Our discussion of the background and procedural history of this litigation appears in Judge Winter's lead opinion, 818 F.2d 145. This portion of the Agent Orange appeal concerns the district court's approval of a fee sharing agreement entered into by the nine-member Plaintiffs' Management Committee ("PMC") in December of 1983. Under the agreement, each PMC member who had advanced funds to the class for general litigation expenses was to receive a threefold return on his investment prior to the distribution of other fees awarded to individual PMC members by the district court. In result, the agreement dramatically increased the fees awarded to those PMC members who had advanced funds to the class for expenses, and concurrently decreased the fees awarded to non-investing PMC members, who only performed legal services for the class.

David Dean, lead trial counsel for the plaintiff class and a non-investing member of the PMC, challenges the validity of the agreement, to which he was a signatory, contending that it violates DR 5-103 and DR 2-107(A) of the ABA Code of Professional Responsibility ("ABA Code"). The ABA Code provisions prohibit an attorney from acquiring a proprietary interest in an action in which he is involved and from dividing a fee with an attorney who is not a member of his firm, unless such division is made pursuant to client consent and is based upon services performed and responsibility assumed. In addition, Dean asserts that such an agreement, which premises the size of a fee on the amount advanced for expenses rather than on services rendered, violates the standards and principles developed in this circuit for the award of attorneys' fees in equitable fund class actions and inevitably places class counsel in a position at odds with the interests of the class itself.

Although not informed of the existence of the fee sharing agreement until September of 1984, four months after the parties reached a settlement, the district court approved the agreement, holding that "there is no reason to believe that the existence of the PMC's fee-sharing agreement had any appreciable untoward effect on the decision to settle." In re "Agent Orange" Product Liability Litigation, 611 F. Supp. 1452, 1461 (E.D.N.Y.1985) ("Agent Orange I "). In essence, the court determined that the substantial financial demands placed upon counsel in complex multiparty litigation require flexibility in reviewing internal fee sharing agreements so as not to discourage future representation of large plaintiff classes. At the same time, however, the district judge ruled that, in all future cases, counsel must notify the court of any fee sharing agreement at the time of its inception. In this way, according to the district judge, "the court at the outset can determine whether to permit the fee allocation agreement to stand before any attorney invests substantial time and funds." Id. at 1463.

Because we find that the agreement before us violates established principles governing awards of attorneys' fees in equitable fund class actions and creates a strong possibility of a conflict of interest between class counsel and those they were charged to represent, we reverse the district court's approval of the agreement. Accordingly, the fees originally allocated by the district court, based on the reasonable value of services actually rendered, will be distributed to the members of the PMC.

In September of 1983 Yannacone and Associates withdrew as attorneys for the class, claiming financial and management hardships. The district court then approved appointment of the PMC as new class counsel. The PMC was comprised of three members--attorneys Stephen Schlegel, Benton Musslewhite and Thomas Henderson. In re "Agent Orange" Product Liability Litigation, 571 F. Supp. 481 (E.D.N.Y.1983). In later months the district court approved the expansion of the PMC to encompass six additional members, including appellant David Dean. Dean, a member of the original panel of class counsel, had been closely involved with the Agent Orange litigation since its inception in 1979. In October of 1983 the district court appointed him to be the attorney responsible for leading the preparation and potential trial of plaintiffs' case.

In December of 1983, as a means of raising the capital necessary for the maintenance and continuation of the lawsuit, the nine PMC members entered into a written fee sharing agreement whereby six of the members each promised to advance the class $200,000 for general litigation expenses. The agreement provided that the investing members would be reimbursed threefold from the pool of attorneys' fees awarded to PMC members upon successful completion of the action. The fees remaining in the pool after the investment pay-outs would be distributed pursuant to a fifty-thirty-twenty percent formula: fifty percent of the remainder would be distributed equally among the nine PMC members, thirty percent would be distributed according to the number of hours each member expended in the case, and twenty percent would be distributed in accordance with certain quality and risk factors relating to each PMC member's work in the action, as determined by a majority vote of the PMC. All PMC members, including Dean, signed the agreement. The district court, however, was not notified of its existence.

