the allure of the principle, we have very little sense of what adequate representation means, how we can measure it, or how we can * Professor of Law, Notre Dame Law School. I thank Roger Trangsrud and participants in faculty forums at Loyola University Chicago School of Law, DePaul University College of Law, George Washington University Law School, and the University of Cincinnati College of Law for discussions and comments on early versions of this Article, and Adam Kern for research assistance. 1. FED. R. CIV. P. 23. Unless otherwise stated, all references to Rules are to the Federal Rules of Civil Procedure. 2. See Hansberry v. Lee, 311 U.S. 32, 45 (1940) (rejecting the representation as inadequate in part because it afforded “opportunities . . . for the fraudulent and collusive sacrifice of the rights of absent parties”); Susan P. Koniak, Feasting While the Widow Weeps: Georgine v. Amchem Products, Inc., 80 CORNELL L. REV. 1045, 1048 (1995) (describing “collusion between class counsel and the defendants” in a mass-tort class-action settlement). 3. See Amchem Prods., Inc. v. Windsor, 521 U.S. 591, 625 (1997) (stating that the inquiry into adequacy of representation “serves to uncover conflicts of interest between named parties and the class they seek to represent”). 4. See Culver v. City of Milwaukee, 277 F.3d 908, 910 (7th Cir. 2002) (recognizing “the danger that the lawyer will sell out the class in exchange for the defendant’s tacit agreement not to challenge the lawyer’s fee request”). 5. See Bruce Hay & David Rosenberg, “Sweetheart” and “Blackmail” Settlements in Class Actions: Reality and Remedy, 75 NOTRE DAME L. REV. 1377, 1377 (2000) (noting that one of the “dangers associated with the settlement of class actions” is “the problem of ‘sweetheart’ settlements, in which the class members’ interests are compromised by class counsel”); Charles Silver, Class Actions—Representative Proceedings, in 5 ENCYCLOPEDIA OF LAW AND ECONOMICS 194, 213 (Boudewijn Bouckaert & Gerrit De Geest eds., 2000) (discussing four types of “sweetheart deals”). 6. See PRINCIPLES OF THE LAW OF AGGREGATE LITIGATION § 1.05 (Tentative Draft No. 1, 2008) [hereinafter AGGREGATE LITIGATION] (suggesting guidelines for dealing with all forms of aggregate litigation to ensure adequate representation); Howard M. Erichson, Informal Aggregation: Procedural and Ethical Implications of Coordination Among Counsel in Related Lawsuits, 50 DUKE L.J. 381, 468 (2000) (arguing that attorneys should, in the absence of formal class certification, use written cooperation agreements to replicate the same protections available to class- action plaintiffs). 1138 Texas Law Review [Vol. 87:1137 guarantee it.7 We use a few heuristics—for instance, “conflicts of interest and collusion are bad”—but the complexities of class-action practice quickly blur even these seemingly bright-line measuring rods.8 If you look for them, conflicts and the possibility of collusion lurk in almost every class action.9 The question therefore becomes how great must the conflict of interest be or how strong must the evidence of collusion be in order to deem a class’s rep- resentation inadequate.10 Some of the difficulty in determining the meaning and measure of adequate representation derives from the ambiguities in Hansberry v. Lee,11 the seminal case that rooted adequate representation, at least in part, in the Constitution.12 A greater part of the difficulty is the ambivalent relationship between the requirement of adequate representation and the underlying goal of class actions. The class action sweeps together parties with similarly situ- ated claims,13 thus avoiding one or more of a number of evils—to the litigants, to the absent class members, or to the court system—that repetitive individual litigation creates.14 But these economic and social benefits can often be reaped even when the representation of the class members is less than adequate. To take a crass example, a class action might achieve more deterrent effect at no greater litigation expense if the class representative and class counsel were allowed to split ninety-nine percent of the recovery due to class members;15 but such an arrangement would surely doom class 7. See Richard A. Nagareda, Administering Adequacy in Class Representation, 82 TEXAS L. REV. 287, 288 (2003) (“For all the agreement on the centrality of adequate representation to the modern class action . . . there remains remarkably little agreement on the content of that concept or how to enforce it.”). 8. See, e.g., Reynolds v. Beneficial Nat’l Bank, 288 F.3d 277, 282 (7th Cir. 2002) (“[I]n light of the modesty of the stakes . . . we are not disposed to regard this [conflict of interest] as fatal.”); In re “Agent Orange” Prod. Liab. Litig., 800 F.2d 14, 19 (2d Cir. 1986) (noting that principles of efficiency and fairness uniquely implicated in class actions may compel a court to disregard technical conflicts of interest in order to allow an attorney to continue representation of the class); JACK B. WEINSTEIN, INDIVIDUAL JUSTICE IN MASS TORT LITIGATION 46 (1995) (“The common practice in mass tort cases . . . unfortunately has been to ignore ethical problems in the name of expediency.”). 9. For a discussion of the inevitability of conflicts of interest, see infra Part II. 10. See Charles Silver & Lynn Baker, I Cut, You Choose: The Role of Plaintiffs’ Counsel In Allocating Settlement Proceeds, 84 VA. L. REV. 1465, 1468–69 (1998) (“There being no way to eliminate conflicts [of interest] from multiple-claimant representations, the only question is how to deal with them.”). 11. 311 U.S. 32 (1940). 12. For a discussion of Hansberry and its ambiguities, see infra notes 66–91 and accompanying text. 13. Class actions can be brought on behalf of a class of claimants or against a class of persons. The former are usually, though misleadingly, referred to as “plaintiff class actions,” and the latter as “defendant class actions.” See Taylor v. Sturgell, 128 S. Ct. 2161, 2172 (2008) (noting that members of a class are in fact “nonparties” bound by the judgment). In this Article, I will refer only to the far more common plaintiff class action, although my analysis applies equally to both types. 14. For an expanded description of these evils, see infra subpart I(B). 15. This arrangement could be economically justified on the ground that, from the viewpoint of deterrence, the recipient of the recovery is irrelevant and the award of the bulk of recovery to the 2009] Rethinking Adequacy of Representation 1141 two principles mean that no judgment can bind a plaintiff unless she consents to join the case as a party. Effectively, therefore, the default principle is the permissive, or voluntary, joinder of plaintiffs: Each plaintiff must agree to join the case. At the federal level, Rule 20(a) embodies this principle, and permits the joinder of plaintiffs when their claims for relief “aris[e] out of the same transaction, occurrence, or series of transactions or occurrences”20 and “any question of law or fact common to all plaintiffs will arise in the action.”21 Although am- biguously expressed in the word “may,”22 the idea underlying Rule 20(a) is to give each plaintiff the choice to join a case.23 Voluntary joinder is highly consonant with the adversarial approach to litigation employed in American courts. The adversarial system places con- trol over critical litigation decisions regarding the shaping of proofs and arguments in the hands of each party.24 Because few decisions so affect the proofs and arguments as the joinder of those with whom and against whom the case will proceed,25 placing the joinder decision in the hands of each and in three other limited areas, see Taylor v. Sturgell, 128 S. Ct. 2161, 2172–73 (2008). Taylor identified these three circumstances as a nonparty’s agreement to be bound by a judgment, a nonparty’s assumption of control over prior litigation, and litigation brought by a nonparty proxy on behalf of a party to prior litigation. Id. 20. FED. R. CIV. P. 20(a)(1)(A). 21. FED. R. CIV. P. 20(a)(1)(B). Because it is difficult to imagine any claims involving the same transaction or occurrence that would not also involve a common question of law or fact, the independent significance of Rule 20(a)(1)(B) is uncertain. See JAY TIDMARSH & ROGER H. TRANGSRUD, COMPLEX LITIGATION AND THE ADVERSARY SYSTEM 108 (1998) (questioning whether “the common ‘question of law or fact’ requirement in Rule 20 [is] superfluous”). 22. See FED. R. CIV. P. 20(a)(1) (“Persons may join in one action as plaintiffs . . . .”). 23. Indeed, the idea of individual consent is so embedded in Rule 20(a) that, despite the arguable ambiguity in the rule, it is difficult to find cases that discuss whether a plaintiff can use Rule 20(a) to join another plaintiff involuntarily. In Cle-Ware Rayco, Inc. v. Perlstein, 401 F. Supp. 1231 (S.D.N.Y. 1975), the court granted a motion to make a third-party defendant an additional plaintiff, but it was not clear from the opinion whether the third-party defendant consented to joinder as a plaintiff or was joined involuntarily. See id. at 1232–33 (joining a corporation as a party plaintiff after a merger between that corporation and the named plaintiff). In Lyne v. Arthur Anderson & Co., No. 91 C 1885, 1991 WL 247576 (N.D. Ill. Nov. 12, 1991), a court refused to allow ten plaintiffs who had joined together under Rule 20(a) to use the rule to join another seven plaintiffs involuntarily, noting that “[t]here is nothing fundamentally fair about joining plaintiffs who have expressed no interest in participating in this litigation.” Id. at *3. 24. See Lon L. Fuller, The Forms and Limits of Adjudication, 92 HARV. L. REV. 353, 383–84 (1978) (“[T]he true significance of partisan advocacy lies deeper, touching . . . the integrity of the adjudicative process itself. . . . Each advocate comes to the hearing prepared to present his proofs and arguments . . . .”). 25. Cf. Hilao v. Estate of Marcos, 103 F.3d 767, 782–87 (9th Cir. 1996) (permitting proof of damages through statistical sampling and a master’s testimony, rather than individualized proof of damages, in a class action containing 10,000 members). Experimental data suggest that the aggregation of claims increases the likelihood of recovery on weak claims, but suppresses the value of strong claims. Irwin A. Horowitz & Kenneth S. Bordens, The Effects of Outlier Presence, Plaintiff Population Size, and Aggregation of Plaintiffs on Simulated Civil Jury Decisions, 12 LAW & HUM. BEHAV. 209, 226 (1988) [hereinafter Horowitz & Bordens, Aggregation]; see Irwin A. Horowitz & Kenneth S. Bordens, The Consolidation of Plaintiffs: The Effects of Number of Plaintiffs on Jurors’ Liability Decisions, Damage Awards, and Cognitive Processing of Evidence, 1142 Texas Law Review [Vol. 87:1137 plaintiff makes perfect sense from the viewpoint of adversarial theory. At a deeper level, both the adversarial system in general and voluntary joinder in particular respond to the same philosophical intuition about individual autonomy: Allowing individuals the freedom to act on and to govern their own legal affairs is a political and moral good.26 The American version of adversarialism embodies another principle: In making basic decisions regarding the structure and prosecution of a case, a plaintiff is entitled to be guided by self-interest. The relationship between autonomy and self-interest is evident, but hardly inevitable; for instance, a Kantian conception of autonomy leads to the recognition of duties that can- not be justified by, and often are inconsistent with, an individual’s self- interest.27 Whatever the philosophical justification, the American litigation system does not expect litigants to be concerned with the effects of their liti- gation decisions on the positions or rights of others. The freedom to act given to the plaintiff in the adversarial system includes the freedom to act in a decidedly self-interested fashion. Thus, a plaintiff is within her rights to join with others when it advances her interests, and equally within her rights to refuse to join with others when doing so does not advance her interests— even when other plaintiffs’ claims, the defendant’s position, or society’s in- terests are made worse by her decision.28 85 J. APPLIED PSYCHOL. 909, 914 (2000) [hereinafter Horowitz & Bordens, Consolidation] (reporting data showing that the likelihood of recovery increases with the inclusion of more plaintiffs, but the average damage award decreases if more than four plaintiffs are aggregated). 26. See STEPHAN LANDSMAN, READINGS ON ADVERSARIAL JUSTICE: THE AMERICAN APPROACH TO ADJUDICATION 33–39 (1988) (defending the adversarial process as important to maintaining a free society); Monroe H. Freedman, How Lawyers Act in the Interests of Justice, 70 FORDHAM L. REV. 1717, 1727 (2002) (“[W]orking within the rule of law in our constitutionalized adversary system, we enhance our clients’ autonomy as free citizens in a free society.”); Fuller, supra note 24, at 372 (discussing the relationship between adjudication and the rule of law). 27. See Immanuel Kant, On the Common Saying: “This May Be True in Theory, But It Does Not Apply in Practice,” reprinted in KANT: POLITICAL WRITINGS 61, 70 (Hans Reiss ed., H.B. Nisbet trans., 2d ed. 1991) (“[T]he concept of duty . . . is far more powerful, incisive and likely to promote success than all incentives borrowed from the . . . selfish principle [of happiness].”); David Luban, Lawyers as Upholders of Human Dignity (When They Aren’t Busy Assaulting It), 2005 U. ILL. L. REV. 815, 826 (“Kantian autonomy represents freedom achieved through stoic self-control and self-command; it means reasoned self-restraint. Freedom of choice represents casting off restraints.”). 28. Due to the costliness of litigation, a plaintiff’s lawsuit almost invariably worsens the defendant’s position. I do not suggest that it is unjust for the plaintiff to invoke the machinery of the legal system when she has a nonfrivolous claim. What I mean when I describe a plaintiff’s “worsening” of the positions of other plaintiffs, defendants, or society is the plaintiff’s imposition of costs above and beyond the direct and ordinary costs of litigating and satisfying a judgment or settlement. To take a paradigmatic situation of worsening a defendant’s position, consider an interpleader-type scenario in which a defendant holds 1,000 shares of stock that multiple plaintiffs claim. If one plaintiff successfully brings an action and recovers the stock, the defendant might be subject to later suits by other plaintiffs, who, as nonparties to the first case, are not bound by the finding in the first case that the stock belongs to the first plaintiff. See supra note 19 and accompanying text. The defendant might thus be subject to numerous judgments when only one is appropriate. To take a paradigmatic situation of worsening other potential plaintiffs’ positions, if we assume that the sum total of claims exceeds a defendant’s assets, an early plaintiff that receives 2009] Rethinking Adequacy of Representation 1143 Viewed from the vantage point of moral philosophy, this license to act in a purely self-interested way, and to ignore the harms that actions cause others, is intriguing. Somewhat simplistically, ethical theories can be divided into the nonconsequentialist (i.e., evaluating the moral significance of actions without regard to their consequences) and the consequentialist (i.e., evaluat- ing the moral significance of actions by evaluating their consequences).29 Nonconsequentialist ethical theories subdivide into two branches: virtue ethics, which emphasizes the development of moral character in the actor,30 and deontological ethics, which emphasizes the moral quality of the action and often leads to the specification of duties, rules, and obligations.31 Conse- quentialist theories subdivide into three branches. The first is egoism, which involves either a claim that people inevitably act only in their own self- interest (psychological egoism), or a claim that they should act only in their own self-interest (ethical egoism).32 The opposite theory is altruism, which full satisfaction can cause later plaintiffs to receive less than a proportional share of the defendant’s assets. The paradigmatic example of increasing social costs results from a refusal to join with similarly situated plaintiffs, thus generating unnecessary expenses associated with the repetitive relitigation of common issues. Another potential cost in entrusting joinder decisions to the plaintiff is a shift, in the plaintiff’s favor, of the probability of recovery when multiple cases are joined. See supra note 25. If we assume that a decision rendered in individual litigation is the most accurate (an admittedly controversial assumption), a shift in the outcome due to joinder makes a judgment less accurate. Loss in accuracy is a cost to the defendant, as well as a social cost. See RICHARD A. POSNER, ECONOMIC ANALYSIS OF LAW § 21.1–21.2, at 593–95 (7th ed. 2007) (arguing that, from the perspective of an individual defendant or society, when the cost of judicial error outweighs the per- case cost of reducing the chance of such error, the result of judicial error is a net social loss, which should not be tolerated). The extent of this cost is uncertain, for the loss in accuracy due to the aggregation of claims might be offset by lower per-plaintiff awards in aggregated proceedings. See Horowitz & Bordens, Consolidation, supra note 25, at 914. 29. Walter Sinnott-Armstrong, Consequentialism, in THE STANFORD ENCYCLOPEDIA OF PHILOSOPHY (Edward N. Zalta ed., 2008) [hereinafter ENCYCLOPEDIA OF PHILOSOPHY], http:// plato.stanford.edu/archives/fall2008/entries/consequentialism (“Consequentialism, as its name suggests, is the view that normative properties depend only on consequences.”). 