United States v. Microsoft Corporation. United States of America v. Microsoft Corporation

Opinion for the Court filed by Circuit Judge SILBERMAN.

Opinion PER CURIAM.

Section 16(e) of the Antitrust Procedures and Penalties Act, known as the Tunney Act, requires the district court to determine whether entry of an antitrust consent decree is “in the public interest.” 15 U.S.C. § 16(e) (1988). In this case, the district court refused to enter a proposed consent decree the Antitrust Division of the Department of Justice negotiated with Microsoft Corporation. We conclude that the proposed consent decree is in the public interest, and that the district court exceeded its authority in concluding to the contrary. We therefore reverse and remand with instructions to enter an order approving the decree.

I.

Microsoft dominates the world market for operating systems software that runs on IBM-compatible personal computers (“PCs”). Operating systems software controls the operation of the computer and manages the interaction between the computer’s memory and attached devices such as keyboards, printers, display screens and disk drives. In 1990, the Federal Trade Commission began investigating Microsoft’s acquisition and maintenance of monopoly power in that market. When faced with the decision whether to file a complaint against Microsoft, however, the Commission deadlocked 2-2, thus suspending the agency’s investigation.

The Antitrust Division of the Department of Justice then initiated its own investigation of Microsoft (apparently a rather rare occurrence), using the FTC’s extensive investigatory file as its starting point. The Department issued 21 Civil Investigative Demands to Microsoft and third parties, reviewed one million pages of documents, and conducted over 100 interviews. The Department also deposed 22 persons, including Microsoft Chairman Bill Gates.

In July 1994, the Department filed a civil complaint under the Sherman Act, 15 U.S.C. §§ 1. and 2 (1988), charging Microsoft with unlawfully maintaining a monopoly of operating systems for IBM-compatible PCs and unreasonably restraining trade of the same through certain anticompetitive marketing practices.

The key anticompetitive practice against which the complaint is aimed is Microsoft’s use of contract terms requiring original equipment manufacturers (“OEMs”) to pay Microsoft a royalty for each computer the OEM sells containing a particular microprocessor (in this case, an x86 class microprocessor), whether or not the OEM has included a Microsoft operating system with that computer. The practical effect of such “per processor licenses,” it is alleged, is to deter OEMs from using competing operating systems during the life of their contracts with Microsoft. The complaint further charges that Microsoft has exacerbated the anticom-petitive effect of the per processor licenses by executing long-term contracts with major OEMs, and by requiring minimum commitments and crediting unused balances to future contracts, thereby extending the contract term and creating an economic disincentive for an OEM to install a non-Microsoft operating system.

The other anticompetitive device alleged in the complaint is Microsoft’s use of overly restrictive nondisclosure agreements with certain independent software vendors (“ISVs”). Those ISVs provide applications software to run “on top of’ Microsoft’s operating system, enabling the user to perform a variety of tasks, such as wordprocessing. Microsoft provides those ISVs with advance test versions of its newest operating system so that the ISVs can develop their software to be compatible with that operating system. The government alleged that Microsoft, as a corollary, has imposed nondisclosure agreements on some ISVs which would restrict their ability to work with competing operating systems companies and to develop com *1452 peting products for an unreasonably long period of time.

The government did not allege and does not contend — and this is of crucial significance to this case — that Microsoft obtained, its alleged monopoly position in violation of the antitrust laws. The government believes that Microsoft’s initial acquisition of monopoly power in the operating systems market was the somewhat fortuitous result of IBM choosing for its PCs the operating system introduced by Microsoft (“MS-DOS”), which, with Microsoft’s successful exploitation of that advantage, led Microsoft to obtain an installed base on millions of IBM, and IBM-compatible, PCs.