The action was settled in May of 1984 and the district court, by order dated June 11, 1984, notified counsel that petitions for attorneys' fees were to be submitted to the court no later than August 31, 1984. A hearing on the issue of fees was scheduled for late September. In ordering the hearing, the district court waived application of Rule 5 of the Local Rules of the Eastern District of New York requiring notice to the class of all fee applications and fee sharing agreements prior to the hearing on such fee petitions. The court gave as its reasons "the need for continued intensive work by the attorneys until the close of the fairness hearings and ... the complexity of the fee applications." Notice of Proposed Settlement of Class Action, reprinted in In re "Agent Orange" Product Liability Litigation, 597 F. Supp. 740, 867 (E.D.N.Y.1984). When the court waived application of the local rule, it was unaware of the PMC fee sharing agreement.

It was not until the PMC submitted its joint fee petition that the court finally learned of the agreement. At the September hearing on the fee petitions, the district judge expressed doubts as to the agreement's propriety and requested further briefing on the issue. Faced with the reservations expressed by the district judge, the PMC members modified their agreement in December of 1984. The revised agreement, and the one now before us, provided that five of the six investing members of the PMC each would advance an additional $50,000 for general litigation expenses, bringing their total investments to $250,000 each. In return for these advances, as well as for the $200,000 advanced by the sixth investing member, the new agreement provided for the same threefold return as did the original agreement. The fifty-thirty-twenty percent formula for the distribution of the remaining portion of the fees, however, was eliminated. In its place, the revised agreement called for the remainder to be distributed pro rata to each PMC member "in the proportion the individual's and/or firm's fee award bears to the total fees awarded."1 Agent Orange I, 611 F. Supp. at 1454.

On January 7, 1985, the district court issued a Memorandum and Order awarding over $10 million in fees and expenses to the various counsel whose work had benefitted the class, applying the principles of fee distribution in equitable fund actions set forth in City of Detroit v. Grinnell Corp., 495 F.2d 448 (2d Cir.1974) ("Grinnell I ") and City of Detroit v. Grinnell Corp., 560 F.2d 1093 (2d Cir.1977) ("Grinnell II "). In re "Agent Orange" Product Liability Litigation, 611 F. Supp. 1296 (E.D.N.Y.1985) ("Agent Orange II "). As later amended and supplemented, the district court's decision awarded over $4.7 million in fees to the nine members of the PMC on an individually apportioned basis. David Dean, due to his lengthy involvement in the class action and the exceptional quality of his work, was awarded $1,424,283.75, or over thirty percent of all fees awarded to the PMC. Each of the six investing members of the PMC was awarded a much lower percentage of the entire PMC fee award, with one investor being awarded only $41,886. The highest award to an investor was $515,163.

Once the fee sharing agreement was applied to these awards, however, the amount of fees each PMC member was to receive changed dramatically. In Dean's case, application of the agreement reduced his award to $542,310, a reduction of $881,973. In contrast, Newton Schwartz, an investing member of the PMC to whom the district court awarded $41,886, was now to receive $513,026, equivalent to an hourly rate of $1,224.81. The awards to all other investing members were similarly enhanced and, in turn, the awards to the two other non-investing members were diminished, resulting in a distortion of the district court's individual PMC member fee awards. The total of all fees awarded by the court to the members of the PMC, of course, remained unchanged.2

In May of 1985, Dean moved in the district court to overturn the fee sharing agreement, claiming that it violated professional ethics and did not protect the rights of the class. In a Memorandum and Order issued June 27, 1985, the court denied Dean's motion and upheld the agreement, albeit with some reluctance. The court found, as a factual matter, that no conflict of interest had arisen in the litigation from the fee sharing agreement and, consequently, that the interests of the class in obtaining a fair and reasonable settlement had not been impinged. Agent Orange I, 611 F. Supp. at 1461. Initially, the court recognized its obligation to review the agreement in its capacity as protector of the rights of the plaintiff class. It then went on to examine the propriety of the agreement under DR 2-107(A) and DR 5-103 of the ABA Code and the practical effect of the agreement on the PMC's representation of the class.