30. Rosalind Hursthouse, Virtue Ethics, in ENCYCLOPEDIA OF PHILOSOPHY, supra note 29, http://plato.stanford.edu/archives/fall2003/entries/ethics-virtue. Virtue theories are sometimes referred to as “aretaic theories” after the Greek word arete, or “virtue.” See Lawrence B. Solum, Public Legal Reasoning, 94 VA. L. REV. 1449, 1463 (2006) (referring to virtue theories as aretaic theories and describing the Greek origin of the term). 31. Larry Alexander & Michael Moore, Deontological Ethics, in ENCYCLOPEDIA OF PHILOSOPHY, supra note 29, http://plato.stanford.edu/archives/win2007/entries/ethics-deonto logical. Some deontological theories exclude consideration of the consequences of actions, but others allow such consideration while denying that the only measure of good action is its consequences. See Solum, supra note 30, at 1463 (“Some deontological theories exclude consequentialist reasons altogether . . . . Other deontological moral theories allow for the consideration of consequences but maintain that considerations of fairness trump good consequences in other circumstances.”). 32. Robert Shaver, Egoism, in ENCYCLOPEDIA OF PHILOSOPHY, supra note 29, http://plato. stanford.edu/archives/win2002/entries/egoism. A related theory is rational egoism, which argues that it is rational for a person to act in her self-interest and irrational not to. Id. Egoist theories allow for the possibility that cooperation among people is possible, but they assert that cooperation occurs only when the mutual best interests of the parties dictate cooperation. Id. 1146 Texas Law Review [Vol. 87:1137 lose rights that the American adversarial system affords nonclass litigants. Class members do not get the opportunity to decide with whom and against whom to file suit; they do not control the forum, the timing of their cases, or the presentation of the proofs and arguments concerning their claims. Mem- bers of the class will probably never talk with their lawyer; their lawyer will likely never know the details and circumstances of their individual claims. Indeed, class members might never know that they are parties in the case. Yet the judgment obtained on the class’s behalf binds the members to the outcome. The license to act self-interestedly by refusing to join another’s lawsuit collapses.41 Strong reasons must accompany such a fundamental reordering of the litigation paradigm. At the federal level, Rule 23(b) provides four such reasons.42 First, a class action can be certified when the plaintiffs’ autonomy to bring separate actions creates a risk of “inconsistent or varying adjudications . . . that would establish incompatible standards of conduct for” the defendant.43 Second, a class action can be certified when one plaintiff’s autonomous decision to bring a lawsuit might result in a judgment “that, as a practical matter, would be dispositive of the interests of the other members” of the class or “would substantially impair or impede their ability to protect their interests.”44 Third, class treatment is proper when “final injunctive re- lief or corresponding declaratory relief is appropriate respecting the class as a whole.”45 Finally, a class action may be maintained when “the questions of law or fact common to class members predominate over any questions af- fecting only individual members” and “a class action is superior to other available methods for fairly and efficiently adjudicating the controversy.”46 It is possible to restate and unify these four scenarios in a single principle: An individual plaintiff’s autonomy to choose not to join with other plaintiffs stops when that choice threatens harm to others47—whether that harm is to other potential plaintiffs, who might find their own interests 41. As I describe infra at notes 54–56 and accompanying text, in some circumstances a class member can retain a measure of autonomy by opting out of the class. 42. In addition, a class action must meet the six requirements of Rule 23(a). These requirements can loosely be grouped into two meta-requirements: efficiency and adequate representation. Gen. Tel. Co. of the Sw. v. Falcon, 457 U.S. 147, 157 n.13 (1982). For further discussion, see infra notes 78–79 and accompanying text. 43. FED. R. CIV. P. 23(b)(1)(A). 44. FED. R. CIV. P. 23(b)(1)(B). 45. FED. R. CIV. P. 23(b)(2). 46. FED. R. CIV. P. 23(b)(3). In assessing these two requirements of predominance and superiority, Rule 23(b)(3) lists four relevant factors to consider: “the class members’ interests in individually controlling the prosecution” of their claims; “the extent and nature” of any individual actions that class members have already brought; “the desirability or undesirability of concentrating the litigation of the claims in the particular forum”; and “the likely difficulties in managing a class action.” FED. R. CIV. P. 23(b)(3)(A)–(D). 47. Again, the concept of “harm” does not encompass the imposition of the ordinary costs of good faith litigation on the defendant or the court system. See supra note 28 and accompanying text. 2009] Rethinking Adequacy of Representation 1147 impaired by the plaintiff’s choice to sue separately; to the defendant, who might be whipsawed by inconsistent judgments of multiple plaintiffs exer- cising their choices to sue separately or else beaten down by the costs of repetitive litigation; or to society, which finds individual litigation too expen- sive a luxury to subsidize or too vagarious a method to deter illegal conduct. The harm must be more than de minimis: Rule 23 allows a class action only when the class of plaintiffs “is so numerous that joinder of all members is impracticable.”48 Once that threshold is crossed, however, Rule 23 sacrifices individual class members’ abilities to pursue self-interested litigation strate- gies to the achievement of a greater good. Cast in this light, the class action embodies a consequentialist perspective to joinder: Avoidance of negative consequences, not the inculcation of virtue in litigants or the fulfillment of a priori litigation duties toward others, is its principal justification. Indeed, Rule 23 works out, in a particular litigation context, the limits of autonomy. Captured in the famous aphorism that my right to swing my fist ends where your nose begins,49 a central problem in consequentialist ethical theory is to specify the point (if any) at which a person’s autonomy to act must yield to the harm that those actions cause others. Rule 23’s answer to that question appears to reject an egoist stance and to embrace a rule-utilitarian approach to class treatment.50 In two important ways, however, Rule 23 retains an orientation toward individual autonomy. First, class members retain some rights to promote or protect their individual claims. In all class actions, class members may seek to intervene to “present claims or defenses, or otherwise to come into the action,”51 and in Rule (b)(3) class actions, class members are entitled to 48. FED. R. CIV. P. 23(a)(1). 49. The saying is usually used as a shorthand description of Mill’s argument for liberty. See JOHN STUART MILL, ON LIBERTY (1859), reprinted in ON LIBERTY WITH THE SUBJECTION OF WOMEN AND CHAPTERS ON SOCIALISM 13–16, 56–57, 75–76, 94–96 (Stefan Collini ed., Cambridge Univ. Press 1989). 50. Utilitarian theory can be divided into “act utilitarianism” and “rule utilitarianism.” Under the former, the moral significance of an action is determined by whether that specific action maximizes utility. A problem with act utilitarianism is the costliness of engaging in this inquiry for every action. Rule utilitarianism seeks to eliminate those costs by establishing rules of behavior that, in the main, maximize utility; even though the rule does not maximize for each action, the savings from not needing to engage in an action-by-action evaluation exceed the losses in utility in specific cases. Sinnott-Armstrong, supra note 29. Rule 23’s four scenarios for class treatment could thus be seen as the rule makers’ conclusion about the occasions when, in the main, the costs of permitting individual litigation exceed the costs of the lost autonomy that class certification entails. Some might argue that Rule 23 even reflects an altruistic (rather than a utilitarian) impulse, allowing individuals to sacrifice their autonomy to bring their own claims and to take up the burden of championing the claims of others. Although I do not deny the possibility that a class representative and class counsel might take on class litigation for altruistic reasons, the ensuing paragraphs describe the reasons that altruism constitutes a poor justification for Rule 23. 51. FED. R. CIV. P. 23(d)(1)(B)(iii), 24(a), 24(b); see THOMAS E. WILLGING ET AL., FED. JUDICIAL CTR., EMPIRICAL STUDY OF CLASS ACTIONS IN FOUR FEDERAL DISTRICT COURTS 56 (1996) (indicating that intervention occurred in 0% to 11% of cases in four class-action data sets). 1148 Texas Law Review [Vol. 87:1137 appear through their own counsel.52 Moreover, filing a class action does not prevent class members from also filing individual cases, or even prevent “dueling” class actions; as long as they beat the class action in the race to judgment or settlement, individual class members can enjoy the benefits of their individual resolution.53 The most significant provision retaining individual autonomy, however, is the opt-out right for (b)(3) class members.54 In effect, the (b)(3) opt-out right returns to the individual the decision of whether to join with others or to strike out independently. By requiring an affirmative decision to opt out, rather than by adopting a Rule 20(a)-like demand that each plaintiff affirma- tively opt in,55 Rule 23(b)(3) flips the consequence for doing nothing from nonjoinder to joinder. This flip in the default rule is hardly insignificant; the apathetic plaintiff, the tardy plaintiff, and the plaintiff who never receives notice of the right to opt out lose control of their lawsuits. Nonetheless, in light of the prevalence of (b)(3) class actions over either (b)(1) or (b)(2) class This opportunity to intervene is something less than the full “master of the complaint” autonomy guaranteed individual litigants, because the appearing class member loses the right to determine when, where, with whom, and against whom to file suit. In addition, most courts granted only permissive intervention under Rule 24(b), id., and courts can impose significant limitations on the opportunity of permissive intervenors to participate, see 7C CHARLES ALAN WRIGHT ET AL., FEDERAL PRACTICE AND PROCEDURE § 1922 (3d ed. 2007). 52. FED. R. CIV. P. 23(c)(2)(B)(iv). 53. See Epstein v. MCA, Inc., 179 F.3d 641, 643–44 (9th Cir. 1999) (holding that class members in a state class action who commenced a separate federal class action were precluded when the state class action won the race to settlement); Rhonda Wasserman, Dueling Class Actions, 80 B.U. L. REV. 461, 462 (2000) (noting that dueling class actions are “rampant” and discussing the pressure to settle first on the part of plaintiffs attorneys). Although its reasoning is suspect, one court has held that a class-certification order in a mandatory class action effectively bars separate actions involving the same claims, but has further held that this bar violates the Anti-Injunction Act, 28 U.S.C. § 2283 (2000), when applied to cases filed in state court. In re Fed. Skywalk Cases, 680 F.2d 1175, 1180, 1183 (8th Cir. 1982); cf. Cooper v. Fed. Reserve Bank of Richmond, 467 U.S. 867, 880 (1984) (holding that the individual claims of class members not encompassed within the class allegations are not precluded by the class judgment). 54. See FED. R. CIV. P. 23(c)(2)(B)(v). Because (b)(1)(A), (b)(1)(B), and (b)(2) class actions lack a comparable opt-out right, they are often called “mandatory” class actions. See Ortiz v. Fibreboard Corp., 527 U.S. 815, 833 n.13 (1999) (contrasting (b)(1) class actions’ lack of opt-out or notice rights to (b)(3) class actions, which include these rights). Courts have on rare occasion nonetheless fashioned an off-book, opt-out right for class members in mandatory class actions when an opt-out right would not create the types of harm that the (b)(1) and (b)(2) class actions were meant to avoid. See, e.g., Eubanks v. Billington, 110 F.3d 87, 94–95 (D.C. Cir. 1997) (holding that courts have discretion to fashion opt-outs in (b)(1) and (b)(2) class actions when necessary to “protect the rights of individual class members”); County of Suffolk v. Long Island Lighting Co., 907 F.2d 1295, 1302 (2d Cir. 1990) (“Rule 23 does authorize a district court to allow a class member to opt out of a [(b)(1) action] under some circumstances . . . .”); cf. Phillips Petroleum Co. v. Shutts, 472 U.S. 797, 811 n.3 (1986) (leaving open the question of whether mandatory class actions are consistent with the Due Process Clause). 55. Cf. 29 U.S.C. § 216 (2000) (providing an opt-in process for Fair Labor Standards Act class actions); John Bronsteen & Owen Fiss, The Class Action Rule, 78 NOTRE DAME L. REV. 1419, 1446–47 (2003) (arguing that class members should be required to opt into a settlement class). 2009] Rethinking Adequacy of Representation 1151 toward their positions—and indifference toward the positions of others—that class members have been required to sacrifice. C. The Requirement of Adequate Representation The bridge spanning the gulf between the interests of class members and the actions of the class representative and class counsel is the doctrine of adequate representation. The doctrine handles two distinct problems, one of incompetence and one of indifference. “Incompetence” concerns class repre- sentatives and class counsel who sincerely (whether for virtuous, deontological, altruistic, or utilitarian reasons) want to represent the interests of class members, but are incapable of effectively doing so because of insuf- ficient financing, experience, talent, probity, or mental capacity. “Indifference” concerns egoist class representatives and class counsel who are willing to represent the interests of class members only to the extent that such representations serve their own interests.62 The desire to address these problems does not, however, fully specify the content of the doctrine. Although class actions have ancient roots, shap- ing the contours of adequate representation is a fairly recent phenomenon.63 The phrase shows up in the 1938 version of Rule 23.64 Before then, it was subsumed within the idea that class actions were maintainable when the rep- resentative’s claim was one of “common interest” (or “general interest”) to a 62. I use “indifference” as an umbrella term to describe two different possibilities. The first is true indifference; the class representative and class counsel do not care one way or the other about the interests of those they represent, except to the extent that such interests are useful to advancing their own interests. The second is actual hostility, in which promoting the interests of some or all class members will injure the interests of the class representative or class counsel. Cases of hostility are rare, but they exist. See supra note 57; infra notes 104–07 and accompanying text; infra text accompanying notes 119, 134. Because they present the same analytical problem—an unwillingness at some point to advance the interests of class members—I consider the situations together. 63. See generally STEPHEN C. YEAZELL, FROM MEDIEVAL GROUP LITIGATION TO THE MODERN CLASS ACTION (1987) (describing the emergence, from its medieval roots, of the twentieth-century class action, which for the first time clearly operated on a “congruence of interest” model). 64. See FED. R. CIV. P. 23(a), 308 U.S. 689 (1939) (revised 1966) (permitting a class action when, among other things, a class representative “will fairly insure the adequate representation of all”). Cases before the 1930s tended not to use the phrase “adequate representation” or variants like “adequacy of representation.” The earliest use of the phrase that I could find was in 1889. See Lancashire Ins. Co. v. Maxwell, 5 N.Y.S. 399, 401 (N.Y. Sup. Ct. 1889) (stating that according to New York’s rules of civil procedure, one prerequisite of aggregate suits is that the plaintiffs “may be adequately and efficiently represented by one or more of their number less than the whole”). The earliest federal reference—and an ambiguous one at that—was in 1912. See Carpenter v. Knollwood Cemetery, 198 F. 297, 300 (D. Mass. 1912) (citing 1 ROGER FOSTER, FEDERAL PRACTICE 317 (4th ed. 1909)) (asserting that one trustee is held “adequately to represent” the rest of the trustees in a case involving a deed that secured the rights of real property in a large number of beneficiaries). 1152 Texas Law Review [Vol. 87:1137 group of parties too numerous to join.65 But we can trace the rise of the modern concept of adequate representation to 1940, with Hansberry v. Lee.66 Hansberry involved a racially restrictive covenant that was arguably invalid.67 But in a prior class action, Burke v. Kleiman,68 in which a class of signatory landowners sued one of their number who had violated the covenant, the class and the defendant stipulated, and the court found, that the covenant was valid.69 When an African-American family, the Hansberrys, bought another property subject to the same covenant, one of their defenses against an eviction proceeding brought by a class of landowners was the covenant’s invalidity.70 The plaintiffs argued that the Hansberrys were pre- cluded from raising the argument—that the question of validity had been decided in Burke, and that, as successors in interest to a class member in that action, the Hansberrys were bound by that decision.71 The Illinois courts agreed,72 but the U.S. Supreme Court reversed.73 The basis for the Supreme Court’s decision was the inadequacy of the representation that the Hansberrys received in Burke. The Court explained: It is familiar doctrine of the federal courts that members of a class not present as parties to the litigation may be bound by the judgment where they are in fact adequately represented by parties who are present . . . . . . . . It is one thing to say that some members of a class may represent other members in a litigation where the sole and common interest of the class in the litigation, is either to assert a common right or to chal- lenge an asserted obligation. It is quite another to hold that all those who are free alternatively either to assert rights or to challenge them are of a single class . . . . Such a selection of representatives for pur- poses of litigation, whose substantial interests are not necessarily or even probably the same as those whom they are deemed to represent, does not afford that protection to absent parties which due process requires. . . . Apart from the opportunities it would afford for the 65. See Beatty v. Kurtz, 27 U.S. (2 Pet.) 566, 566 (1829) (recognizing that “this is one of those cases in which certain persons . . . having a common interest, may sue in behalf of themselves and others . . . for purposes common to all, and beneficial to all”); Frederick v. Douglas County, 71 N.W. 798, 799 (Wis. 1897) (quoting a statute authorizing suit “when the question is one of common or general interest of many persons”). 