It is undisputed that the software market is characterized by “increasing returns,” resulting in natural barriers to entry. Because the costs of producing software are almost exclusively in its design, marginal production costs are “virtually zero.” Professor Arrow, the government’s consultant and a Nobel-prize winning economist, described the importance of Microsoft’s large installed base in an increasing returns market as follows:

A software product with a large installed base has several advantages relative to a new entrant. Consumers know that such a product is likely to be supported by the vendor with upgrades and service. Users of a product with a large installed base are more likely to find that their products are compatible with other products. They are more likely to be able successfully to exchange work products with their peers, because a large installed base makes it more likely that their peers will use the same product or compatible products. Installed base is particularly important to the economic success of an operating system software product. The value of the operating system is in its capability to run application software. The larger the installed base of a particular operating system, the more likely it is that independent software vendors will write programs that run on that operating system, and, in this circular fashion, the more valuable the operating system will be to consumers.

In a not uncommon technique, the Department of Justice filed a proposed consent decree along with its complaint, which embodied the Department’s and Microsoft’s settlement of the case. The consent decree, which essentially tracks the complaint and is effective for 78 months following its entry, prohibits Microsoft from entering into per processor licenses, licenses with a term exceeding one year (unless the customer opts to renew for another year), licenses containing a minimum commitment, and unduly restrictive nondisclosure agreements. To prevent Microsoft from using other exclusionary practices to achieve effects similar to those achieved by the practices challenged in the complaint, the proposed decree also prohibits certain other arrangements such as lump-sum pricing and variants of per processor licensing. The decree applies to Microsoft’s most popular operating systems products (MS-DOS, Windows and Windows 95) and successor versions or operating systems marketed as replacement products. The decree does not, however, cover ‘Windows NT” products, which are designed for sophisticated “high end users” and which do not enjoy a substantial portion of the market for such products.

Pursuant to Section 16(b) of the Tunney Act, 15 U.S.C. § 16(b) (1988), the Department of Justice published the proposed decree in the Federal Register, accompanied by a competitive impact statement, and invited comment. See 59 Fed.Reg. 42,845 (1994). Only five comments were received during the statutory period, to which the government responded on October 31, 1994.

At the first substantive status conference on September 29, 1994, the district judge informed the parties that over the summer he had read a book about Microsoft — Hard Drive 1 — because he “thought it would be a good idea maybe to know as much about Microsoft as probably they’re going to know about me.” Much of the ensuing discussion focused on accusations against Microsoft con- *1453 tamed in the book. The district judge asked whether the government’s lawyers had read the book and whether they had investigated the allegations made by its authors. In particular, the judge focused on the allegation that Microsoft engages in “vaporware,” which he described in differing terms but ultimately defined as “the public announcement of a computer product before it is ready for market for the sole purpose of causing consumers not to purchase a competitor’s product that has been developed and is either currently available for sale or momentarily about to enter the market.” United States v. Microsoft Corp., 159 F.R.D. 318, 334 (D.D.C.1995) (“Opinion”). 2 (The judge insisted that even truthful product prean-nouncements would violate the securities laws, if not the antitrust laws.)

According to the district judge, Hard Drive also claimed that Microsoft’s own applications developers have unfair access to information about Microsoft’s operating systems, giving them an undue advantage over competitors in developing applications software that is compatible with Microsoft’s operating systems. The government did not include “vaporware” or unfair access charges in its complaint against Microsoft.

At a subsequent status hearing on November 2, 1994, the district judge again referred to Hard Drive and its “vaporware” allegations, noting that the book “does allege some very serious practices,” and telling the government that he wanted to be satisfied that the allegations in the book were not true. The Department was instructed to inform interested persons that they had until December 5,1994 to seek leave to participate in the court’s hearing on the consent decree. Only I.D.E. Corporation, which had participated in the comment process, sought leave to participate in the hearing. But on January 10, 1995, (over a month late and over the objection of both the government and Microsoft), the law firm of Wilson, Sonsini, Goodrich & Rosati, on behalf of three computer industry companies (“Doe Companies”), filed a 96-page memorandum (plus a 215-page appendix) arguing that the proposed consent decree was inadequate because it would not result in increased competition in the operating systems market, nor prevent Microsoft from monopolizing the rest of the software industry. The Doe Companies claimed that because of the unusual “increasing returns” nature of Microsoft’s market position, it would be extremely difficult to dislodge Microsoft from its dominant status and return the market to a state of equilibrium, or competition. Moreover, they claimed that Microsoft had the capacity to leverage its installed base in the operating systems market so as to dominate the related markets for applications and other software products. The Doe Companies also attached two documents purporting to show that Microsoft had engaged in “vaporware.” 3 Wilson, Sonsini’s brief was accompanied by a motion requesting that the district court permit the late filing, and also permit the purported computer industry companies to remain anonymous, asserting “fear” that they would be subject to unexplained retaliation from Microsoft. The district court, without a hearing on the need for or propriety of the Doe Companies’ proceeding anonymously, granted the motion over the government’s and Microsoft’s objections.