As to DR 2-107(A), which prohibits an attorney from splitting his fee with another attorney not of the same firm unless he has the consent of his client and the "division is made in proportion to the service performed and responsibility assumed by each," the court determined that the PMC should be viewed as an ad hoc law firm "formed for the purpose of prosecuting the Agent Orange multidistrict litigation," Agent Orange I, 611 F. Supp. at 1458. The court reasoned that the business realities of the litigation required the PMC to be able to perform those functions ordinarily performed by actual law firms, such as splitting fees among its members. The district court also noted that the Model Rules of Professional Conduct ("Model Rules") adopted by the ABA in 1983, although not adopted in New York, reflect "an increased recognition" of these business realities by permitting fee sharing agreements based upon services rendered or upon written acceptance of joint responsibility by the attorneys if the client is advised of the participation and does not object and the total fee is reasonable. Model Rule 1.5(e). Recognizing the practical problem of client consent in class actions, however, the district court concluded that its duty to protect the rights of the class ordinarily could not be performed unless the attorneys involved notified the court of the existence of such an agreement "as soon as possible," Agent Orange I, 611 F. Supp. at 1459.

As to DR 5-103, which prohibits an attorney from acquiring a proprietary interest in an action in which he is involved, the court found that the investing members acquired no independent interest in the action because the financial return from any initial advance for expenses was to be paid from the fees otherwise awarded to the PMC members, and thus would not affect the class fund. While the court did recognize that a conflict of interest could arise from such an agreement, it cautioned that complex class actions require a more sophisticated analysis of ethical codes than ordinary two-party cases in order not to "unnecessarily discourage counsel from undertaking the expensive and protracted complex multiparty litigation often needed to vindicate the rights of a class." Id. at 1460. Accordingly, the district court held that a case-by-case analysis of such fee sharing agreements to identify potential conflicts of interest should be adopted.

The court conceded that an agreement of the sort before it conceivably could create an interest on the part of the investors to settle early, regardless of the benefit to, or interest of, the class. This is because an attorney whose fee is based upon the amount of funds advanced for expenses in an action will receive the same fees "whether the case is settled today or five years from now." Id. The court reasoned, however, that any possible interest to settle early would have been offset by the theoretical incentive to extend such litigation created by the lodestar formula and concluded that, as a factual matter, no conflict had arisen here.

The court then set forth five additional, though nondispositive, reasons for approving the agreement. First, the returns on the investments did not affect the class fund, since they were paid from the fee awards of PMC members. Second, the court recognized that the "business" of law will at times require creative, yet ethical, methods for economical and efficient operation. Third, without the funds advanced by the PMC members, it was possible that the litigation would have collapsed and neither the attorneys nor the class would receive any payments. Fourth, the court noted that the PMC members could have earned substantial returns, though not quite threefold, on these same funds if they had undertaken more traditional investments. Fifth, if the PMC members had received the amount of fees requested in their joint petition, nearly thirty million dollars, the extent of the distortion of the fees by the investment agreement would have been insubstantial.

In sum, the district court determined that the practical needs of this form of litigation required an inventive method of fund raising in order to guarantee effective representation of class rights. At the same time, however, it labeled as "troubling" the PMC's failure to inform the court of the existence of the agreement until months after a settlement had been reached. Id. at 1462. In light of class counsel's fiduciary obligations to the class and the court's role as guardian of class rights in relation to settlement review, the district court found that both the class and the court had a right to be notified of the existence of such an agreement. To this end, the court proclaimed that in all future cases, class counsel would be obligated to make the existence of a fee sharing agreement known to the court at the time of its formation.

Dean's appeal presents an issue of first impression: whether an undisclosed, consensual fee sharing agreement, which adjusts the distribution of court awarded fees in amounts which represent a multiple of the sums advanced by attorneys to a class for litigation expenses, satisfies the principles governing fee awards and is consistent with the interests of the class.