66. For a discussion of the facts and historical context of the case, see Jay Tidmarsh, The Story of Hansberry: The Rise of the Modern Class Action, in CIVIL PROCEDURE STORIES 233 (Kevin M. Clermont ed., 2d ed. 2008). 67. Hansberry v. Lee, 311 U.S. 32, 37–38 (1940). 68. 277 Ill. App. 519 (App. Ct. 1934). 69. Id. at 522; see also Tidmarsh, supra note 66, at 243–44. 70. Hansberry, 311 U.S. at 38; see also Tidmarsh, supra note 66, at 262. 71. Hansberry, 311 U.S. at 38; see also Tidmarsh, supra note 66, at 262. 72. Hansberry, 311 U.S. at 38; see also Tidmarsh, supra note 66, at 271. 73. Hansberry, 311 U.S. at 38. 2009] Rethinking Adequacy of Representation 1153 fraudulent and collusive sacrifice of the rights of absent parties, we think that the representation in this case no more satisfies the require- ments of due process than a trial by a judicial officer who is in such situation that he may have an interest in the outcome of the litigation in conflict with that of the litigants.74 Sensibly enough, the Hansberry Court thought that the representative in Burke could not adequately represent both the interests of those home owners who, like the representative herself, wanted to enforce the covenant and the interests of those home owners who, like the Hansberrys, wanted not to en- force the covenant: [T]hose signers or their successors who are interested in challenging the validity of the agreement and resisting its performance are not of the same class in the sense that their interests are identical so that any group who had elected to enforce rights conferred by the agreement could be said to be acting in the interest of any others who were free to deny its obligation.75 Hansberry presented a case of indifference rather than a case of incompetence. The self-interest of the class representative in Burke was to enforce the restrictive covenant. The best way to do so was to bring a class action, which bound every signatory to the determination that the covenant was enforceable. The class representative and class counsel76 were at best indifferent, and more likely were hostile, to the interests of those members of the class that had no interest in enforcing the covenant.77 Thus, Hansberry harbored no illusions about the goodness of humanity; it accepted the reality that class representatives were going to act self- interestedly—or egoistically. Its solution was to require a tight alignment of the interests of class representatives and the interests of class members. Even if class representatives were self-interested, the rising tide of their self- 74. Id. at 42–45 (citations omitted). 75. Id. at 44. 76. Although Hansberry focused on the deficiencies of the class representative in Burke, the defendants in Hansberry also alleged that the class counsel in Burke was deficient. Brief of Petitioners at 45, Hansberry, 311 U.S. 32 (No. 29). In particular, the defendants argued that class counsel colluded with a willing defense counsel to enter a fraudulent stipulation that the covenant was valid. Id. This allegation of collusion was never proven, although it formed the backdrop for the Supreme Court’s decision. See Hansberry, 311 U.S. at 45 (noting “the opportunities” that the representation of conflicting interests “would afford for the fraudulent and collusive sacrifice of the rights of absent parties”); Tidmarsh, supra note 66, at 267–70. 77. Presumably, all signatories to the covenant had an interest in its enforcement when they signed the agreement during 1927 and 1928. But those interests began to change over time, and the covenant ran with each property for twenty-five years. Tidmarsh, supra note 66, at 239. The combination of the Great Depression, the changing racial composition of the surrounding neighborhoods, white flight, and the ready supply of African-American families willing to pay a premium for housing made some of the original signatories regret their decisions. See id. at 240 (“The inexorable law of supply and demand has already washed away restrictive covenants in other areas, and it seemed unlikely that the Association could keep its homeowners from eventually breaking rank.”). In addition, some new owners of the property likely had different attitudes toward the covenants than the original owners. Nagareda, supra note 7, at 288. 1156 Texas Law Review [Vol. 87:1137 requirements, but ones that are hard to tease out of a doctrine concerned with conflicts and collusion. That disconnection is itself a small point. But it feeds into two larger problems in the progression from Hansberry to Amchem. The first is the movement from dichotomous to polycentric interests. In Hansberry there were essentially only two possible positions that class members might have wished to take: either the covenant was valid and therefore enforceable, or the covenant was invalid and therefore unenforceable. In a world with such dichotomous choices, it is easy to recognize conflicts of interest. In Amchem, however, the interests were far more complex and diffuse. The factual and legal positions of class members varied considerably, and class members had a range of possible interests other than the dichotomous poles of “I favor set- tlement” and “I favor no settlement.” Hansberry’s identity-of-interests test works best when the interests in litigation break into readily opposing, clear- cut camps; Amchem shows that interests are often shades of grey. Because it involved only two dichotomous positions, Hansberry never needed to specify exactly how tightly aligned the interests had to be in order for a class repre- sentative to be regarded as adequate. Amchem might have done so, but the majority ducked the issue.87 So we do not know whether a representative can stand in only for those claimants whose interests are exactly the same shade of grey, or, if not, how close the shades of grey must be. Hansberry’s test for adequacy seems poorly suited for a world that involves a multiplicity of in- terests among class members. Second, and relatedly, in the vocabulary of ethics, the class action is a consequentialist—specifically, a utilitarian—device, which is principally concerned with preventing certain harms to potential class members, to the defendant, or to society.88 A doctrine requiring avoidance of conflicts of interest or collusion has a deontological ring to it; it is a duty-based requirement that must be satisfied regardless of the consequences. In Hansberry, it was not clear that adequate representation was a deontological principle. On one view, Hansberry “constitutionalized” adequate represen- tation to achieve a social good; it allowed an attack on a racially restrictive covenant to go forward. The petitioners in Hansberry had asked the Supreme Court to declare all racially restrictive covenants invalid, making the very arguments against restrictive covenants that the Court would adopt eight years later in Shelley v. Kraemer.89 In 1940, the Supreme Court was 87. At one point Amchem appeared to suggest that the settlement could have been saved if subclasses composed of members with comparable interests had been certified, instead of one large class that threw together people with varying interests. 521 U.S. at 627. But this suggestion came only in the Court’s quotation of a Second Circuit opinion, and in any event the Court did not make clear how many subclasses—or with what array of interests—would have been acceptable. 88. See supra note 47 and accompanying text. 89. 334 U.S. 1 (1948). 2009] Rethinking Adequacy of Representation 1157 not yet willing to go that far;90 adequacy of representation thus became a convenient vehicle that gave the Hansberry petitioners the chance to contest a particular covenant without the Court embroiling itself in far trickier con- stitutional and political questions whose outcome would have been in considerable doubt.91 By the time of Amchem, however, any sense that adequate representation was a utilitarian doctrine crafted with an eye toward achieving a good outcome was gone. The majority’s opinion both opened and closed with a lament about the social and judicial ills that asbestos litigation had caused.92 It recognized that the Amchem settlement might have alleviated those ills in some significant measure.93 Despite these benefits, however, the Court felt constrained to find a lack of adequate representation without weighing the benefits of settlement against the costs of less than ideal representation94—clearly a nonconsequentialist approach. The point of this critique is not to catch the Court in a bit of philosophical incoherence. Rather, it is to show that the doctrine of adequate representation—at least if it is understood to require strict avoidance of con- flicts of interest or collusion—has drifted from the moorings of the law to which it was attached. Class actions are designed to prevent harm to simi- larly situated claimants, to defendants, or to society. Class actions also impose costs,95 including two costs—depressed judgment (or settlement) val- ues for class members and a concomitant loss in optimal deterrence of the defendant96—that can occur when conflicts of interest or collusion make representation inadequate. Viewed consequentially, representation should be deemed inadequate only when the marginal cost of a more imperfect 90. One of the Court’s difficulties was that, in the recent past, it had dismissed an appeal of a case challenging the constitutionality and legality of racially restrictive covenants for want of a substantial federal question. Corrigan v. Buckley, 271 U.S. 323, 331–32 (1926). 91. For this interpretation of Hansberry, see YEAZELL, supra note 63, at 235–36. 92. Amchem, 521 U.S. at 597–99, 628–29. 93. Id. at 628–29; see id. at 629 (Breyer, J., dissenting) (“[T]he need for settlement in this mass tort case, with hundreds of thousands of lawsuits, is greater than the Court’s opinion suggests.”). 94. This turn seems particularly out of place because Hansberry v. Lee located the adequacy-of- representation doctrine in the Due Process Clause. 311 U.S. 32, 45 (1940); see also id. at 42 (stating that the due process and full faith and credit clauses shape the doctrine). Today, due process is mainly regarded as a consequentialist doctrine, balancing the costs of greater procedural protections against their benefits. See Mathews v. Eldridge, 424 U.S. 319, 334–35 (1976) (identifying three factors the Court considers when determining whether a government action is in harmony with procedural due process: (1) the private interest that will be affected by the state action; (2) the effectiveness of the proposed procedural requirements in safeguarding against due process deprivations; and (3) the burdens such procedural requirements would place upon the government). For further discussion, see infra notes 193–94 and accompanying text. 95. One example is the potential cost of inaccuracy due to an increased probability of recovery from class aggregation. See supra note 28. 96. On the deterrence effect of class actions, see David Rosenberg, Class Actions for Mass Torts: Doing Individual Justice by Collective Means, 62 IND. L.J. 561, 570 (1987), which explains that adhering to a disaggregative dispute-resolution system with regard to mass torts reduces the deterrence effect that threatened liability normally has on defendants. 1158 Texas Law Review [Vol. 87:1137 representation, when added to the other costs of class treatment, exceeds the costs to similarly situated claimants, to the defendant, or to society that non- class litigation would impose. But this is not the inquiry in which Amchem engages.97 Finally, treating adequate representation as an absolute bar to class treatment ignores a critical fact. Conflicts of interest are not occasional oc- currences involving a shady class counsel or two. Rather, as I describe in the next Part, conflicts of interest are built into the very fabric of our present class-action regime. If we are serious about enforcing as tight an alignment of interests as Hansberry and Amchem seem to require, we must abandon Rule 23, or rewrite it. II. Adequacy’s Conundrum: The Inevitability of Conflicts of Interest in American Class Actions The four circumstances in which Rule 23(b) permits class treatment almost invariably bring into a single class people with antagonistic interests. This Part details the reasons for this perhaps surprising conclusion. In brief, the four class-action scenarios described in Rule 23(b) seek to prevent harms to other claimants (Rules 23(b)(1)(A) and, to some extent, 23(b)(2)), to de- fendants (Rules 23(b)(1)(B) and, to some extent, 23(b)(2)), or to society (Rule 23(b)(3)). To minimize these harms, people with diverging interests must be placed side-by-side in the class, and those interests must be adjusted one against the other. The observation that conflicts inhere in many class actions is hardly new.98 This Part, however, is the first systematic effort to prove why that observation is correct. In doing so, this proof raises the disconcerting possi- bility that virtually all class actions exceed the due process strictures that Hansberry imposes on adequate representation, and are therefore unconstitu- tional. 97. Had Amchem undertaken such an inquiry, it is not clear what the outcome would have been. The settlement contained complex provisions that tended to set awards around the historical averages for asbestos settlements, but also tended to suppress high-end awards and made no allowance for inflation, thus, over time, suppressing all awards beneath real-dollar historical averages. There is some reason to believe that the willingness of class counsel to accept a low settlement was related to the arguable conflicts of interest that they had, both in simultaneously representing individual claimants and class claimants and in simultaneously representing present class claimants and future class claimants. For a full description of the settlement’s terms, see JAY TIDMARSH, FED. JUDICIAL CTR., MASS TORT SETTLEMENT CLASS ACTIONS: FIVE CASE STUDIES 51–54 (1998), available at http://www.fjc.gov/public/pdf.nsf/lookup/Tidmarsh.pdf/$file/Tidmarsh. pdf. 98. See, e.g., Silver & Baker, supra note 10, at 1468 (“Conflicts of interest and associated tradeoffs among plaintiffs are an unavoidable part of all group lawsuits and all group settlements.”). 2009] Rethinking Adequacy of Representation 1161 • Scenario Two: The claimants do not have identical remedial claims. The self-interested class representative might hope that, after using the class’s number to help establish liability, she can effectively denigrate or devalue the claims of fellow class members, thus getting a larger share of the judgment. For instance, assume that class treatment raises the chance of recovery only to 70%, but the class representative expects to be able to shade the remedial proof to enhance her claim to $12,000. Still assuming pro rata litigation costs of $3,000, her ex- pected recovery is $6,300, so she will seek class certification. • Scenario Three: The self-interested plaintiff’s preferred counsel agrees to take the case only as a class action, so that class treatment becomes the only viable way to bring suit. For instance, using the numbers described in the first example, the plaintiff might expect that retaining the preferred counsel will increase the chance of winning from 40% to 80%. On these numbers, it is rational for the class repre- sentative to assent to a class action. • Scenario Four: The putative class counsel, with an eye only on the fee award, advises the putative class representative to seek class status without advising her that she stands to lose money from class treatment—and the putative representative lacks the legal acumen to realize that class treatment is not in her best interests. Whichever reason is at work, class treatment ends up bundling together plaintiffs whose interests are not aligned. The class representative and the members of the class are locked in a competition for a limited resource— whether an injunction that cannot be perfectly tailored to all their interests or a fund that is insufficient to satisfy fully all their claims. The interests of the class representative, class counsel, and class members might converge on certain issues—in particular, on the issue of the defendant’s liability.107 On the issue of remedy, however, class representatives and class members who are competing with each other for a resource that cannot fully satisfy all claimants lack an “identity of interests.” Thus, there is reason to doubt that the self-interested class representative, who wants the relief shaped as fa- vorably to her interests as possible, will adequately represent the interests of those with different interests. True enough, but a doubting Thomas might point to two situations in which this self-interest will not manifest itself. The first occurs when the class members do not disagree about either the theory of liability or the na- ture of the relief they seek. For example, under the relevant law, the choice might be either injunction X or no injunction; every potential plaintiff wishes to obtain exactly the same remedy, which will apply to and equally benefit 107. But not always. To the extent that the nature of the defendant’s liability determines the shape of the relief, plaintiffs interested in different relief might not agree on which theory of liability to pursue. 