Opinion for the Court filed by Circuit Judge STEPHEN F. WILLIAMS.
STEPHEN F. WILLIAMS, Circuit Judge:
When the government and AT & T settled the government’s antitrust suit against AT & T, the Department of Justice pledged to report to the district court every three years as to whether there was a continuing need for the “line of business” restrictions that the settlement decree, the “Modified Final Judgmént”, imposed on the Bell operating companies (or BOCs). In 1987 the Department proposed that many of the restrictions be removed, including those on entry into the information services market. The district court rejected the information services proposal and several others. On review, this court largely affirmed, but reversed and remanded as to information services, finding that the district court had employed the wrong standard in reviewing the Department’s proposal. See United States v. Western Electric Co., 900 F.2d 283, 289-93, 305-09 (D.C.Cir.1990) (“Triennial Review Opinion”).
On remand, the district court modified the decree, in an opinion that is something of a pushmi-pullyu. See United States v. Western Electric Co., 767 F.Supp. 308 (D.D.C.1991) (“Remand Decision”). While on the one hand asserting that “none of [the Department of Justice’s intermediate contentions] is supported by credible evidence,” id. at 330, the court also concluded that under the standard established by this court in the Triennial Review Opinion it was bound to defer to those conclusions, id. The district court thus removed the ban on entry into [272]*272information services. It stayed its own order of modification, but this court set the stay aside. United States v. Western Electric Co., 1991-2 Trade Cas. (CCH) ¶ 69,610, 1991 WL 238308 (D.C.Cir.1991).
A variety of media and information services interests challenge the modification. Although the district court in certain places mischaracterized the standard under which it was to review the Department’s proposal, the evidence was such that, under the standard laid down in the Triennial Review Opinion, rejection of the Department’s proposal would have been error. We affirm.
The standard for review of modifications on mutual consent
Under the Antitrust Procedures and Penalties Act, 15 U.S.C. §§ 16(b)-(h), known as the Tunney Act, the district court reviewing a consent decree must determine whether the entry of such judgment is “in the public interest.” Id. at § 16(e). Although application of the Tunney Act to decree modifications has been contested, see Note, Modifications of Antitrust Consent Decrees: Over a Double Barrel, 84 Mich.L.Rev. 134 (1985), all participants have assumed the point, as did we in the Triennial Review Opinion, 900 F.2d at 295. There we held that the “public interest test”, as applied to a modification assented to by all parties to a decree,1 “directs the district court to approve an uncontested modification so long as the resulting array of rights and obligations is within the zone of settlements consonant with the public interest today.” 900 F.2d at 307 (emphasis in original). That formulation made clear that it was not up to the court to reject an agreed-on change simply because the proposal diverged from its view of the public interest. Rather, the court was bound to accept any modification that the Department (with the consent of the other parties, we repeat) reasonably regarded as advancing the public interest. In our final sentence on the standard, we underscored the deferential character of the inquiry:
The court should bear in mind flexibility of the public interest inquiry: the court’s function is not to determine whether the resulting array of rights and liabilities “is one that will best serve society,” but only to confirm that the resulting “settlement is “within the reaches of the public interest.’ ”
Id. at 309 (emphasis in original) (citing and quoting United States v. Bechtel Corp., 648 F.2d 660, 666 (9th Cir.1981), in turn quoting United States v. Gillette Co., 406 F.Supp. 713, 716 (D.Mass.1975)).
The appellants address the standard of review issue by selectively citing other language in the Triennial Review Opinion, assigning an absurd meaning to that language, and contending that therefore the language could not really be read to describe the standard for judicial evaluation of an uncontested modification. Finally — the point of their exercise — appellants leap to the conclusion that once we have banished the whipping boy of their creation, nothing will remain except a standard of de novo judicial evaluation. See Appellants’ Brief at 54-68.
The language out of which appellants spin this argument is our statement that “Because the ‘public interest’ test must take its meaning from the nation’s antitrust laws ..., the appropriate question ... is whether the proposed modification would be certain to lessen competition in the relevant market.” Triennial Review Opinion, 900 F.2d at 308 (emphasis added). By reading “certain” to imply absolute certainty, appellants give the language a meaning that would extinguish Tunney Act review. Indeed, the district court voiced the same idea, saying that “the certainty standard may well be an impossible one”. Remand Decision, 767 F.Supp. at 331. See also id. at 330-31 n. 116 (saying that our decision required opponents of noncontested modification to establish anticompetitive effects even more persuasively than under the “beyond-a-reasonable-doubt standard applicable to criminal cases”).