At the outset, we note that the fees in this case were awarded pursuant to the equitable fund doctrine, first set forth in Trustees v. Greenough, 105 U.S. (15 Otto) 527, 26 L. Ed. 1157 (1882), and Central Railroad & Banking Co. v. Pettus, 113 U.S. 116, 5 S. Ct. 387, 28 L. Ed. 915 (1885). The underlying rationale for the doctrine is the belief that an attorney who creates a fund for the benefit of a class should receive reasonable compensation from the fund for his efforts. Central Railroad, 113 U.S. at 125. Because the calculation of fees necessarily will affect the funds available to the class, this circuit has adopted a lodestar formula for fee computation. Grinnell II, 560 F.2d at 1099; Grinnell I, 495 F.2d at 471. The lodestar seeks to protect the interests of the class by tying fees to the "actual effort made by the attorney to benefit the class." Grinnell II, 560 F.2d at 1099. Accordingly, fees are calculated by taking the number of hours reasonably billed and multiplying that figure by an hourly rate "normally charged for similar work by attorneys of like skill in the area." Id. at 1098. Once calculated, the court may, in its discretion, increase or decrease this figure by examining such factors as the quality of counsel's work, the risk of the litigation and the complexity of the issues. Id. Discretion to adjust the lodestar figure upward because of superior quality, however, is limited to exceptional situations and must be supported by "specific evidence" and "detailed findings" by the district court. Pennsylvania v. Delaware Valley Citizens' Council for Clear Air, --- U.S. ----, 106 S. Ct. 3088, 3098, 92 L. Ed. 2d 439 (1986). Adherence to these principles is essential not only to avoid awarding windfall fees to counsel, but also to "avoid every appearance of having done so," Grinnell I, 495 F.2d at 469.

Of equal importance to our analysis is Fed.R.Civ.P. 23(e), which requires court approval of any settlement of a class action suit and squarely places the court in the role of protector of the rights of the class when such a settlement is reached and attorneys' fees are awarded. Grinnell II, 560 F.2d at 1099. In fulfilling this role, courts should look to the various codes of ethics as guidelines for judging the conduct of counsel. Agent Orange I, 611 F. Supp. at 1456. In addition, where only retrospective review of counsel's conduct is available, courts should not be limited to an examination of the actual effects of such conduct on the litigation, but rather, as the ABA Code and Grinnell I imply, the appearance and potential effect of the conduct should be reviewed as well. See Grinnell I, 495 F.2d at 469; ABA Code of Professional Responsibility Canon 9 (1975).

The ultimate inquiry, therefore, in examining fee agreements and setting fee awards under the equitable fund doctrine and Fed.R.Civ.P. 23(e), is the effect an agreement could have on the rights of a class. Because we find that the agreement here conflicts substantially with the principles of reasonable compensation in common fund actions set forth in Grinnell I and Grinnell II, and that it places class counsel in a potentially conflicting position in relation to the interests of the class, we reverse.

Initially, it is beyond doubt that the agreement, by tying the fee to be received by individual PMC members to the amounts each advanced for expenses, completely distorted the lodestar approach to fee awards. In setting fees here, the district judge meticulously examined counsel's fee petitions in accordance with the Grinnell decisions and arrived at individual awards for each PMC member based upon the services that each had provided for the class. By providing for threefold returns of advanced expenses, however, the agreement vitiated these principles. The distortion was so substantial as to increase the fees awarded to one investor by over twelve times that which the district judge had determined to be just and reasonable, and, in a second case, to decrease the otherwise just and reasonable compensation of a non-investor by nearly two-thirds.