1162 Texas Law Review [Vol. 87:1137 all.108 Here, our Thomas might rejoin, the rising tide of the plaintiff’s self- interested desire to get injunction X really does lift all class members’ boats. Even in this situation, however, conflict exists. In this example, the only rea- son that a self-interested putative class representative would seek class certification is to offload some of the costs of litigation onto class members. Although it is difficult to have much sympathy for class members who are hoping to be free riders on an individual suit, the self-interest of class members—a self-interest that I have not discussed before, but one that exists every bit as much as the self-interest of the class representative—is not to certify the case, for class members will be required to pay a pro rata share of litigation expenses for an injunction that they otherwise would have obtained for free. We can (and do) ignore the free-rider problem in (b)(1)(B) and (b)(2) cases, but ignoring it does not mean that the antagonism of interests goes away.109 The second situation a doubting Thomas might raise involves the use of subclasses, which Amchem has suggested as a solution to the polycentric in- terests of class members.110 The idea of subclasses is to subdivide plaintiffs 108. It is not clear how many cases this hypothetical encompasses. For the sake of argument, I am willing to admit that some cases fall within this description. 109. Because putative class members cannot opt out of (b)(1)(B) and (b)(2) class actions, see supra note 54, the free-rider problem cannot be assuaged by allowing class members to exit. It is interesting to note that the conflict of interest here does not concern the substance of the claim, but rather the litigation strategy: whether to seek class certification. Neither Hansberry nor Amchem involved a disagreement over litigation strategy, nor has the Court ever held that disagreements over strategy are a sufficient reason for finding a conflict of interest. Indeed, it is difficult to imagine that it would, for then no class action could be certified when any class member objected to certification. What makes this situation different from a simple disagreement over strategy, however, is the reason for the disagreement: Class members wish to see the class not certified because they will obtain a better outcome (i.e., the same injunction at less expense) without certification. Conversely, the class representative is making herself better off by making class members worse off. In both Hansberry and Amchem, comparable disagreements over the remedial outcome were sufficient to find a conflict of interest. The one situation in which no antagonism appears to exist occurs when it is too costly for any claimant to bring the injunctive claim on her own. In such a case—the injunctive equivalent of the “negative-value suit” discussed infra in the text following note 124—it appears to be in every class member’s interest to join the class. Nonetheless, as the discussion of negative-value suits will show, internal conflicts can arise in this situation. Moreover, unless the expected value of the injunction is very close to the cost of prosecuting the claim, so that the claim becomes viable only when litigation costs are spread across every member of the class, the free-rider problem still exists: An absent class member would prefer for the class definition to be broad enough to spread the litigation costs among enough people to make the class action viable, but narrow enough to exclude her. It is also worth noting that determining a lack of conflict by calculating the net effect of a class action on the interests of the class members is a consequentialist approach to determining conflicts of interest—precisely the approach that Amchem’s deontological analysis eschewed. See Amchem Prods., Inc. v. Windsor, 521 U.S. 591, 621 (1997) (“[T]he standards set for the protection of absent class members serve to inhibit appraisals of the chancellor’s foot kind—class certifications dependent upon the court’s gestalt judgment or overarching impression of the settlement’s fairness.”). 110. Amchem, 521 U.S. at 627; see FED. R. CIV. P. 23(c)(5) (“When appropriate, a class may be divided into subclasses that are each treated as a class under this rule.”); see supra note 87 and accompanying text. Amchem’s suggestion was mild: After noting that class representatives “served 2009] Rethinking Adequacy of Representation 1163 into smaller groups, each of which shares an identity of interests. Although useful in theory, subclassing is far less helpful in practice for two reasons. First, each subclass must be large enough not to run afoul of the numerosity requirement111 but still contain within it no remedial antagonism among the members. Although it is possible to hypothesize such situations,112 they are rare in the real world. Second, subclassing along interest lines institutional- izes the conflict among competing interests, rather than eliminating it. Because subclasses are still part of a larger single class, the conflicts among class members remain. Third, although separate classes might overcome this problem, it is tricky to imagine how a court could engineer the filing—and consolidation into a single litigation—of separate class actions, each of which represents distinct, conflicting interests. Separate classes or subclasses require multiple class representatives and multiple class counsel. But separate classes or sub- classes undercut the incentives for self-interested class representatives and class counsel to seek certification. Assuming that willing representatives for multiple classes or subclasses are available (and this is hardly a given), mul- tiple classes loosen the self-interested class representative’s control of the litigation and her ability to steer the litigation in her favor, thus reducing her expected recovery from class treatment; multiple representatives further re- duce the number of people over whom the costs of litigation can be spread, thus increasing the pro rata costs. In the example above, if separate classing or subclassing reduces the expected recovery from 75% to 65%, and in- creases per-person litigation expenses to $500, the expected recovery from class treatment falls to $6,000—less than the $6,500 that a self-interested plaintiff might expect from individual treatment. Similarly, assuming that multiple willing class counsel are available, a self-interested lawyer will re- alize that successfully settling the case becomes harder with more classes, and the fee that the counsel can expect will be substantially less than she could expect in a scenario in which she was the sole lawyer for a larger class. generally as representative for the whole, not for a separate constituency,” it quoted a Second Circuit decision that suggested subclassing as a solution to the problem of adverse interests within a settlement class. Amchem, 521 U.S. at 627 (discussing In re Joint E. & S. Dist. Asbestos Litig., 982 F.2d 721, 742–43 (2d Cir. 1992), modified on reh’g sub nom. In re Findley, 993 F.2d 7 (2d Cir. 1993)). The Court never endorsed subclassing directly. 111. FED. R. CIV. P. 23(a)(1). In general, a class of 100 or more members meets this requirement; even smaller numbers can sometimes be satisfactory. See, e.g., Robidoux v. Celani, 987 F.2d 931, 935–36 (2d Cir. 1993) (permitting inexact numbers of a potential class size to satisfy the numerosity requirement); WRIGHT ET AL., supra note 80, § 1762 (emphasizing that there are no arbitrary rules establishing a required class size and there only needs to be a showing that joining all the members is impracticable). 112. For example, in the limited-fund hypothetical discussed supra in note 103 and accompanying text, rather than individual claimants, each entity with a $20,000 claim could be an unincorporated association or trust composed of thousands of members. On the further assumption that the members of each association have no legally cognizable internal conflicts about the proper distribution of its $20,000 claim, subclasses or separate classes composed of the members of each association would be numerous enough and also would not be remedially conflicted. 1166 Texas Law Review [Vol. 87:1137 • Scenario One: Later-filed suits seeking injunction Y or injunction Z threaten the ability of the defendant to provide the first plaintiff with the injunction X that she seeks, or threaten to undo the injunction. • Scenario Two: A class action increases the chances of winning or spreads the class representative’s cost of litigation across the class (or both), and these expected advantages outweigh any loss in the quality of the remedy that the plaintiff would have achieved in individual liti- gation. • Scenario Three: By taking control of the entire litigation, the self- interested plaintiff steers the case toward her preferred relief. For instance, the plaintiff might believe her claim to injunction X (or to the res) is better than that of other potential plaintiffs. A class action un- der (b)(1)(A) or (b)(2) becomes a way to spread the costs of litigation among class members, as well as a way to obtain a clear entitlement to injunction X or to secure clear title to the res that the competing claims of other class members might undo. • Scenario Four: The only way to attract a lawyer who will increase the chances of victory is to dangle the lure of the recovery of class- action fees; the increased chances of recovery exceed any potential loss in the quality of an individual remedy. • Scenario Five: Tantalized by the promise of larger class fees, class counsel misleads the putative class representative into believing that class treatment will lead (on balance) to a better remedy, even if it will not. In addition, another scenario is possible in the (b)(1)(A) and (b)(2) situations: A self-interested defendant could seek or consent to class certifi- cation to avoid the risk of multiple or inconsistent liability.120 As with the (b)(1)(B) and (b)(2) classes, the class now contains members that have competing interests. Indeed, by definition, the members of the class wish to subject the defendant to “inconsistent or varying” law- suits and “incompatible” remedies. Some of the class members have an interest in achieving one remedy, and others in achieving another remedy. Although in limited circumstances constituting separate classes according to group interests might provide an escape from the conundrum of conflicting 120. Although requests by defendants to certify class actions against them are in theory possible, see Rossetto v. Pabst Brewing Co., 71 F. Supp. 2d 913, 917 (E.D. Wis. 1999), rev’d on other grounds, 217 F.3d 539 (7th Cir. 2000), they are as rare as hens’ teeth. More typically, the defendant is likely to negotiate a settlement with a putative class representative and class counsel, and then support the plaintiffs’ request for class certification. See, e.g., Ortiz v. Fibreboard Corp., 527 U.S. 815, 82226 (1999) (noting a defendant’s attempt to reach a “global settlement” before supporting class certification). 2009] Rethinking Adequacy of Representation 1167 class interests, as a practical matter that solution is unlikely to be effective for the reasons already explained.121 Therefore, the (b)(1)(A) and (b)(2) “harm to defendants” class actions contain the same constitutional infirmity as the (b)(1)(B) and (b)(2) “harm to absent class members” class actions. If the Due Process Clause in fact re- quires an identity of interests among class members, a court cannot certify a (b)(1)(B) or (b)(2) defendant-harm class action. C. Conflicts of Interest when a Class Action Is the Superior Aggregation Mechanism: Rule 23(b)(3) Classes Unlike the (b)(1) and (b)(2) class actions, the (b)(3) class action is not designed specifically to prevent harm to absent plaintiffs or to the defendant. The (b)(3) class action applies principally to monetary claims; plaintiffs with injunctive claims will use Rules 23(b)(1) and 23(b)(2).122 The assumption underlying the (b)(3) class action is that the defendant has sufficient assets to satisfy putative class members’ monetary claims.123 Unlike a limited-fund or interpleader-type situation, serial suits by individual plaintiffs will not leave later-filing plaintiffs without adequate monetary remedies. Nor will a series of such suits leave the defendant exposed to multiple or inconsistent obliga- tions within the meaning of Rule 23(b)(1)(A); as courts have long recognized, inconsistent monetary judgments—some of which order a defen- dant to pay damages to some plaintiffs and others of which find the defendant not liable to other plaintiffs for the same behavior—do not consti- tute a harm to the defendant within the meaning of this rule.124 Instead, the focus of Rule 23(b)(3) is a more diffuse harm to society as a whole. Two types of social harm are the central cases. The first is inade- quate deterrence. The paradigmatic case is “large-scale, small-stakes” litigation, also known as “negative-value suits.” Assume that a credit-card company has illegally overcharged ten million customers $2 apiece. It is unlikely that very many, if any, individual cases will be filed; the costs of litigating each case exceeds the expected recovery (thus giving the suit a negative net value), and no attorney will work for such a small potential 121. See supra notes 110–14 and accompanying text. 122. In limited circumstances, a court can award monetary relief incidental to an injunctive claim in a (b)(2) class action. Courts are presently divided over how “incidental” the relief must be, but under no construction can Rule 23(b)(2) be used when the case is principally about monetary recovery. Compare Allison v. Citgo Petroleum Corp., 151 F.3d 402, 415 (5th Cir. 1998) (narrowly construing the monetary relief available in (b)(2) suits), with Robinson v. Metro-North Commuter R.R., 267 F.3d 147, 157 (2d Cir. 2001) (rejecting Allison’s approach). 123. Cf. Ortiz, 527 U.S. at 859–60 (reversing certification of a (b)(1)(B) settlement class action when the defendant retained some assets, thus making the action unsusceptible to limited-fund treatment). 124. See In re “Agent Orange” Prod. Liab. Litig., 506 F. Supp. 762, 789 (E.D.N.Y. 1980) (“Rule 23(b)(1)(A) is not meant to apply . . . where the risk of inconsistent results in individual actions is merely the possibility that the defendants will prevail in some cases and not in others, thereby paying damages to some claimants and not others.”). 1168 Texas Law Review [Vol. 87:1137 recovery. But this failure to deter the credit-card company has a significant social cost. By aggregating all ten million cases and by spreading the costs of litigation across all the claims, the suit becomes economically viable, a lawyer will now be willing to handle the case in return for a fee based on the class’s recovery, and society achieves a better level of deterrence. The second central case for the (b)(3) class action involves excessive litigation costs.125 Here the paradigmatic case is “large-scale, large-stakes” litigation, in which the expected net recoveries make individual lawsuits eco- nomically viable. Because the cases tend to arise out of a common pattern of conduct by the defendant and tend to involve a limited array of injuries, the same issues are litigated in case after case, courtroom after courtroom. At a certain point, this repetitive litigation becomes a socially unnecessary expense.126 A class action that brings together all of the claimants promises to reduce these costs, as well as the strain on the judiciary, substantially.127 But (b)(3) status is not always in an individual class member’s best interest. The recovery on a class claim can be lower128 and slower129 than recovery on an individual claim. Class treatment can also be unattractive to 125. Opt-out (b)(3) class actions impose certain costs not found in individual litigation, such as the often pricey cost of giving notice to class members of their right to opt out. See FED. R. CIV. P. 23(c)(2)(B) (delineating the requirements for individual notifications to class members). Therefore, when I refer to “excessive litigation costs,” I mean net excessive costs—the costs that repetitive individual litigation creates less the costs that (b)(3) litigation imposes. 126. There is an argument that a certain amount of repetitive litigation is socially useful. According to Professor McGovern’s thesis on maturity, large-scale, large-stakes cases often go through cycles, from the early cycle in which defendants win most cases to a cycle when plaintiffs respond to the early defense strategies and break through with some significant victories, before finally reaching a mature equilibrium when the chances of victory can be predicted with some statistical accuracy. See Francis E. McGovern, An Analysis of Mass Torts for Judges, 73 TEXAS L. REV. 1821, 1841–43 (1995) (discussing the three phases of mass-tort cases under a maturity theory). Arguably, the most accurate assessment of the value of cases occurs only after the equilibrium has been reached. Therefore, before reaching this equilibrium, repetitive litigation is justified as long as its marginal costs are less than the case’s marginal contribution to the establishment of a more accurate equilibrium. Once an equilibrium is reached, however, the costs of repetitive litigation cannot be justified. 