This notion of utter certainty, and its correlative of abject deference to the Department, is a straw man. The focus of the [273]*273disputed sentence was the nature of the factors that fit within the “public interest”, which the court explained were antitrust considerations — the probable effects on competition within the relevant market. The sentences of the Triennial Review Opinion that address the role of the district court in relation to Department of Justice submissions (the passages quoted above on the “zone” of settlements that fall within the “reach” of the public interest) make clear that the district court may reject an uncontested modification only if it has exceptional confidence that adverse antitrust consequences will result— perhaps akin to the confidence that would justify a court in overturning the predictive judgments of an administrative agency. Baltimore Gas & Electric Co. v. NRDC, 462 U.S. 87, 103, 103 S.Ct. 2246, 2255, 76 L.Ed.2d 437 (1983) (when examining predictions within an agency’s expertise, “a reviewing court must generally be at its most deferential”); Telocator Network of America v. FCC, 691 F.2d 525, 550 n. 191 (D.C.Cir.1982); NARUC v. FCC, 525 F.2d 630, 638-39 (D.C.Cir.1976). We recognize that we are not, strictly speaking, reviewing agency action. However, as the Antitrust Division’s assessment of likely competitive effect involves the same kind of predictive judgment, and the court’s role is similarly secondary, that relation at least provides a helpful analogy.
The above view accords with other courts’ understanding of the court-agency relation under the Tunney Act. As the Ninth Circuit found in Bechtel, “The balancing of competing social and political interests affected by a proposed antitrust decree must be left, in the first instance, to the discretion of the Attorney General.” 648 F.2d at 666. And in United States v. Mid-America Dairymen, Inc., 1977-1 Trade Cas. (CCH) ¶ 61,508 (W.D.Mo.1977), while the court spoke of making an “independent public interest determination”, it also found that it was not “to make de novo determination of facts and issues”, and that its task was to determine whether the Department of Justice’s explanations were “reasonable under the circumstances.” Id at 71,979-80. See also United States v. Yoder Bros. Inc., 1989-2 Trade Cas. (CCH) ¶ 68,723 at 61,795 (N.D. Ohio 1986).
Insofar as appellants claim that the deferential standard adopted in the Triennial Review Opinion (and in Bechtel and the other cases noted above) is at odds with the Tunney Act and other judicial interpretations, see Appellants’ Brief at 28-54, the claim is one that they must pursue either by a suggestion for rehearing en banc or by petition for certiorari. We briefly note one special aspect of their argument, however. They suggest that it is inappropriate to defer to the Department when it is acting in a litigation mode, citing Crandon v. United States, 494 U.S. 152, 177-78, 110 S.Ct. 997, 1011, 108 L.Ed.2d 132 (1990) (Scalia, J., concurring). Here the appellants are looking into the wrong end of the telescope. Parties to a contract have voluntarily decided to modify it; the interesting question is not why courts should allow modification, but why they should consider upsetting it. The answer appears to lie in the judicial involvement in the process. Cf. Sam Fox Publishing Co. v. U.S., 366 U.S. 683, 689, 81 S.Ct. 1309, 1312, 6 L.Ed.2d 604 (1961) (review to assure absence of bad faith or malfeasance in entry of consent decree); Note, Modifications of Antitrust Consent Decrees: Over a Double Barrel, 84 Mich.L.Rev. 134, 146 (1985) (arguing that judicial role in modification is grounded in consent decree’s having itself been a judicial act). We need not explore these issues further, except to say that, even apart from the circuit law established in the Triennial Review Opinion, appellants are wide of the mark in supposing that courts should approach the parties’ agreement to a modification with the same wariness as they would an adversary party’s litigating position.
Insofar as appellants argue that our Triennial Review Opinion did not impose a standard of “strict certainty”, see Appellants’ Brief at 61, of course we agree — at least if that is understood to involve more deference than is conventional between court and agency on expert, predictive judgments. Despite the phrases quoted above from the Remand Decision, it is unclear what deference the district court actually extended. If it had demanded absolute certainty, i.e., indulged in total, rubber-stamp deference, we would have to remand for reconsideration under the correct standard — unless the record before the [274]*274court was such that any district court rejection of the proposed modification would have been reversible error. See Pullman Standard v. Swint, 456 U.S. 273, 292, 102 S.Ct. 1781, 1792, 72 L.Ed.2d 66 (1982). In fact the record satisfies that standard. Accordingly we need not enter the elusive inquiry as to what standard the district court “really” applied. Nor need we consider the scope of our review of the district court findings. Although Fed.R.Civ.P. 52 prescribes the “clearly erroneous” standard for actions tried to the court without a jury, the better analogy may be to appellate review of district court application of the “substantial evidence” standard. In that context the district court decision drops out and the appellate court reviews the agency decision directly, its sole deference being to the agency. Novicki v. Cook, 946 F.2d 938, 941 (D.C.Cir.1991); Costello v. Agency for Int’l Development, 843 F.2d 540, 543 n. 5 (D.C.Cir.1988). We need not address the issue of deference to the findings of the district court, however, for to the extent that they contradict the Department’s conclusions they are indeed clearly erroneous.
Record evidence on the effects of removing the ban on entry into information services
In advocating removal of the information services ban, the Department made the following contentions: (1) that there was no substantial risk that removal of the ban would lessen competition either through the BOCs’ use of their “bottleneck” power over local loop connections (i.e., intra-exchange services) to discriminate against competing information service providers, or through cross-subsidization of their own information services; (2) that under current conditions regulation would play a substantial role in minimizing any anti-competitive risks; and (3) that removal of the ban would benefit consumers by enhancing competition in information services. See Memorandum of the United States in Support of Motions for Removal of the Information Services Restriction, Joint Appendix (“J.A.”) 797; see also Remand Decision, 767 F.Supp. at 330.