There is authority for a court, under certain circumstances, to award a lump sum fee to class counsel in an equitable fund action under the lodestar approach and then to permit counsel to divide this lodestar-based fee among themselves under the terms of a private fee sharing agreement. E.g., Ruskay v. Jensen, No. 71-3169, slip op. at 10-13 (S.D.N.Y. Sept. 18, 1981); In re Magic Marker Securities Litigation, [1979 Transfer Binder] Fed.Sec.L.Rep. (CCH) p 97,116, at 96,195 (E.D.Pa. Sept. 16, 1979); Valente v. Pepsico, Inc., [1979 Transfer Binder] Fed.Sec.L.Rep. (CCH) p 96,921, at 95,863 (D.Del. June 4, 1979), appeal dismissed, 614 F.2d 772 (3d Cir.1980); In re Ampicillin Antitrust Litigation, 81 F.R.D. 395, 400 (D.D.C.1978); Del Noce v. Delyar Corp., 457 F. Supp. 1051, 1055 (S.D.N.Y.1978). We reject this authority, however, to the extent it allows counsel to divide the award among themselves in any manner they deem satisfactory under a private fee sharing agreement. Such a division overlooks the district court's role as protector of class interests under Fed.R.Civ.P. 23(e) and its role of assuring reasonableness in the awarding of fees in equitable fund cases. See Kamens v. Horizon Corp., [1981 Transfer Binder] Fed.Sec.L.Rep. (CCH) p 98,007, at 91,218 & n. 4 (S.D.N.Y. May 26, 1981); Steiner v. BOC Financial Corp., [1980 Transfer Binder] Fed.Sec.L.Rep. (CCH) p 97,656, at 98,490 (S.D.N.Y. Oct. 10, 1980); cf. Jones v. Amalgamated Warbasse Houses, Inc., 721 F.2d 881, 884 (2d Cir.1983) ("if the court finds good reason to do so, it may reject an agreement as to attorneys' fees just as it may reject an agreement as to the substantive claims"), cert. denied, 466 U.S. 944, 104 S. Ct. 1929, 80 L. Ed. 2d 474 (1984). In addition, this approach overlooks the class attorneys' "duty ... to be sure that the court, in passing on [the] fee application, has all the facts" as well as their "fiduciary duty to the ... class not to overreach." Lewis v. Teleprompter Corp., 88 F.R.D. 11, 18 (S.D.N.Y.1980).

A careful examination of those decisions permitting internal fee sharing agreements to govern the distribution of fees reveals no case where return on investment was a factor. More important, in a number of those cases the courts apparently assumed that the internal fee sharing agreement would be based substantially on services rendered by individual counsel. E.g., Ruskay, slip op. at 14 n. 4 ("Since the court has satisfied itself that the proposed distribution will not result in compensation beyond services performed, it declines to overrule the agreement."); In re Ampicillin Antitrust Litigation, 81 F.R.D. at 400 ("Since the fee application purports to be based upon the rates and time spent by the several attorneys, it is presumed that these factors also weigh heavily in this internal agreement.").

Accordingly, while the practice of allowing class counsel to distribute a general fee award in an equitable fund case among themselves pursuant to a fee sharing agreement is unexceptional, we find that any such agreement must comport essentially with those principles of fee distribution set forth in Grinnell I and Grinnell II. This does not mean that a fee sharing agreement must replicate the individual awards made to PMC members under the district court's lodestar analysis. Even after the court makes the allocation, the attorneys may be in a better position to judge the relative input of their brethren and the value of their services to the class. See In re Ampicillin Antitrust Litigation, 81 F.R.D. at 400. Nor does this mean that class counsel need follow, line by line, the lodestar formula in arriving at an agreement as to fee distribution. Obviously, the needs of large class litigation may at times require class counsel, in assessing the relative value of an individual attorney's contribution, to turn to factors more subjective than a mere hourly fee analysis. It does mean that the distribution of fees must bear some relationship to the services rendered.

In our view, fees that include a return on investment present the clear potential for a conflict of interest between class counsel and those whom they have undertaken to represent. "[W]henever an attorney is confronted with a potential for choosing between actions which may benefit himself financially and an action which may benefit the class which he represents there is a reasonable possibility that some specifically identifiable impropriety will occur." Zylstra v. Safeway Stores, Inc., 578 F.2d 102, 104 (5th Cir.1978). The concern is not necessarily in isolating instances of major abuse, but rather is "for those situations, short of actual abuse, in which the client's interests are somewhat encroached upon by the attorney's interests." Court Awarded Attorney Fees, Report of the Third Circuit Task Force, 108 F.R.D. 237, 266 (Oct. 8, 1985). Such conflicts are not only difficult to discern from the terms of a particular settlement, but "even the parties may not be aware that [they exist] at the time of their [settlement] discussions," id. This risk is magnified in the class action context, where full disclosure and consent are many times difficult and frequently impractical to obtain. In re Mid-Atlantic Toyota Antitrust Litigation, 93 F.R.D. 485, 490-91 (D.Md.1982); Gould v. Lumonics Research Ltd., 495 F. Supp. 294, 297 n. 6 (N.D.Ill.1980).