127. Although it might seem that the reduction of these costs is an advantage to the parties as well as to society, that is not necessarily the case. For instance, from a defendant’s viewpoint, high litigation costs act as a barrier to entry for some plaintiffs. As long as the defendant’s litigation costs are less than the additional liability costs that the defendant would incur from a lower barrier to entry, the defendant will prefer to expend money on litigation costs. In particular, a litigation class action typically threatens a liability judgment far in excess of the savings that the defendant incurs in lower litigation costs. 128. See Horowitz & Bordens, Aggregation, supra note 25, at 226 (reporting on experimental results showing that “a plaintiff with a quite strong case . . . appears to be better served by being disaggregated, particularly with reference to punitive damages”). For data on typical recoveries in class actions, see Thomas E. Willging & Shannon R. Wheatman, Attorney Choice of Forum in Class Action Litigation: What Difference Does It Make?, 81 NOTRE DAME L. REV. 591, 638–40 & tbl.15 (2006), which provides data showing that in state class actions the median recovery was $850,000 and the median per capita recovery was $350, while in federal class actions the median recovery was $300,000 and the median per capita recovery was $517. 129. See JAMES S. KAKALIK ET AL., DISCOVERY MANAGEMENT 90 tbl.A-3 (1998) (noting the positive correlation between time to disposition and class certification). 2009] Rethinking Adequacy of Representation 1171 becomes more acute as the case moves from questions of general liability to the questions specific to each plaintiff (such as individual causation, damages, and case-specific defenses) that usually inhabit (b)(3) class actions. This lack of enthusiasm for proving the noncommon issues of class members is somewhat counterbalanced by the interests of class counsel, who has an incentive to maximize the fee and therefore, within limits, to enhance the value of class members’ claims. But that fact merely highlights a conflict of interest that pertains between the class representative and class counsel. Moreover, class counsel has an incentive to litigate noncommon issues only to the extent that doing so maximizes the fee—in other words, to expend time and money on increasing the size of the monetary recovery only to the point that an additional unit of time and money earns a greater fee than coun- sel can earn for other legal work. This point is not necessarily the point at which the recovery for either the class representative or the class members is at its maximum.141 Thus, counsel’s interest does not always align tightly with the interests of either the class representative or the class members. The misalignment of interests between the class and its counsel extends beyond the question of fees. One of the enduring realities of most (b)(3) class actions is that class counsel advances the funds for the litigation (including the often considerable money required to give notice).142 Losing a class action can visit financial ruin on class counsel. The combination of fronted costs and expected attorneys fees typically makes class counsel the largest stakeholder in the class action. In a large-scale, small-stakes case, the counsel’s stake can exceed that of any plaintiff by five or even six orders of magnitude. Even in large-scale, large-stakes cases, the disparity can be enormous. For instance, in the Blood Factor settlement class action,143 class development of cy pres and other doctrines for providing substituted remedies bears out the expected result. See In re Airline Ticket Comm’n Antitrust Litig., 307 F.3d 679, 689 (8th Cir. 2002) (approving cy pres distribution of unclaimed settlement proceeds, but modifying the distribution plan); Democratic Cent. Comm. v. Wash. Metro. Area Transit Comm’n, 84 F.3d 451, 454–56 (D.C. Cir. 1996) (describing cy pres and other approaches to distribution of funds when it was too costly to determine the victims of excessive transit charges); cf. Molski v. Gleich, 318 F.3d 937, 953–56 (9th Cir. 2003) (rejecting the settlement in a mandatory class action, and also rejecting the finding that the class representative and class counsel provided adequate representation, where the representative received $5,000, counsel received $50,000, class members received nothing, and the defendant instead made contributions to third-party advocacy groups). This cost–benefit calculus is not the deontological identity-of-interest analysis that Hansberry and Amchem appear to demand. 141. See supra note 610 and accompanying text; supra text accompanying notes 107 and 119. 142. Class counsel commonly advances the costs of class-action litigation. See 5 ALBA CONTE & HERBERT B. NEWBERG, NEWBERG ON CLASS ACTIONS § 15:21, at 78 (4th ed. 2002) (stating that cost advancements on a contingency basis are critical in class actions); supra note 106 and accompanying text. 143. This case, In re Factor VIII or IX Concentrate Blood Products Litigation (Blood Factor), 159 F.3d 1016 (7th Cir. 1998), was a class action for settlement brought by one class representative, Susan Walker, on behalf of all hemophiliacs infected with HIV that had used the defendant’s products, as well as their infected spouses and children, their estates, and all persons with derivative claims such as loss of consortium and emotional distress. See DEBORAH R. HENSLER ET AL., CLASS 1172 Texas Law Review [Vol. 87:1137 counsel received $40 million for fees and reimbursement of costs, while each of the 6,500 class members received $100,000.144 And counsel must deal with competitive pressures from overlapping class actions brought by other lawyers who pose a constant risk of settling the case out from underneath class counsel, thus leaving counsel holding the bag of uncompensated costs and time.145 Therefore, the risk tolerance of the class counsel is often differ- ent from that of the class representative and class members. In such a circumstance, it is understandable—even if not ethically laudable—when class counsel acts to protect that investment in the case rather than undertak- ing riskier strategies that match up better with the risk positions of the class.146 The superiority inquiry adds to the tension within the class. Superiority examines the issue of class certification from the viewpoint of society, en- suring that a class action is a better vehicle for resolving a case than other possible methods, such as individual litigation, consolidation,147 multidistricting,148 and so on.149 It is likely that some class members would have been better served by one of these alternate methods. But none of those methods better serves the interests of the class representative or class counsel; if they did, the representative or counsel would have employed them to begin with. The opt-out feature of (b)(3) class actions arguably softens these conflicting interests. It could be argued that class members waive any possible conflicts of interest by failing to opt out when afforded the ACTION DILEMMAS: PURSUING PUBLIC GOALS FOR PRIVATE GAIN 305 (2000) (describing the class of plaintiffs); id. at 315 n.59 (noting that Susan Walker is a fictitious name used to protect the privacy of the actual class representative). 144. Blood Factor, 159 F.3d at 1018; HENSLER ET AL., supra note 143, at 305–07. 145. There are cases in which defendants arguably pitted counsel in two overlapping class actions against each other in a reverse auction to drive down the settlement’s price. See Reynolds v. Beneficial Nat’l Bank, 288 F.3d 277, 282–83 (7th Cir. 2002) (remanding the case to determine if a reverse auction occurred); Epstein v. MCA, Inc., 126 F.3d 1235, 1250 (9th Cir. 1997) (withholding full faith and credit from a related state-court settlement because the conflicts of interest between class counsel in state-court action and plaintiff class amounted to inadequate representation), withdrawn on reh’g, 179 F.3d 641, 650 (9th Cir. 1999). 146. Although I do not explore them here, these pressures are also a reality in (b)(1) and (b)(2) class actions. Because those classes are less common, and they rarely involve monetary awards akin in size to (b)(3) awards, the degree of conflict between the interests of the class and the interests of counsel is often less. Cf. John C. Coffee, Jr., Conflicts, Consent, and Allocation After Amchem Products—Or, Why Attorneys Still Need Consent to Give Away Their Clients’ Money, 84 VA. L. REV. 1541, 1542, 1544–50 (1998) (demonstrating that serious conflicts of interest exist at the stage of allocating money among class members). 147. See FED. R. CIV. P. 42(a). 148. See 28 U.S.C. § 1407 (2000). 149. For a full discussion of potential aggregation methods, see generally JAY TIDMARSH & ROGER H. TRANGSRUD, COMPLEX LITIGATION: PROBLEMS IN ADVANCED CIVIL PROCEDURE 10– 200 (2002). 2009] Rethinking Adequacy of Representation 1173 opportunity to do so.150 But this argument proves too much, for it wipes out the doctrine of adequate representation in (b)(3) class actions. It puts those who do not opt out at the mercy of self-interested class representatives and class counsel, creating an incentive for rational class members to opt out and frustrating the hoped-for gains from class treatment. Indeed, treating a fail- ure to opt out as a waiver turns the idea of adequate representation on its head; Rule 23’s promised trade-off for giving up the right to bring an indi- vidual action is adequate representation—not the absence of adequate representation.151 It is possible to specify one (albeit uncommon) condition in which a class representative lacks a conflict of interest with class members. The situation occurs when a class of claimants have a joint entitlement to receive a lump-sum award, and the proper distribution of the award among the bene- ficiaries is not in dispute.152 Further assuming that any variations in the factual positions of the claimants create no varying legal positions or theories,153 there are no noncommon issues. This hypothetical situation is the monetary equivalent of the “injunction X” or “no injunction” situation de- scribed in connection with (b)(1)(B) and (b)(2) class actions.154 But the (b)(3) class action does not present a comparable free-rider conflict between the class representative and the class members. The reason is that the (b)(3) class action generates the fund of money from which expenses will be paid; had the class representative brought an individual action, the common fund would have been charged for these expenses in any event.155 Unless the 150. Hansberry v. Lee, which created the constitutional requirement of adequate representation, did not involve an opt-out class action. 311 U.S. 32, 39–44 (1940). 151. See Phillips Petroleum Co. v. Shutts, 472 U.S. 797, 810 (1985) (“Unlike a defendant in a normal civil suit, an absent class-action plaintiff is not required to do anything. He may sit back and allow the litigation to run its course, content in knowing that there are safeguards provided for his protection.”). In addition, the waiver argument fails to account for the possibility that, even if representation is adequate at the time that the opt-out right is provided, subsequent events could render the class representation inadequate. See id. at 812 (“[T]he Due Process Clause of course requires that the named plaintiff at all times adequately represent the interests of the absent class members.”); cf. FED. R. CIV. P. 23(e)(4) (permitting, but not requiring, a court to give a second opt- out right at the time of settlement). 152. For instance, the claimants might be the numerous beneficiaries of a trust or the members of a corporation or unincorporated association suing on behalf of the trust, corporation, or association because of an action that injured the trust or corporation, not just the interests of some beneficiaries of shareholders. Cf. FED. R. CIV. P. 23.1–23.2 (derivative actions and actions by unincorporated associations). 153. For example, in a case brought by shareholders against a corporation, the actions of the officers or the board of directors might be called into question; if the officers or board also own shares, the class’s conflicting positions take this situation outside of the narrow exception that I am describing. 154. See supra notes 108–09 and accompanying text. 155. See Bloomer v. Liberty Mut. Ins. Co., 445 U.S. 74, 88 n.15 (1980) (acknowledging “the established power of a court of equity to charge beneficiaries with a proportionate share of the costs of creating a common fund through litigation”). 1176 Texas Law Review [Vol. 87:1137 the conflicting interests within the class is simple: The representation provided to a class member is adequate if and only if the actions of the class representative and class counsel can reasonably be expected to place that class member in no worse a position than that class member would have en- joyed had she retained control of her own case. More simply, the principle might be stated: Do no harm. So stated, the principle seems self-evident. Representation that leaves some class members worse off than they would have been without class cer- tification must be inadequate, right? Surprisingly, the answer has not always been “yes.” The poster-child counterexample occurred in Kamilewicz v. Bank of Boston Corp.,162 in which a state-court, negative-value class settle- ment resulted in a $2.19 recovery and a $91.33 assessment of attorneys fees for some class members—thus leaving these class members owing $89 as a result of their lawsuit.163 The federal court refused to reexamine the state court’s decision upholding the reasonableness of the settlement.164 Admittedly, cases with Kamilewicz’s egregious facts are rare;165 typically, a successful class action garners a positive recovery for class members. But that fact does not mean that the representation was adequate in the sense that I describe; adequacy does not hinge on a positive result in absolute terms, but rather a positive result in comparative terms, so that the net result from class treatment equals or exceeds the net result from a class member’s retention of control over her own case. It is not difficult to imagine circumstances in which a class action might result in a positive result for class members but still fail the adequacy-of-representation test that I propose. At the same time, the requirement is fairly minimal—a modest and elementary principle of justice rather than an extraordinary and heroic burden.166 It is easy to comprehend and, in most circumstances, easy to ap- ply. Indeed, it is possible to state the condition of adequate in mathematical form. Let PI represent that probability of a class member’s recovery in 162. 92 F.3d 506 (7th Cir. 1996). 163. The total amount awarded for attorneys fees was in dispute, but lay somewhere between $8.5 million and $14 million. Id. at 508. 164. The underlying state-court class action involved an alleged overcharge in a mortgage escrow account. Kamilewicz, 92 F.3d at 508. The plaintiffs claimed in their federal suit that class counsel and the defendants in the state-court class action had colluded in the settlement, thus violating federal racketeering laws and their civil rights, and further committing fraud, conversion, and malpractice. Id. at 509. The district court dismissed the suit on the basis of the Rooker– Feldman doctrine, which holds that federal courts cannot act as appellate courts over the decisions of state courts. Id. at 509–12. The Supreme Court denied certiorari. Kamilewicz v. Bank of Boston Corp., 520 U.S. 1204 (1997). For a scathing critique of the decision, see Susan P. Koniak, How Like a Winter? The Plight of Absent Class Members Denied Adequate Representation, 79 NOTRE DAME L. REV. 1787, 1808–17 (2004). 165. But see Koniak, supra note 164, at 1797 (suggesting that class-action litigation is rife with abuse). 166. Indeed, the “do no harm” principle is a generalized statement of the reason that no conflict of interest exists in the Rule 23(b)(3) class action that I described supra at notes 152–58 and accompanying text. 2009] Rethinking Adequacy of Representation 1177 individual litigation, LI the size of the recovery in individual litigation, FI the attorneys fees in individual litigation, and CI the costs of individual litigation. And let PC represent that probability of a class member’s recovery in class litigation, LC the size of the recovery in class litigation, FC the member’s pro rata share of attorneys fees in class litigation, and CC the member’s pro rata share of costs in class litigation. Adequate representation exists when: (PI × LI) − (FI + CI) ≤ (PC × LC) − (FC + CC), where 0 ≤ (PC × LC) − (FC + CC). 1. Explaining the Operation of the Principle.—This formulation allows me to explore a number of the central features of this adequate-representation principle. First, as the P × L term reflects, adequacy of representation de- pends on expected recoveries; it is determined before the fact, not based on how the case actually turns out.167 Second, as the F + C term reflects, ade- quacy of representation depends on net, rather than gross, recovery. Individual litigation can lead to an award (LI) that is higher than a class award (LC). 