On January 18, 1995, the United States filed a motion for entry of the decree (later joined orally by Microsoft at the district court’s January 20, 1995 hearing) and attached an affidavit from Professor Arrow. As noted, Professor Arrow agreed with ami- *1454 ci that in an increasing returns market there is a possibility of monopolization, which may be inefficient; but, he claimed, this process is entirely natural. He specifically rejected the notion that the government should intervene where, as he believed was the case here, the market success of the dominant firm was not the result of anticompetitive practices. Professor Arrow concluded that only artificial barriers, such as the licensing practices addressed in the decree, should be regulated or prohibited.

The next day the district court issued an order identifying issues to be addressed at the hearing scheduled for the following day. The parties were requested to explain why, among other things, the consent decree did not contain provisions that would (1) bar Microsoft from engaging in “vaporware,” (2) establish a wall between the development of operating systems software and the development of applications software at Microsoft, and (3) require disclosure of all instruction codes built into operating systems software designed to give Microsoft an advantage over competitors in the applications software market. See Opinion, 159 F.R.D. at 326-27 n. 15. The Computer & Communications Industry Association (“CCIA”) filed a motion for leave to intervene, or alternatively, to participate as amicus curiae.

The district court allowed I.D.E. Corporation, CCIA and the Doe Compames to participate in the January 20, 1995 hearing. 4 All three urged disapproval of the decree. The district judge devoted substantial time to questioning counsel about “vaporware” and pressing the government for information regarding its investigation of “vaporware” allegations — information which the government declined to provide on the ground that the such allegations were unrelated to the violations charged in the complaint. 5

On February 14, 1995, the district court issued an order denying the government’s motion to approve the consent decree. The judge stated that he could not find the proposed decree to be in the public interest for four reasons:

First, the Government has declined to provide the Court with the information it needs to make a proper public interest determination. Second, the scope of the decree is too narrow. Third, the parties have been unable and unwilling adequately to address certain anticompetitive practices, which Microsoft states it will continue to employ in the future and with respect to which the decree is silent. Thus, the decree does not constitute an effective antitrust remedy. Fourth, the Court is not satisfied that the enforcement and compliance mechanisms in the decree are satisfactory.

Opinion, 159 F.R.D. at 332.

The judge’s understanding of the extent of additional information he “need[ed]” to make *1455 the public interest determination under the Tunney Act was quite considerable and was of a character that the government contended was not only outside the scope of the Tunney Act, but was also improper for a judge to seek. The court required at a “minimum”:

(1) The broad contours of the investigation ie., the particular practices and conduct of the defendant that were under investigation along with the nature, scope and intensity of the inquiry;

(2) With respect to such particular practices and conduct, what were the conclusions reached by the Government;

(3) In the settlement discussions between the Government and defendant: (a) what were the areas that were discussed, and (b) what, if any, areas were bargained away and the reasons for their non-inclusion in the decree;

(4) With respect to the areas not discussed at the bargaining table or not bargained away, what are the plans for the Government to deal with them ie., is the investigation to continue, and, if so, at what intensity, or if the investigation is to be closed, then the Government must explain why it is in the public interest to do so.

Opinion, 159 F.R.D. at 332.

The judge’s second objection, going to the scope of the decree, was predicated on his concern that it does not apply to all of Microsoft’s operating systems. The decree, it will be recalled, explicitly excludes from its coverage ‘Windows NT.” In an apparent reference to Microsoft’s contention that its product preannouncements do not fit the definition of “vaporware” and do not constitute antitrust violations, the court further noted that “taking into account Microsoft’s penchant for narrowly defining the antitrust laws, the Court fears there may be endless debate as to whether a new operating system is covered by the decree.” Id. at 333.