The district court recognized that the agreement provided an incentive for the PMC to accept an early settlement offer not in the best interests of the class, because "[a]n attorney who is promised a multiple of funds advanced will receive the same return whether the case is settled today or five years from now." Agent Orange I, 611 F. Supp. at 1460. Given the size and complexity of the litigation, it seems apparent that the potential for abuse was real and should have been discouraged. Unlike the district court, however, we conclude that the risk of such an adverse effect on the settlement process provides adequate grounds for invalidating the agreement as being inconsistent with the interests of the class. The conflict obviously lies in the incentive provided to an investor-attorney to settle early and thereby avoid work for which full payment may not be authorized by the district court. Moreoever, as soon as an offer of settlement to cover the promised return on investment is made, the investor-attorney will be disinclined to undertake the risks associated with continuing the litigation. The conflict was especially egregious here, since six of the nine PMC members were investing parties to the agreement.

The district court's factual finding, that the adequacy of the settlement demonstrated that the agreement had no effect on the PMC's conduct, is not dispositive. The district court's retrospective appraisal of the adequacy of the settlement cannot be the standard for review. The test to be applied is whether, at the time a fee sharing agreement is reached, class counsel are placed in a position that might endanger the fair representation of their clients and whether they will be compensated on some basis other than for legal services performed. Review based on a fairness of settlement test would not ensure the protection of the class against potential conflicts of interest, and, more important, would simply reward counsel for failing to inform the court of the existence of such an agreement until after a settlement.

We also reject the district court's finding that its authority to approve settlement offers under Fed.R.Civ.P. 23(e) acts to limit the threat to the class from a potential conflict of interest. At this late stage of the litigation, both class counsel and defendants seek approval of the settlement. The court's attention properly is directed toward the overall reasonableness of the offer and not necessarily to whether class counsel have placed themselves in a potentially conflicting position with the class. It would be difficult indeed for a court at this stage to hold that, regardless of the terms of the settlement, class counsel had not fulfilled its obligation to the class. Given this focus and other administrative concerns that may come to bear, we find the approval authority, in this context, to be insufficient to assure that the ongoing interests of the class are protected. See Alleghany Corp. v. Kirby, 333 F.2d 327, 347 (2d Cir.1964) (Friendly, J., dissenting) (at this stage of litigation, "[a]ll the dynamics conduce to judicial approval of such settlements"), cert. dismissed, 384 U.S. 28, 86 S. Ct. 1250, 16 L. Ed. 2d 335 (1966); In re Mid-Atlantic Toyota Antitrust Litigation, 93 F.R.D. at 491 (court authority to review settlement offers not adequate to safeguard against dangers of conflict of interest); Coffee, The Unfaithful Champion: The Plaintiff As Monitor In Shareholder Litigation, 48 Law & Contemp. Probs. 5, 26-27 (Summer 1985) (judicial review not a significant barrier to collusive settlements).

Equally unpersuasive is the district court's determination that the potential incentive to settle early is offset by an incentive, fostered by the lodestar formula, to prolong the litigation. While a number of commentators have asserted that use of the lodestar formula encourages counsel to prolong litigation for the purpose of billing more hours, e.g., Wolfram, The Second Set of Players: Lawyers, Fee Shifting, and the Limit of Proportional Discipline, 47 Law & Contemp. Probs. 293, 302 (Winter 1984), the formula's effect in this regard is far from clear, see Coffee, supra, at 34-35 ("the claim that the lodestar formula results in excessive fees is nonetheless a red herring"); Mowrey, Attorneys Fees In Securities Class Action and Derivative Suits, 3 J.Corp.Law. 267, 343-48 (1978) (attorneys' fees awards by district courts have not risen since adoption of lodestar analysis); see also 7B C. Wright, A. Miller & M. Kane, Federal Practice and Procedure Sec. 1803, at 508 (1986) (no empirical data show any incidence of district courts awarding excessive fees). Moreover, the court's authority in reviewing fee petitions and approving or disapproving hours billed in an equitable fund action works as a substantial and direct check on counsel's alleged incentive to procrastinate. In re Equity Funding Corporation of America Securities Litigation, 438 F. Supp. 1303, 1328 (C.D.Cal.1977); 7B C. Wright, A. Miller & M. Kane, supra, Sec. 1803, at 511. Consequently, we do not view the lodestar system as countervailing the clear interest in early settlement created by the private agreement.