168 Compared to individual litigation, however, a class action can increase P and decrease F + C, which together raise the net expectancy de- spite the lower award.169 In effect, therefore, the “do no harm” principle requires the class representative and class counsel to spread the gains antici- pated from a rising P and a falling F + C (net of losses anticipated by a falling L) among class members, so none are disadvantaged by class treat- ment. Third, the lower limit of the formula (0 ≤ (PC × LC) − (FC + CC)) is necessary to account for negative-value suits like Kamilewicz, which would normally not be brought as independent litigation. Without the lower limit, a self-interested class representative or class counsel can engage in a signifi- cant amount of self-dealing in such cases. To take Kamilewicz as an example,170 assume that in order to recover the class members’ $2.19 in overcharges, an individual suit would have cost $2,000. With an expected recovery of −$1,997.83, the class members almost surely will not bring the suit; thus, the effective value of their suit is $0. If −$1997.83 rather than $0 is the number that constrains a self-interested class representative or class counsel, then the actions of the class counsel in Kamilewicz could have wors- ened the financial position of the Kamilewiczes by as much as $1,997 without being regarded as inadequate—a far more unjust result than the $89 injustice that occurred. Placing the lower limit of zero on the adequacy 167. Thus, a class member cannot claim the representation was inadequate merely because the class lost the case, while the class member would have won an individual case. 168. See supra note 103 and accompanying text. 169. For an illustration of this statement, see supra notes 104–06 and accompanying text. 170. See supra notes 162–64 and accompanying text. 1178 Texas Law Review [Vol. 87:1137 principle reflects the reality that many large-scale, small-stakes claims are not independently viable. Fourth, the calculation of adequacy must be done for each class member. It is not enough that, overall, the benefits of class treatment exceed the benefits of individual control; that issue has already been determined by other provisions in Rule 23(a).171 Rather, the Due Process Clause guarantees each and every class member adequate representation. This demand can be tricky when class members have a variety of types of claims, types of injuries, and strength of claims, and also have temporal variations. To take the hypothetical example previously discussed, assume a $1 million limited fund with 100 potential claimants, each of whom has an equal and identical $20,000 claim.172 Assume as well that the claims of thirty of the claimants have already matured, and those of the other seventy will mature in several years. Individual actions by the present thirty claimants will deplete the bulk of the fund ($600,000, or $20,000 apiece for each claimant), and leave the remaining seventy claimants to recover a fractional share ($5,714 apiece) of their claims in their individual suits. Self-interested class counsel might wish to bring this case as a (b)(1)(B) class action in order to obtain fees based on recovery of the full $1 million fund. If the class action is successful, class counsel might well contend for a simple (and equitable) distribution of $10,000 to each class member, less costs and fees. If so, it is unlikely that class counsel has adequately represented the thirty present claimants, even though counsel has adequately represented the seventy future claimants and even though, from a social viewpoint, the settlement causes no overall harm (the windfall to the seventy offsets the losses to the thirty, and the per capita cost of implementing the remedy is likely less as well).173 The only circum- stance in which representation of the thirty class members would be adequate is when the costs of individual litigation (FI + CI) are so high that the net in- dividual recovery for the thirty claimants is greater than the net recovery under an equal-share class settlement.174 171. See supra note 78 and accompanying text. 172. See supra text accompanying note 103. 173. This example shows a limitation of the “do no harm” principle: its potential to achieve less efficiency than other possible solutions. I consider this criticism infra in subpart III(B)(3). For a case on which this example is loosely based, see Ortiz v. Fibreboard Corp., 527 U.S. 815, 857–59 (1999), in which the court rejected certification of a (b)(1)(B) class in part because the settlement treated claimants equally even though class members whose claims accrued before 1959 had stronger claims than class members whose claims accrued after 1959. 174. For instance, if we assumed that a present claimant’s fees and costs for recovering $20,000 were $10,000, and the likelihood of recovery were 70%, then the net expected recovery for the thirty claimants in individual litigation would be $4,000 (0.7 × $20,000 − $10,000). On the other hand, if we assumed that the total fees and costs in the class action were $500,000 (or $5,000 per class member), then a settlement that proposed an equal division of the fund ($10,000 apiece) would lead to a net recovery of $5,000 (1 × $10,000 − $5,000) for the thirty present class members, and the class counsel’s representation in agreeing to an equal settlement would be adequate. 2009] Rethinking Adequacy of Representation 1181 class counsel were compelled by a sense of virtue, duty, or altruism to make the class members better off than individual litigation would, the adequacy principle that I propose might be unnecessary.182 But the principle admits of no illusions on that score; it is intended to act as a constraint on the self- interested who wish to achieve the maximum benefit for themselves. As such, the “do no harm” principle imposes a consequentialist limit on their actions—a limit that speaks in the same language that they do. Indeed, both internal arguments183 and external arguments184 support the principle. To begin with internal arguments, the “do no harm” principle aligns the doctrine of adequate representation with the other central require- ments of Rule 23—efficiency and the avoidance of harms—as well as with the structure of the joinder rules, with the modern understanding of proce- dural due process, and with the results in Hansberry and Amchem. As I have described, the foundational requirement for all class actions, which Rule 23(a) seeks in part to ensure, is the efficiency of the class action in rela- tion to individual litigation.185 Likewise, Rule 23(b) specifies certain harms that class actions must avoid.186 Both requirements are consequentialist in orientation; Rules 23(a) and (b)(3) are utilitarian in orientation, while Rules 23(b)(1) and (b)(2) have more solicitude for preventing harms to indi- viduals even if they are not necessarily social harms. As we have also seen, Rule 23 still retains an egoist or self-interested cast in certain regards, espe- cially in its willingness to place the decision to seek class certification in the hands of class representatives and class counsel.187 The “do no harm” princi- ple mediates among these diverging consequentialist tendencies and harmonizes them. It leaves room for some self-interested behavior—it does not require the representative or counsel to seek the best outcome for the class or for each class member. At the same time, it assures that the class action will do at least as much good as individual litigation would do, thus achieving certain social benefits. And it avoids harming absent class members, which is a concern that manifests itself specifically in Rule 23(b)(1)(B) and (b)(2). The “do no harm” formula is not the only prin- 182. The principle is a floor on action, not a ceiling. 183. By “internal” arguments, I refer to arguments derived from legal rules, structures, and precedents. These arguments might also be described as formalist, in the sense that existing legal rules, structures, and precedents constrain or determine the choice of a proper legal outcome. See Frederick Schauer, Formalism, 97 YALE L.J. 509, 511–29 (1988) (defining common uses of the term “formalism” as, variously, the “denial of choice” or the “limitation of choice”); Emily Sherwin, A Defense of Analogical Reasoning in Law, 66 U. CHI. L. REV. 1179, 1189–93 (1999) (stating that arguments by analogy help to make law more stable and less error-prone than does open-ended reasoning). 184. By “external” arguments, I mean arguments grounded in sources other than legal structures, rules, and precedents. 185. See supra note 78 and accompanying text. 186. See supra notes 42–46 and accompanying text. 187. See supra notes 517–61 and accompanying text. 1182 Texas Law Review [Vol. 87:1137 ciple that can mediate among these tendencies,188 but it sits nicely as a compromise among them. For the same reason, the formula is aligned with principles and ideas that drive joinder law more generally. As I have discussed, the default join- der rule, Rule 20, allows plaintiffs to engage in egoist joinder behavior—but it ultimately checks that behavior by considering both the efficiency of a plaintiff’s joinder decision and the fairness of that decision to the defendant.189 In addition, Rule 19(a)(1)(B)(i) requires joinder of absent parties when the failure to join such parties “as a practical matter impair[s] or impede[s] the person’s ability to protect the interest.”190 Thus, at the broader level of joinder, the compromise among egoism, utilitarianism, and altruism that manifests itself in Rule 23 also finds a place. Therefore, the “do no harm” principle, which likewise finds middle ground that partially validates each of these tendencies, fits comfortably within the range of ideas that ani- mate American joinder decisions. A third internal argument is the “do no harm” principle’s consistency with the modern approach to due process. Although Hansberry itself was a bit dodgy about the exact source of the constitutional right to adequate representation—it located the source of the right in both the Due Process Clause and the Full Faith and Credit Clause191—the right is cast purely in due process terms today.192 The Supreme Court provided the modern test for determining the constitutionality of deprivations of individualized adversarial procedure in Mathews v. Eldridge193: departures from adversarial process must have expected benefits that equal or exceed the costs to an individual from the loss of traditional adversarial process.194 Therefore, as a type of 188. For instance, I consider another possible principle—a rule that adequacy requires class representatives and class counsel to improve the positions of class members in relation the second- best joinder alternative. See infra notes 248–55 and accompanying text. 189. See supra notes 20–23, 36–38 and accompanying text; see also FED. R. CIV. P. 19(a)(1)(B)(ii) (requiring joinder when a plaintiff’s joinder decision “leave[s] an existing party subject to a substantial risk of incurring double, multiple, or otherwise inconsistent obligations”). 190. This language parallels that of Rule 23(b)(1)(B), which shows a similar solicitude for protecting the rights of absent plaintiffs. The two protections do not conflict; joinder under Rule 19, however, is subject to the requirements of Rule 23. FED. R. CIV. P. 19(d). 191. See supra note 94. 192. See, e.g., Richards v. Jefferson County, 517 U.S. 793, 805 (1996) (holding that due process prevents the assertion of the preclusive effect of a prior judgment against a nonparty that was not adequately represented in the prior case); Phillips Petroleum Co. v. Shutts, 472 U.S. 797, 812 (1985) (explaining that due process requires “that the named plaintiff at all times adequately represent the interests of the absent class members”). 193. 424 U.S. 319 (1976). 194. Id. at 319, 335. The Supreme Court continues to adhere to this formulation. See Boumediene v. Bush, 128 S. Ct. 2229, 2268 (2008) (citing to the Mathews v. Eldridge test to support the “idea that the necessary scope of habeas review in part depends upon the rigor of any earlier proceedings”); Hamdi v. Rumsfeld, 542 U.S. 507, 509, 529 (2004) (using the Mathews v. Eldridge test to determine the scope of an enemy combatant’s opportunity to contest his detention). On the economic underpinnings of Mathews v. Eldridge, see POSNER, supra note 28, § 21.1, at 593– 94. 2009] Rethinking Adequacy of Representation 1183 departure from traditional adversarial theory,195 class actions satisfy the Mathews v. Eldridge formulation of due process as long as the loss to each class member from not being able to individually litigate her claim does not exceed the expected benefits of class treatment. One of the problems of the conflict-of-interest approach for determining adequacy of representation as a constitutional matter is its poor match-up with the decidedly consequentialist test of Mathews v. Eldridge; under the conflict-of-interest approach, repre- sentation can be inadequate even if the conflict costs less than the gains from class treatment.196 On the other hand, the “do no harm” approach matches up well with the Mathews v. Eldridge formulation for due process: Its insistence that class members not be made worse by class treatment guarantees that, as long as the class action reaps social benefits,197 Mathews v. Eldridge-style due process has been satisfied.198 The final internal argument is the principle’s consistency with the outcomes in both Hansberry and Amchem. In Hansberry, the position staked out by the class representative in the first class action had the potential to harm those class members—including future property owners who would buy the property during the twenty-five year term of the racially restrictive covenant—who, for economic or social reasons, wished to sell to African- Americans. On a “do no harm” theory, answering the question of adequate representation is not as simple as showing the possibility of different litiga- tion positions among class members. Rather, the issue is whether, at the time that the first class action was certified, obtaining a judgment finding that the covenant was valid was expected to put any class member in a worse position than she would have been in had she been allowed to control her own litiga- tion. To answer that question, a court would have needed to compare (1) the expected value—for each present and future homeowner during the remain- ing duration of the covenant—of the home in a white-only neighborhood, after subtracting the cost of bringing the first class-action litigation but add- ing the expected cost of an individual’s possible future suit to enforce the covenant, against (2) the expected value—for each present and future home owner—of the home without any restrictive covenants. If the latter value was greater than the former, then the Hansberry class action violated the “do no harm” principle. On the facts of Hansberry, it appeared that this was the 195. See supra text accompanying note 41. 196. See supra notes 88–97 and accompanying text. 197. The requirements of class certification ensure that these benefits will accrue; class treatment must be more efficient than individual litigation, see supra note 78 and accompanying text, and must avoid harms that individual litigation causes, see supra notes 42–48 and accompanying text. 198. Of course, the Mathews v. Eldridge test could also be satisfied on a lesser showing than the “do no harm” approach, for it allows individual litigants (or class members) to be made worse off as long as the social gains from class treatment are great enough. Thus, the “do no harm” approach is not the constitutional minimum that Mathews v. Eldridge might tolerate. 1186 Texas Law Review [Vol. 87:1137 autonomy.209 For Locke, the natural law imposed necessary limits;210 for Kant, the categorical imperative required each individual to treat every indi- vidual as an end, not a means;211 for Mill, the limiting principle was the maximization of utility;212 for Rawls, it was the veil of ignorance.213 The “do no harm” principle borrows from this tradition, relying most on Mill’s famous observation that one individual’s right to swing a fist ends at another person’s nose,214 on Kant’s categorical imperative, and on the Rawlsian veil of ignorance. In effect, class representatives and class counsel become the agents placed in charge of the claims of class members; at the same time, they are also looking after their distinct and independent interests. The “do no harm” principle insists that in exercising their autonomous right to advance their own litigation interests, they do not act in a way that harms the litigation interests of those for whom they serve as agents. It guarantees that the “society” created by the class representative and class counsel works together for mutual advantage; the representative and counsel cannot repre- sent those that the class action places at a disadvantage in relation to their autonomous right to bring individual litigation. It does not deny the class representative and class counsel the opportunity to achieve better outcomes for themselves than they could achieve through individual litigation. Indeed, the provision of such an opportunity is a necessary incentive to attract repre- sentatives and counsel to file class actions that, as we have seen, beneficially prevent certain types of undesirable harm.215 But in exercising this personal freedom—this right to swing their arms, if you will—representatives and counsel must stop when they threaten to hit the “noses” of class members. By guaranteeing absent class members the same expected (if not better) out- come as individual litigation would have yielded, class representatives and class counsel also respect absent class members as their own ends, rather than simply as a means to achieve the ends of the representative and counsel. Moreover, according to the Rawlsian veil of ignorance, arrangements that benefit some more than others can be regarded as just when all those affected by the arrangement—not knowing whether they will occupy a fa- vored or unfavored position under the arrangement—will nonetheless 209. See William B. Rubenstein, Divided We Litigate: Addressing Disputes Among Group Members and Lawyers in Civil Rights Campaigns, 106 YALE L.J. 1623, 1645, 1645–46 (1997) (“The primary problem with the individualist model is the central downside of liberalism generally: a satisfactory account of its limits.”). 210. See LOCKE, supra note 208, § 96 (finding that once individuals choose to become part of a community, they “submit to the will and determination of the majority,” where the majority “determines, . . . by the law of nature and reason, the power of the whole”). 