As to the third point — the essence of the amici’s objection — the judge concluded that the decree does not provide an effective antitrust remedy because it does not “pry open” the market to competition, ie., remedy the monopolist position Microsoft has achieved through supposed illegal means. The judge was especially concerned that the decree does not address “a number of other anti-competitive practices that from time to time Microsoft has been accused of engaging in by others in the industry.” Id. at 334. Among such practices were “vaporware,” Microsoft’s “use[ ] [of] its dominant position in operating systems to give it an undue advantage in developing applications software,” and its manipulation of its operating systems to render competing applications software inoperable or more difficult for consumers to use. Id. at 334-35.

Finally, the court determined that the consent decree did not oblige Microsoft to adopt sufficient internal compliance mechanisms. Based on its perception that Microsoft had misled the court about whether it engaged in “vaporware,” the district court concluded that Microsoft’s current staff of “50 or so in-house lawyers, along with its outside retained counsel,” were insufficient to monitor the decree adequately. Id. at 336.

The United States and Microsoft appeal from the order refusing to enter the decree, asking this court to remand with instructions to enter the decree. Microsoft appeals as well from the order allowing the anonymous participation of the Doe Companies (and CCIA’s participation), and asks that the case be remanded to another district court judge because it contends that Judge Sporkin has demonstrated personal bias against the company.

Since both parties to the decree have appealed the district court’s order, these consolidated cases present the rare situation in which there is no appellee. Accordingly, we have allowed the Doe Companies, CCIA and I.D.E. Corporation to continue in their roles as amici.

II.

Both the government and Microsoft contend that the district judge vastly exceeded his authority under the Tunney Act, and that as a matter of law they are entitled to the court’s entry of the consent decree. Before considering their arguments, however, we are obliged to determine that we have jurisdiction to entertain this appeal. Amici con *1456 tend that we do not. The government (and Microsoft) rely on 28 U.S.C. § 1292(a)(1) (1988), which authorizes interlocutory appeals from orders of the district court “granting, continuing, modifying, refusing or dissolving injunctions....” By refusing to enter the consent decree, the district court, it is argued, has refused to grant an injunction within the meaning of that statute. The leading case on point is Carson v. American Brands, Inc., 450 U.S. 79, 101 S.Ct. 993, 67 L.Ed.2d 59 (1981), where the Supreme Court held that a district court’s refusal to enter a consent decree that had the practical effect of denying an injunction was immediately appealable if the order had a “serious, perhaps irreparable, consequence,” and could only be “effectually challenged” by immediate appeal. Id. at 84, 101 S.Ct. at 996-97, citing Baltimore Contractors, Inc. v. Bodinger, 348 U.S. 176, 181, 75 S.Ct. 249, 252-53, 99 L.Ed. 233 (1955).

The decree at issue here does, as in Carson, call for an injunction: Microsoft would be permanently enjoined from using what the government contends are anticompetitive licensing contracts for its PC operating systems. “Indeed, prospective relief [is] at the very core of the disapproved settlement.” Id. To be sure, in Carson the district court indicated its disapproval of the relief sought by a civil rights plaintiff because it was arguably too extensive; whereas here, the district judge is objecting because the relief does not appear to him to go far enough. Amici thus describe the district judge’s order as a refusal to limit the potential relief available rather than a refusal to grant injunctive relief. That seems to us to be only a deft semantic characterization. It matters not whether a district judge objects to the injunctive relief as too strong or not strong enough: in either case, the judge is refusing to grant the injunction except under conditions that the parties will not accept. Nor is there any doubt in this case that the district judge had reached a firm determination. His order and opinion make that quite clear.