Additionally, potential conflicts of interest in class contexts are not examined solely for the actual abuse they may cause, but also for potential public misunderstandings they may cultivate in regard to the interests of class counsel. Susman v. Lincoln American Corp., 561 F.2d 86, 95 (7th Cir.1977); Prandini v. National Tea Co., 557 F.2d 1015, 1017 (3d Cir.1977). While today we hold that the settlement reached here falls within that range of reasonableness permissible under Fed.R.Civ.P. 23(e), we are not insensitive to the perception of many class members and the public in general that it does not adequately compensate the individual veterans and their families for whatever harm Agent Orange may have caused. To be sure, the settlement does not provide the individual veteran or his family substantial compensation. Given the facts of this settlement, the potentially negative public perception of an agreement that awards an investing PMC member over twelve times the amount the district court has determined to be the value of his services to the class provides additional justification for invalidating the agreement and applying the lodestar formula.

We find the various additional rationales for approving the fee sharing agreement set out in the district court's decision equally unpersuasive. First, the fact that the returns on the advanced expenses did not directly affect the class fund is of little consequence, since we have already determined that the district court's responsibility under Grinnell I and Grinnell II, as well as under Fed.R.Civ.P. 23(e), goes beyond concern for only the overall amount of fees awarded and requires attention to the fees allocated to individual class counsel. Second, while we sympathize with counsel regarding the business decisions they must make in operating an efficient and manageable practice and agree that a certain flexibility on the court's part is essential, we are not inclined to extend this flexibility to encompass situations in which the bases for awarding fees in an equitable fund action are so clearly distorted. Third, whether this class action would have collapsed without an agreement calling for a threefold return is a matter of speculation. Any such collapse, however, would have been due to the pervasive weaknesses in the plaintiffs' case. Fourth, we find wholly unconvincing the district court's suggestion that the investors could have made a sizeable return on their funds if they had invested them in other ventures. We take notice of the fact that a threefold return on one's money is a rather generous return in any market over a short period of time. Fifth, while the effect of this fee sharing agreement might have been dwarfed to the point of insignificance if the fees awarded to counsel had been much greater, this simply is too speculative to defend the agreement as not affecting the interests of the class. Finally, we do not find class counsel to have formed an ad hoc partnership. They merely are a group of individual lawyers and law firms associated in the prosecution of a single lawsuit, and they lack the ongoing relationship that is the essential element of attorneys practicing as partners.

We do agree with the district court's ruling that in all future class actions counsel must inform the court of the existence of a fee sharing agreement at the time it is formulated. This holding may well diminish many of the dangers posed to the rights of the class. Only by reviewing the agreement prospectively will the district courts be able to prevent potential conflicts from arising, either by disapproving improper agreements or by reshaping them with the assistance of counsel to conform more closely with the principles of Grinnell I and Grinnell II. In the present case, however, where the district court was not made aware of the agreement, and the potential for a conflict of interest arising was substantial, the adoption of a rule for future cases in no way alleviates the fatal flaws of this agreement and does not offset the need for its invalidation.

Although appellant Dean is successful on this appeal, his conduct has been far from praiseworthy. He freely consented to the formation of the agreement in December of 1983 and later to its revision in 1984. He did not even inform the district court of the existence of the agreement or of his objections to it until long after the settlement was reached. If he had called the agreement into question immediately, a great deal of time and expense could have been saved.

Having determined that the fee sharing agreement violates the principles for awarding fees in an equitable fund action and places class counsel in a position potentially in conflict with the interests of the class which they represent, we reverse. We award all the PMC members the fees to which the district court determined they were entitled.