211. Hans Reiss, Introduction to KANT: POLITICAL WRITINGS, supra note 27, at 1, 18. 212. MILL, supra note 49, at 14. 213. RAWLS, supra note 206, at 11. 214. See supra note 49 and accompanying text. 215. See supra subpart I(B). 2009] Rethinking Adequacy of Representation 1187 consent to it.216 Thus, assume that we place the class representative, the class counsel, and all class members behind a veil of ignorance, so that they do not know which position they will occupy in a class action. We then ask them to choose one of three governing principles for the behavior of the class repre- sentative and class counsel: a principle that permits no conflicts of interest; a principle that tolerates conflicts of interest and further allows the class repre- sentative and class counsel to act in a self-interested way that risks harm to class members; and a principle that tolerates conflicts of interest when no one is made worse by the conflict and some are made better. The position on which all (including the risk-averse) should be able to agree is the last—the “do no harm” principle.217 Like the “no conflicts” position, no one ends up worse off from class treatment; like the “allow all conflicts” position, some are made better off. The “do no harm” principle creates no risk of loss, and promises the possibility of a gain, from class treatment. It is therefore the principle on which all, sitting behind the veil of ignorance, can agree on. What the “do no harm” principle rejects is the “greatest good for the greatest number” solution of utilitarianism. It is easily possible to construct a Kamilewicz-type class-action hypothetical in which most class members end up better off from class as opposed to individual litigation by, say, $5 apiece, while just a few class members end up worse off by $1 apiece. Under a “greatest good” measure of adequacy, the representation is fine. Under the “do no harm” approach, it is not. The “do no harm” approach retains a greater respect for the individualist model of American adjudication, which places a premium on each litigant’s liberty and autonomy.218 It also forces those who gain from class treatment to redistribute some of those gains to others, so they become at least indifferent to the outcome.219 Even if class- 216. RAWLS, supra note 206, at 118–23. 217. Rational wealth maximizers might prefer the second position in some circumstances. Assume that the class action contains one million people plus one class counsel. Assume as well that the expected outcome from individual litigation is a gain of $1, but the expected outcome from the class action is a loss of 50 cents for every person in the class except for the class counsel, who will receive a fee of $2 million. Under this scenario, a rational individual sitting behind the veil of ignorance might support the class action, for there is a one in 1,000,001 chance of obtaining $2 million (thus an expected gain of roughly $2), and a one million in 1,000,001 chance of being made worse $1.50 (roughly, an expected loss of $1.50). When we consider risk-averse individuals, however, the “do no harm” principle becomes the principle on which all, sitting behind a veil of ignorance, would choose. 218. See Rubenstein, supra note 209, at 1644 (“Current procedural and ethical rules encourage group members and attorneys to pursue their own individual paths in filing, pursuing, and constructing test cases.”); supra notes 24–28 and accompanying text. 219. Utilitarian theory recognizes the reality that harm must sometimes be visited on a few in order to achieve the greater good; thus, a switchman can permissibly route a runaway train onto the track with one worker rather than keep it on the track with five workers. See Judith Jarvis Thomson, The Trolley Problem, 94 YALE L.J. 1395, 1395 (1985). These hard ethical dilemmas arise when harm to some is inevitable, and the question is how to keep harm to its bare minimum. But that problem is different from harming someone when there is no reason to do so. In the context of class actions, when the class achieves gains that exceed its costs, asking some members of the class to bear losses, when it is possible to distribute some of the gains achieved by others in 1188 Texas Law Review [Vol. 87:1137 action status takes away the autonomy of class members to act on their own litigation interests, it guarantees them the expected benefits of that autonomy. The distribution of the gains of class treatment among the class mem- bers can also be explained on economic grounds. The “do no harm” approach to adequate representation forces class representatives, class counsel, and class members who benefit from class treatment to internalize the costs that the class action might otherwise impose on other class members. Without a mechanism to internalize these costs, class representatives and class counsel will be too willing to offload the costs of the class action onto some class members, and to set out on courses of con- duct that enrich themselves while neglecting the costs of their behavior. Indeed, “do no harm” adequacy of representation is a specific application of the principle of Pareto improvement, the economic concept that holds an action to be more efficient when at least one person is made better off by the action and no one is made worse off.220 Thus, in comparison to individual litigation, the “do no harm” principle ensures greater efficiency. Admittedly, the “do no harm” principle does not go so far as to insist on Pareto optimality, the economic concept that defines efficiency as the state of affairs in which no further Pareto improvements can be made.221 In neoclassical economics, Pareto optimality is often tempered by the Kaldor– Hicks refinement, which holds that an action is efficient if wealth transfers to achieve Pareto optimality could be made, even when they are not in fact made.222 One reason that the Kaldor–Hicks refinement has been thought necessary is because of the difficulty of adopting large-scale social or economic policies that make no one worse off and the related costliness of identifying those who are made worse off and transferring wealth to them.223 The “do no harm” principle rejects the Kaldor–Hicks refinement, and insists on transfers to those who might be made worse off by class treatment. The reason is that, unlike large-scale social or economic policies whose various permutations can create polycentric distributions of winners and losers,224 class actions are fairly self-contained and organized around the discrete goal of maximizing the value of a class’s legal claims. Therefore, it is relatively the class to those that would be harmed, is impermissible. The “do no harm” principle requires this distribution. 220. See POSNER, supra note 28, § 1.2, at 12. 221. See THOMAS J. MICELI, THE ECONOMIC APPROACH TO LAW 4 (2004) (explaining that an allocation is Pareto efficient or Pareto optimal if there is no other allocation that is Pareto superior to it). 222. This state is also sometimes described as “potential Pareto efficiency.” Id. at 5–6; POSNER, supra note 28, § 1.2, at 13. 223. See MICELI, supra note 221, at 5–6 (describing the problems with using Pareto efficiency to evaluate proposed policy changes and explaining that economists employ Kaldor–Hicks efficiency to address these problems); A. MITCHELL POLINSKY, AN INTRODUCTION TO LAW AND ECONOMICS 8–9 (3d ed. 2003) (illustrating the conflict between efficiency and equity). 224. Cf. Fuller, supra note 24, at 394–95 (discussing polycentrism in the adoption of wage and price legislation whose various permutations create different advantaged and disadvantaged parties). 2009] Rethinking Adequacy of Representation 1191 number and strength, and in terms of the injuries that they have suffered. Class actions are not economic leveling devices intended to enforce a strict equality of outcome on their members. A second form of this criticism is more troubling—that the “do no harm” principle can foster greater inequalities than individual litigation would. The criticism has particular salience in the large-scale, small-stakes context. To take a hypothetical example, assume a negative-value suit in which 10 million customers have lost $2 apiece as a result of an illegal over- charge by a credit card company. The “do no harm” principle would be satisfied if the case settled for $10,000,003, with the class representative re- ceiving $3, the class counsel receiving $10 million, and the remaining class members receiving nothing. Because the remaining class members are no worse off than they would have been with individual control (their claims were, as a practical matter, worthless), and because the class representative and class counsel are better off, the representation in this case is, under the “do no harm” principle, adequate. My initial reaction to this evident difficulty was to switch the “do no harm” principle to the weak Pareto-optimal condition of “do some good”: in other words, to require the class representative and class counsel to distribute the overall gains from class treatment in such a way that every class member is better off as a result of class treatment. But that change does little to solve the problem; for the “do some good” condition is satisfied if class counsel agrees to take $9,990,000.01 in fees and costs, gives $3 to the class representative, and gives one cent to the remaining 9,999,999 class members. Indeed, on these facts, the “do some good” solution is less efficient than the “do no harm” principle because of the costs associated with distributing the penny to each class member. The “do no harm” principle does not bar the hypothetical $10,000,003 settlement. “Do no harm” adequacy does not require equality of treatment among class members, and it can create greater inequities between the class representative and class members than those that predated the lawsuit (when the class representative and class members had equal $2 claims that, as a practical matter, were equally worthless).232 The principle is intended to check self-interest and collusion that make class members worse off. It is a minimal principle, rather than a maximal version of adequacy that requires the gains from class treatment to be distributed among claimants in propor- tion to the strength and extent of their claims—in effect, it is an “equal gains for equal claims” principle. 232. This example is an application of one common criticism of Pareto improvements and optimality—that they can foster greater social inequalities even as they improve efficiency. See POSNER, supra note 28, § 1.2, at 13–14 (discussing how inequalities in the distribution of wealth affect the consumption and production decisions of parties to an economic transaction); AMARTYA KUMAR SEN, COLLECTIVE CHOICE AND SOCIAL WELFARE 22 (1970) (“An economy can be optimal in this sense even when some people are rolling in luxury and others are near starvation as long as the starvers cannot be made better off without cutting into the pleasures of the rich.”). 1192 Texas Law Review [Vol. 87:1137 Despite its surface appeal, one problem with an “equal gains for equal claims” principle is the difficulty of applying it in many class actions. When class members present an array of temporally and geographically dispersed claims and injuries, an “equal gains for equal claims” principle requires the collection of a great deal of information, as well as difficult judgments about which claims, claimants, and distributions are in fact “equal.”233 Such a principle would therefore eat deeply into—and perhaps surpass—the gains in reduced litigation costs and avoidance of harms that are the raison d’être of Rule 23. The implementation of the principle will also make the tasks of serving as class representative and class counsel more onerous and less ap- pealing and will make collateral attacks on the adequacy of representation, which undermine the finality of class litigation, more likely.234 In short, an “equal gains for equal claims” principle will likely lead to fewer class actions than a “do no harm” principle, thereby thwarting the benefits that class ac- tions provide. That fact alone is not a reason to reject such a principle, but there is another reason as well. Real-world dynamics temper the inequities inherent in the “do no harm” principle and push the parties toward the “equal gains for equal claims” principle without incurring the added costs of adopting the latter principle across the board. To explain, begin by assuming that the class action reaches trial, and the factfinder (jury or judge) finds the defendant liable. In considering the appropriate remedy, the self-interested class representative will want to pitch the remedy (whether injunctive or monetary) as favorably to herself as possible. The self-interested class counsel is likely to push back against the preference, at least in cases seeking monetary relief; for the size of the attorneys fee grows as counsel enhances the value of the claims of the entire class. Moreover, to use to the credit-card hypothetical again, it is inconceivable that, after finding the defendant liable, 233. The “do no harm” principle also requires the collection of some of this same information. See infra note 235 and accompanying text. But it does not require a judge to classify class members into equal groups for the purposes of determining adequacy. 234. The scope of a class member’s right to attack a judgment collaterally has divided the courts. Compare Epstein v. MCA, Inc., 179 F.3d 641, 648 (9th Cir. 1999) (restricting the right of collateral attack), with Stephenson v. Dow Chem. Co., 273 F.3d 249, 257 (2d Cir. 2001), aff’d in part by an equally divided Court and vacated in part, 539 U.S. 111 (2003) (permitting collateral attack when class members were inadequately represented in the class action), and Wal-Mart Stores, Inc. v. Visa U.S.A., Inc., 396 F.3d 96, 110 (2d Cir. 2005) (distinguishing Stephenson and restricting the right of collateral attack). The literature is equally divided. See, e.g., Geoffrey C. Hazard, Jr. et al., An Historical Analysis of the Binding Effect of Class Suits, 146 U. PA. L. REV. 1849, 1946–47 (1998) (tracing the historical progress of class suits, with particular focus on the consistent failure to adequately resolve the issue of class suits’ preclusive effect on absentees); Henry Paul Monaghan, Antisuit Injunctions and Preclusion Against Absent Nonresident Class Members, 98 COLUM. L. REV. 1148, 1202 (1998) (“[A]bsent class members should be allowed to make their collateral due process challenges in a forum of their own choosing.”). For attempts to find a middle ground, see Marcel Kahan & Linda Silberman, The Inadequate Search for “Adequacy” in Class Actions: A Critique of Epstein v. MCA, Inc., 73 N.Y.U. L. REV. 765, 789 (1998); and William B. Rubenstein, Finality in Class Litigation: Lessons from Habeas, 82 N.Y.U. L. REV. 790, 867 (2007). 2009] Rethinking Adequacy of Representation 1193 a jury or judge will order a $3 recovery for the class representative and a one-cent recovery for the remainder of the class. It is important to recall that, in order to satisfy a “do no harm” principle, the class representative and class counsel must have presented to the trial judge evidence of the nature, strength, and extent of the claims of class members. Armed with the knowl- edge that the claims of every class member are equal in strength and size, the trial judge, and the judges on appeal, will not permit such an inequitable dis- tribution to survive post-trial review of the judgment.235 Therefore, as a realistic matter, the only circumstance in which the inequity created by a “do no harm” principle might be realized is the settlement context. Even with settlement, however, real-world checks pre- vent gross inequalities in treatment. A class settlement requires judicial approval, and judicial approval can be given only if the court finds the set- tlement to be “fair, reasonable, and adequate.”236 As the last paragraph noted, the judge will possess information regarding the nature of each class member’s claim. Given the judge’s knowledge, a settlement containing gross disparities in the distribution of settlement proceeds, in which the class representative and class counsel capture virtually all of the remedy that ex- ceeds the remedy that class members would receive in individual litigation, is unlikely to secure judicial approval.237 Moreover, in positive-value monetary suits, a gross distributional disparity is likely to cause some class members to opt out of the settlement if they have the opportunity to do so.238 For these reasons alone, a self-interested class counsel, who is ultimately indifferent to the distribution of the remedy, would be unlikely to acquiesce to seriously inequitable provisions that threaten either the certification of the class (and hence the ability to obtain a fee) or the size of the fee. A final dynamic that minimizes the possibility of inequitable treatment under the “do no harm” approach is the likelihood that a settlement would establish a claims- 235. Admittedly, self-interested class representatives knowing this fact have an incentive to skew the evidence presented in the initial “do no harm” showing in order to make their cases appear stronger, and therefore deserving of a greater eventual remedy. But there are clear limits on the extent of the skewing that can occur; if class representatives make their claims appear too disparate from that of the remainder of the class, they will fail the commonality and typicality requirements of Rules 23(a)(2) and (a)(3). See Gen. Tel. Co. of the Sw. v. Falcon, 457 U.S. 147, 156–60 (1982) (reversing the class-certification order because the district court committed error by presuming that the class representative’s claim was typical of other racial-discrimination claims against the defendant corporation). 236. FED. R. CIV. P. 23(e)(2). 