Amici nevertheless contend that the government has not shown any “serious consequences” that justify the interlocutory appeal. The government, it is argued, is only concerned about its settlement statistics. We disagree. The district court’s refusal to enter the decree puts the government to a difficult, perhaps Hobson’s, choice. It must either drop its case against Microsoft entirely and allow Microsoft to continue to engage in practices which the government believes are anticompetitive; or, it is compelled to litigate and presumably proceed under a vastly expanded complaint that in effect asserts that Microsoft engaged in activities which the government does not believe are illegal (ie., achieving its dominant position in the first place), or seeks remedies which the government does not believe are justified by the evidence. Moreover, as the government points out, the consent decree is part of a negotiated settlement. A district judge’s refusal to accept the decree — particularly upon the grounds advanced — cannot but have enormous practical consequences for the government’s ability to negotiate future settlements. Cf. Carson, 450 U.S. at 88 n. 14, 101 S.Ct. at 998 n. 14 (order refusing to enter consent decree would undermine Title VII’s strong preference for encouraging voluntary settlement of employment discrimination claims). The Tunney Act was not intended to create a disincentive to the use of the consent decree. See, e.g., S.Rep. No. 298, 93d Cong., 1st Sess. 7 (1973) (“The [Senate Judiciary] Committee wishes to retain the consent judgment as a substantial antitrust enforcement tool.”); H.Rep. No. 1463, 93 Cong., 2d Sess. 6 (1974) (expressing intent to preserve the policy of the antitrust laws to encourage settlement by consent decree), reprinted in 1974 U.S. Code Cong. & Admin. News 6535, 6536-37. We conclude, therefore, that the government easily meets the Carson standard. We have jurisdiction over the appeal.

Microsoft has also appealed the judge’s refusal to enter the consent decree. Actually, in this situation, it is doubtful if we would have jurisdiction unless both parties to the decree appealed. Certainly if the government accepted the district judge’s view of the case and wished to proceed to trial, the propriety of the judge’s order would be moot; and, if Microsoft was no longer willing to agree to the government’s conditions, the *1457 issue would similarly be moot. 6 Therefore, both parties must be entitled to appeal the district judge’s refusal to enter the decree. Microsoft in addition appeals one order that the government does not: the order allowing participation by the Doe Companies and CCIA. Microsoft asserts that the district court erred in permitting the Doe Companies to participate anonymously.

Microsoft argues that the order fits within the “collateral order” doctrine of Cohen v. Beneficial Industrial Loan Corp., 337 U.S. 541, 69 S.Ct. 1221, 93 L.Ed. 1528 (1949), which permits interlocutory review of collateral orders “that are conclusive, that resolve important questions completely separate from the merits, and that would render such important questions effectively unreviewable on appeal from final judgment in the underlying action.” Digital Equip. Corp. v. Desktop Direct, Inc., — U.S. -, -, 114 S.Ct. 1992, 1995-96, 128 L.Ed.2d 842 (1994). Normally, of course, an order permitting or denying amici participation could not qualify under the doctrine. Here, however, the Doe Companies, the prime opponents of the decree and the most vigorous accusers of Microsoft, appeared anonymously. That is quite a departure from normal procedure, and raises profound questions of fundamental fairness and perhaps even due process. It might well be that a party forced to confront an anonymous plaintiff (or the functional equivalent of a plaintiff acting as an amici) could suffer injury that might not be redress-able in an ultimate appeal. Anonymity may well confer a kind of immunity which permits a plaintiff to hurl rhetorical weapons that could cause a unique kind of harm not faced in ordinary litigation. Cf. McIntyre v. Ohio Elections Comm’n, — U.S. -, -, 115 S.Ct. 1511, 1524, 131 L.Ed.2d 426 (1995) (holding that Ohio election law forbidding all anonymous political leafletting violated the First Amendment but noting that “[t]he right to remain anonymous may be abused when it shields fraudulent conduct”). However, in our view, it is unnecessary to decide whether Microsoft would have been entitled to appeal this order independently. Since we have concluded that we have jurisdiction over the basic appeal, the district judge’s order permitting the amici to participate is not really interlocutory. It comes before us as part and parcel of the record in this case.

III.

Appellants contend that the district judge misinterpreted the Tunney Act — indeed interpreted that statute so as to raise serious questions regarding its constitutionality — by basing his rejection of the decree on considerations which implicate the executive branch’s prosecutorial discretion. The thrust of the judge’s concerns were directed to his dissatisfaction with the framework of the complaint fashioned by the Department. He thought it much too modest to deal with the imperfections in the relevant market and their cause — at least as he perceived them. Appellants contend that the judge did not simply make the proper inquiry into whether the decree was appropriate to the complaint, but instead asked whether the complaint itself was adequate. By doing so, it is argued, the judge improperly intruded on the government’s prosecutorial role. The judge’s demand that he be informed of the contours of the investigation, the settlement discussions, and the government’s future investigative plans, indicates that the judge impermissibly arrogated to himself the President’s role “to take care that the laws be faithfully executed.”