237. But, unfortunately, it is not impossible, as the Kamilewicz story proves. See supra notes 162–64 and accompanying text. 238. In a (b)(3) settlement class action, class members have a right to opt out. FED. R. CIV. P. 23(c)(2)(B)(v). In a litigation class action that was previously certified, a court has the power to provide a second opt-out right at the time of the settlement. FED. R. CIV. P. 23(e)(4); see also AGGREGATE LITIGATION, supra note 6, § 3.11 (recommending that a second opt-out right ordinarily be given). But see David Rosenberg, Mandatory-Litigation Class Action: The Only Option for Mass Tort Cases, 115 HARV. L. REV. 831, 840–66 (2002) (arguing that plaintiffs should not be permitted to opt out of class actions). 1196 Texas Law Review [Vol. 87:1137 expected recovery from individual litigation, this principle is not more efficient. A different principle that appears to be a Pareto improvement is this: the class representative and class counsel are adequate representatives of a class member if and only if they act in such a way that they obtain for that class member a net expected recovery that is at least as good as the recovery that the class member could obtain from the best available method other than the present class action for resolving the case. In effect, this too is a “do no harm” principle; but the baseline against which harm is measured is not indi- vidual litigation, but rather all other possible methods by which the case might be resolved—including multidistrict litigation, consolidation, and even competing class actions filed elsewhere.248 First, although this principle sounds like a Pareto improvement over the “do no harm” principle, in some circumstances it is not; the principle can lead to a finding of inadequacy (and hence no class treatment) even though class treatment would be the most efficient result.249 Second, the dynamics of class-action litigation, which I have previously described, are sufficient in many cases to push a judgment or settlement in a direction that is Pareto su- perior to the position the class members would occupy under the “do no harm” principle.250 Third, the reason that I chose individual litigation as the baseline against which to measure Pareto improvements—as opposed to an- other baseline such as the best alternative other than the present class 248. I am grateful to Sarah Lawsky for the precision with which she stated this principle. 249. Consider the hypothetical situation in which the defendant holds a $1 million fund, against which 100 claimants have identical $20,000 claims. Assume as well that eighty of the claimants have present claims (the other twenty claims having not yet matured), that the claims involve a federal question, and all eighty claimants file in federal court. Finally, assume that there are three possible ways to adjudicate the dispute: (1) Individual actions, in which each plaintiff has a 40% probability of success and each case involves $7,000 in fees and costs (incurred by claimants only if the case is successful); (2) a multidistrict (MDL) proceeding, in which the eighty present claimants are consolidated for pretrial purposes, the probability of success on each claim rises to 75%, and the per-plaintiff expected fees and costs (incurred only if the case is successful) are $4,200; and (3) a mandatory class action of all 100 claimants, in which the probability of success is 75% and the per- plaintiff expected costs and fees (incurred only if the case is successful) are $3,000. In this case, the net expected recovery in individual litigation is $5,200 ($20,000 gross recovery less $7,000 in expenses, discounted by the 0.4 probability). The net expected recovery per claimant in the mass- joinder action is $6,225 ($12,500 gross recovery less $4,200 in expenses, discounted by the 0.75 probability). The net expected recovery per claimant in the class action is $5,250 ($10,000 gross recovery less $3,000 in expenses, discounted by the 0.75 probability). On these facts, the class action “does harm” in relation to the mass-joinder option; the class representative and class counsel therefore cannot adequately represent the class under the “do no harm compared to other methods” principle. The problem, however, is that the MDL approach is inefficient in relation to the class- action approach. First, it creates a risk of harm either to the absent twenty claimants (if the MDL approach renders the defendant insolvent) or to the defendant (if the defendant is called upon to make good on the twenty remaining $20,000 claims). Second, the cost of obtaining judgment in the MDL situation is $336,000 ($4,200 per plaintiff multiplied by eighty plaintiffs), a less efficient result than the $300,000 ($3,000 per class member multiplied by 100 class members) cost under the class-action approach. 250. See supra notes 235–39 and accompanying text. 2009] Rethinking Adequacy of Representation 1197 action—is the orientation of our joinder system toward individual control as the default litigation position.251 The same orientation exists in the Mathews v. Eldridge formulation of due process: The constitutionality of any departure from adversarial process is measured against the outcome achieved in indi- vidual litigation, not against other second-best alternatives.252 It is perhaps a trite, but nonetheless accurate, observation that “adequate” representation is not perfect representation, and we have always measured adequacy in rela- tion to the opportunities that a litigant might enjoy in a separate lawsuit. Fourth, the majority of class actions are Rule 23(b)(3) actions, and Rule 23(b)(3) requires as a condition of certification that class treatment be “superior to other available methods for fairly and efficiently adjudicating the controversy.”253 Nothing in this language suggests that the (b)(3) class action must make the position of each class member better in relation to second-best alternatives.254 Conversely, one factor that is relevant in determining superiority is Rule 23(b)(3)(A), which provides that a court can consider class members’ “interests in individually controlling the prosecution . . . of separate actions.”255 Thus, the idea that a class action must not worsen the interests of class members in relation to individual liti- gation is built into the superiority analysis in a way that the idea of not worsening the position of class members in relation to second-best alternatives is not. The second, maximalist criticism of the “do no harm” principle is to attack its inefficiency in relation to another possible adequacy principle: a principle that the constitutional duty of class representatives and class coun- sel is to maximize the value of the claim of the class and to ignore the harm to the positions of individual class members completely. This principle re- quires representatives and counsel to pursue the Pareto-optimal outcome (as reconfigured by the Kaldor–Hicks refinement). Admittedly, the “do no harm” principle does not require the class representative or class counsel to engage in actions that maximize the value of the claims of the class as a whole. From a utilitarian perspective, this fact is problematic. The logical conclusion of utilitarianism is to maximize the good. In theory, one action achieves that maximum and can therefore be regarded as ethically superior to any other action. The “do no harm” principle stops short of requiring that 251. See supra subpart I(A). 252. See supra note 194 and accompanying text. 253. FED. R. CIV. P. 23(b)(3). 254. In effect, the superiority analysis of Rule 23(b)(3) operates like a principle of Pareto superiority, but with the Kaldor–Hicks refinement: It requires that a class action be a Pareto improvement over alternative methods of recovery, but it does not require that the benefits of that improvement be spread across all members of the class so that no class member is less worse off than the member would be under the second-best alternative. If it were otherwise, the superiority requirement would defeat some class actions that were more efficient than the second-best alternative. 255. FED. R. CIV. P. 23(b)(3)(A). 1198 Texas Law Review [Vol. 87:1137 action and therefore cannot be regarded as the ethically superior rule. Nor does the principle lead to the greatest economic efficiency. But this criticism can be leveled against most of the law, including the law of joinder and the law of class actions. Individual control of litigation under Rule 20 does not necessarily lead to maximum social utility in every case. More to the point, class actions under Rule 23(b)(1) and (b)(2), which contain no superiority requirement, will not achieve the maximum social utility in every case; a measure of adequacy requiring wealth maximization would prevent such class actions from being certified.256 Indeed, if pursuing the most efficient outcome were the constitutional baseline for measuring adequacy, then all class representation, save for the single wealth- maximizing strategy, must be deemed inadequate. Mathews v. Eldridge’s formulation of due process is not so stringent; it does not require wealth- maximizing procedure, but rather holds that any procedure that is a Pareto improvement over adversarial process is constitutional. Moreover, by forc- ing the class representative and class counsel to adopt a wealth-maximizing strategy, which might not allow them to capture for themselves some gains associated with class treatment, one incentive for representatives and counsel to seek certification is reduced, possibly reducing the effective level of deterrence that class actions can achieve. As I have described, the efficiency requirements of Rules 23(a)(1)– (3),257 the superiority requirement of Rule 23(b)(3),258 and the dynamics of class-action litigation259 already do the heavy lifting of maximizing utility. The “do no harm” principle also works to promote efficiency to a degree: It prevents backsliding toward less efficient solutions due to the actions of self- interested class representatives and class counsel who seek to improve their own positions by causing greater harm to others. The point of the principle is not to maximize wealth directly, but to check selfish behavior by those who are not committed to maximizing the class’s wealth or otherwise caring for their interests. The “do no harm” principle does not prevent class represen- tatives and class counsel who wish to do so from maximizing class wealth— as long as the consequence of doing so is not visited on a few whose posi- tions are made worse. The principle respects the substance, if not the form, of individual control, on which American society sets a high value. Finally, by not inflexibly requiring the pursuit of the single path of wealth maximization, the “do no harm” principle allows class representatives and class counsel some latitude of action without raising the fear that every 256. The harms that Rules 23(b)(1) and (b)(2) protect against are not necessarily harms in an economic sense. For instance, in the limited-fund context of Rule 23(b)(1)(B), it is a matter of indifference from the viewpoint of deterrence and efficiency whether the fund is given only to the early claimants or instead is spread equitably among all claimants. A no-class-action alternative leads to the former result, and the 23(b)(1)(B) alternative leads to the latter. 257. See supra note 78 and accompanying text. 258. See supra notes 147–49 and accompanying text. 259. See supra notes 235–39 and accompanying text. 2009] Rethinking Adequacy of Representation 1201 of Rule 23(a) on its head: One of the requirements of Rule 23(a) designed to ensure adequacy of representation is that the class representative must be a member of the class,267 and the guardian is not. In any event, reconceiving the roles of class counsel, judge, or guardian ad litem assumes that the designated overseer can decipher the interests of the class and employ a metric to determine when those interests are ade- quately represented. Without a metric, oversight seems futile, or at least will result in oversight among class actions as widely disparate as the measure of a chancellor’s foot. With a metric—such as the “do no harm” principle— there is no need for complex institutional arrangements; the judge (aided, perhaps by a traditional adjunct such as a magistrate judge or master) can determine adequacy directly. c. Continual Opt-Out Rights.—A third solution is to allow class members a continual right to opt out of a class action. Opt-out rights are rarely invoked,268 so there is still the realistic concern for self-interested be- havior in such a setting. Opt-out rights are expensive, at least if we are serious about providing notice to class members at every stage at which class members might wish to opt out. Permitting opt-outs also reduces the preclu- sive effect of a class action, thus spawning satellite litigation. A general opt- out right would significantly restructure the present Rule 23, which allows (in almost all cases) an opt-out right only in (b)(3) class actions.269 Finally, once class members have opted out, the very harms that class actions are in- tended to prevent (to the class representative, the defendant, the class members, or society) can occur. d. Opt-In Requirements.—A fourth solution is to require class members to opt in: Class actions consist only of those members who affirmatively choose to opt into the class.270 At a theoretical level, this solu- tion avoids all of the concerns for self-interested behavior of class actions; an opt-in model is an exercise of a litigant’s autonomy, akin to joinder under Rule 20, and thus any failings of the class representative and class counsel to 267. See FED. R. CIV. P. 23(a) (“One or more members of a class may sue . . . on behalf of all . . . .”). 268. See WILLGING ET AL., supra note 51, at 52, 52–53 (“[T]he median percentage of members who opted out was either 0.1% or 0.2% of the total membership of the class and 75% of the opt-out cases had 1.2% or fewer class members opt out.”). 269. See supra note 54 and accompanying text. 270. This is a return to the “spurious” class action idea that predated the present Rule 23(b)(3), which instead set the default as an opt-out rule. Owen Fiss and John Bronsteen are modern proponents of an opt-in solution. See John Bronsteen, Class Action Settlements: An Opt-In Proposal, 2005 U. ILL. L. REV. 903, 906 (proposing that class-action settlements, which are currently binding on any class member who does not expressly opt-out, should instead be binding only on those who explicitly opt into the settlement); Bronsteen & Fiss, supra note 55, at 1453 (suggesting a change in class-action rules that would require class members to consent to the settlement before it would become binding on them). 1202 Texas Law Review [Vol. 87:1137 protect class members’ interests can be treated as a waiver by the class member of individual rights. At a practical level, however, an opt-in strategy gets even less preclusive effect than an opt-out strategy, and increases the likelihood and scope of the potential harm from multiple lawsuits. It can re- quire an expensive notice campaign. Moreover, the concern for self- interested behavior by class representatives and class counsel does not disap- pear in an opt-in class action; the doctrine of waiver simply ignores the problem.271 IV. Conclusion Colleagues who have heard my proposal have sometimes asked whether the “do no harm” measure of adequate representation will lead to more or fewer class actions. To some, the principle seems likely to lead to certifica- tion of fewer class actions, with the undesirable result that more widespread wrongdoing by large entities will go undeterred and more victims will remain uncompensated. To others, the principle seems likely to lead to more class certifications, both because it replaces an unworkable identity-of-interests approach with something more flexible, and because it might have the spill- over effect of making courts willing to rein in some of their recent hostility to (b)(3) class actions. It seems out of place in an Article that has stressed a consequentialist approach to the adequacy problem to answer this question by saying, “I don’t know, and I don’t care. The principle is right regardless of its consequences.” So my first inclination—to say that adoption of the “do no harm” principle ensures that those class actions that are certified will have the beneficial consequence of increasing social welfare—dodges the question. My second inclination is to believe that the principle is unlikely to have a significant effect on the number of class actions filed. Even though class actions inherently place people with conflicting interests into a single class,272 class actions are frequently certified. That fact suggests that courts are already finding (perhaps subconscious) ways to reconcile the adequacy requirement with the presence of conflicts of interest. Indeed, the “do no harm” principle might well be the rule of thumb already operating in some courts. Elsewhere, the operative rule of thumb is probably more restrictive than the “do no harm” rule in some courts and less restrictive in others. In global terms, therefore, the adoption of a “do no harm” principle is probably a wash. One of the principal advantages of replacing these rules of thumb with the “do no harm” principle is the transparency of an above-the-table rule that can replace a bevy of uncertain and unknown rules by which individual judges likely measure adequacy of representation today. The “do no harm” 271. See supra note 151 and accompanying text. 272. See supra Part II. 2009] Rethinking Adequacy of Representation 1203 principle is the type of simple, fair rule that makes all of Rule 23 into a pull harness to achieve the efficient handling of mass litigation and the simulta- neous reduction of harm to class members, to defendants, and to society. If we employ it, we will come closer to having the right number of class actions, whether that number is more or less than the number of class actions today.