Amici, defending the judge’s order, argue that it merely focused on whether the remedy provided in the decree was adequate to the allegations in the complaint. Appellants respond, however, that even to the extent that the judge’s order is directed to the adequacy of the remedy to the allegations actually charged — which they insist is only a minor theme in the judge’s opinion — the judge nevertheless exceeded his authority. Under our own precedent dealing with uncontested modifications of a consent decree, we have repeatedly said that a district judge *1458 must approve such modifications so long as the proposal falls “within the reaches of the public interest.” United States v. Western Elec. Co., 900 F.2d 283, 309 (D.C.Cir.1990) (“Triennial Review Opinion ”) (emphasis in original) (quoting United States v. Bechtel Corp., 648 F.2d 660, 666 (9th Cir.), cert. denied, 454 U.S. 1083, 102 S.Ct. 638, 70 L.Ed.2d 617 (1981), in turn quoting United States v. Gillette Co., 406 F.Supp. 713, 716 (D.Mass.1975)).

At the heart of this case, then, is the proper scope of the district court’s inquiry into the “public interest.” Is the district judge entitled to seize hold of the matter— the investigation into the putative defendant’s business practices — and decide for himself the appropriate combined response of the executive and judicial branches to those practices? With respect to the specific allegations in the government’s complaint, may the court interpose its own views of the appropriate remedy over those the government seeks as a part of its overall settlement? To be sure, Congress, in passing the Tunney Act, intended to prevent “judicial rubber stamping” of the Justice Department’s proposed consent decree. H.R.Rep. No. 1463, supra, at 8, reprinted in 1974 U.S. Code Cong. & Ajdmin. News at 6538. The Court was to “make an independent determination as to whether or not entry of a proposed consent decree [was] in the public interest.” S.Rep. No. 298, supra, at 5. Yet, Congress did not purport to alter antitrust precedent applying the public interest in reviewing consent decrees. H.R.Rep. No. 1463, supra, at 11, reprinted in 1974 U.S. Code Cong. & Admin. News at 6539. The difficulty with that stated purpose is that there was virtually no useful precedent — certainly none in which an appellate court had approved a trial court’s rejection of a consent decree as outside the public interest. Cf. Antitrust Procedures and Penalties Act: Hearings on S.782 and S.1088 Before the Subcomm. on Antitrust and Monopolies of the Senate Comm, on the Judiciary, 93d Cong., 1st Sess. 92 (1973) (“Senate Hearings”) (Statement of Thomas E. Kauper, Assistant Attorney General, Antitrust Division, Dept, of Justice) (“[E]xcept in cases where a previous judicial mandate is involved and the consent decree fails to comply with that mandate, or where there is a showing of bad faith or malfeasance, the courts have allowed a wide range of prosecutorial discretion.”).

Although the statute does not give specific guidance, it does speak in rather broad terms. In determining whether the decree is in the public interest, the district court is authorized to “consider”:

(1) the competitive impact of such judgment, including termination of alleged violations, provisions for enforcement and modification, duration or relief sought, anticipated effects of alternative remedies actually considered, and any other considerations bearing upon the adequacy of such judgment;

(2) the impact of entry of such judgment upon the public generally and individuals alleging specific injury from the violations set forth in the complaint including consideration of the public benefit, if any, to be derived from a determination of the issues at trial.

15 U.S.C. § 16(e) (1988).

The Ninth Circuit observed (in a case in which the defendant wished to withdraw from the decree) that that language “suggests that a court may, and perhaps should, look beyond the strict relationship between complaint and remedy in evaluating the public interest.” Bechtel Corp., 648 F.2d at 666. But it went on to determine that “we cannot agree that a district court should engage in an unrestricted evaluation of what relief would best serve the public.” Id. 7

The most prominent post-Tunney Act consent decree, the AT & T consent decree, was modified by the district judge in several respects in accordance with his views of the public interest and, although both the government and AT & T acquiesced, non-parties *1459 to the decree were allowed to intervene for purposes of appealing the district judge’s public interest determination. It does not appear that any of the appellants challenged the constitutionality of the Tunney Act, but Justice Rehnquist, speaking for Chief Justice Burger and Justice White, nevertheless published a dissent from the order affirming the district court’s entry of the decree. See Maryland v. United States, 460 U.S. 1001, 103 S.Ct. 1240, 75 L.Ed.2d 472 (1983). The dissent expressed grave doubt as to the Act’s constitutionality because without a judicial finding of illegality (a consent decree is, of course, a settlement), the statute does not supply a judicially manageable standard for review of the decree, id. at 1004, 103 S.Ct. at 1242, and the considerations that led the Department of Justice to settle are not amenable to judicial review, id. at 1005-06, 103 S.Ct. at 1243-44. See also Heckler v. Chaney, 470 U.S. 821, 831, 105 S.Ct. 1649, 1655-56, 84 L.Ed.2d 714 (1985). For instance, a settlement, particularly of a major case, will allow the Department of Justice to reallocate necessarily limited resources.

The government, cautioning us as to the constitutional difficulties that inhere in this statute, urges us to flatly reject the district judge’s efforts to reach beyond the complaint to evaluate claims that the government did not make and to inquire as to why they were not made. We agree. Although the language of section 16(e) is not precise, we think the government is correct in contending that section 16(e)(l)’s reference to the alleged violations suggests that Congress did not mean for a district judge to construct his own hypothetical case and then evaluate the decree against that case. Moreover, in section 16(e)(2), the court is authorized to consider “the public benefit ... of the determination of the issues at trial.” Putting aside the perplexing question of how the district judge could insure a trial if the government did not wish one, “the issues” referred to must be those formulated in the complaint. Congress surely did not contemplate that the district judge would, by reformulating the issues, effectively redraft the complaint himself. We therefore dismiss the claim that the last line in section 16(e)(1), the catchall clause allowing the district court to entertain “any other considerations bearing upon the adequacy of such judgment,” authorizes the wide-ranging inquiry the district court wished to conduct in this case. That language recognizes, inter alia, that a consent decree might well do unexpected harm to persons other than those “alleging specific injury from the violations set forth in the complaint.” 15 U.S.C. § 16(e)(2) (1988). And the district court might ponder those sort of concerns in determining whether to enter the judgment.

To be sure, the Act also authorizes the district judge to “take testimony of Government officials ... as the court may deem appropriate.” 15 U.S.C. § 16(f)(1) (1988). We do not read this language, however, to authorize the district judge to seek the kind of information concerning the government’s investigation and settlement negotiations that he wished to obtain here. Even when a court is explicitly authorized to review government action under the Administrative Procedure Act, “there must be a strong showing of bad faith or improper behavior” before the court may “inquir[e] into the mental processes of administrative decisionmakers.” Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 420, 91 S.Ct. 814, 825, 28 L.Ed.2d 136 (1971). Here, the district court is not empowered to review the actions or behavior of the Department of Justice; the court is only authorized to review the decree itself. It is unnecessary to consider whether the district court might have broader authority to inquire into the Department’s deliberations, even though not authorized to “review” the Department’s action, if there were a credible showing of bad faith. See Senate Hearings, supra, at 92. There is no such claim here.

The district court was troubled that if its review were limited to the market and practices within that market against which the complaint was directed, the government could, by narrow drafting, artificially limit the court’s review under the Tunney Act. See Opinion, 159 F.R.D. at 332. We think, with all due respect, that the district court put the cart before the horse. The court’s authority to review the decree depends en *1460 tirely on the government’s exercising its prosecutorial discretion by bringing a case in the first place.

That brings us to amici’s contention that the district court was justified in rejecting the decree as providing inadequate remedies, even if the court was barred from reaching beyond the complaint to examine practices the government did not challenge. The district judge (and amici) believed that the decree would not “effectively pry open to competition a market that has been closed by defendant[’s] illegal restraints.” Opinion,