Participants on all sides of the dispute submitted affidavits on these issues, many by very distinguished economists. The issue before the court was limited: to determine whether the Department’s views were well enough substantiated that it (the Department) could reasonably conclude that removal of the ban was in the public interest. The issue before us, under our self-imposed burden of deference to the district court, is whether that court was clearly erroneous to the extent that it found the Department’s conclusion unreasonable.
(1) BOC power to discriminate against competing information service providers and to cross-subsidize. It is undisputed that the BOCs have monopolies over local telephone exchange service in their respective service areas. Remand Decision, 767 F.Supp. at 314-15. The district court was convinced that the BOCs could use that monopoly power to raise the transmission costs of any rival information service provider, thereby raising the costs and reducing the output of information services. Id. at 315. Raising the costs could take the form either of raising prices explicitly or degrading service (as by delayed installation of equipment, low priority restoration of service after an outage, or poor maintenance). See Fisher Aff., par. 15, J.A. 2031-32.
In assessing the ability of BOCs to raise rivals’ costs, we assume arguendo that devices for avoiding the local loop altogether, such as satellite dishes, are inadequate. See Remand Decision, 767 F.Supp. at 315. But the record provides strong support — which the district court did not address at all — for the view that, quite apart from such end-runs, BOCs will be unable to discriminate against competing providers. Such discrimination could theoretically occur at either end of the information service transaction — the provider’s or the customers’. Take the provider’s end first. Professor Fisher notes that providers may prevent such discrimination by exploiting competition between BOCs, and between BOCs and non-BOC telephone companies, by moving or threatening to move their distribution facilities (the point where their information is fed into the telephone system) to a different region or to an independent company within the BOC’s region. Fisher Aff., pars. 14-16, J.A. at 2031-33. Further, as Professors Stigler and [275]*275Carlton say, large information service providers not only can but already do bypass the BOCs by constructing private networks. Two of the appellants’ primary experts acknowledge the ability of high-volume providers to use by-pass to defeat any discrimination attempts. See Hall Aff., at 16, J.A. 2167; Noll Aff., at par. 8, J.A. 2584. While some providers may well be too small to defeat discrimination this way, it is hard to see what advantage a BOC could draw from beating down small rivals whose customers could readily shift to the BOC’s larger competitors. Carlton/Stigler Aff., pars. 34-35, J.A. at 1857-58.
This takes us to the customer end. See Noll Aff., par. 8, J.A. at 2584. Despite the local monopoly, the record indicates that even here there is much to undercut the threat of discrimination. First, it is evidently difficult, if not impossible, for BOCs to distinguish between information service provider messages and others, as both may come over regular voice lines. Carlton/Stigler Aff., par. 28, J.A. at 1854; Hausman Aff., par. 34, J.A. at 2235. Moreover, as Professor Fisher notes, a BOC would not only have to distinguish between voice and data messages, but also between those data messages coming from a competitor and those from a noncompetitor (e.g., interoffice transmissions by corporations). Fisher Aff., par. 17, J.A. at 2033. The necessary scale of BOC detections would be enormous; Fisher observes that on one information system alone (DunsNet) customers gain access 60,-000 times a day. Id. at par. 18, J.A. 2033-34.
The discrimination hypothesis appears to assume, moreover, that local interconnections are a major element of the total costs of providing information services. This seems untrue. One major information service provider well-known to lawyers, Commerce Clearing House, reports that only 15-20% of its costs are due to telecommunications, of which 30% are local exchange costs. Fisher Aff., par. 19, J.A. at 2034. Thus only about 4.5% to 6% of CCH’s costs are susceptible to BOC discrimination. Even a 50% increase in local exchange costs would raise CCH’s costs by only 3%. Another affiant cites evidence that local exchange costs amount to only 1% of providers’ total costs. See Hausman Aff., par. 24, J.A. 2231.
The BOCs also offered evidence that non-BOC local exchange companies such as GTE have entered the information services market without using their bottleneck control anti-competitively. Remand Decision, 767 F.Supp. at 316. The district court dismissed the GTE experience on the ground that GTE had no ability or incentive to discriminate due to its “dispersed geographical base.” Id. at 321; see also United States v. GTE Corp., 603 F.Supp. 730, 734 (D.D.C.1984). The distinction is far less compelling than it first appears, as its validity depends on the idea that local exchange dispersion would make it materially easier for an information service provider to escape from discrimination at the provider end. As we have already seen, however, the providers’ means of escape from the BOCs are quite good — all that need be relocated is the origin of the provider’s distribution system. Further, the difference in degree of dispersion is by no means overwhelming. First, the BOCs are not AT & T, but seven independent and potentially competing companies, each of which has independent telephone companies operating within its region. Fisher Aff., at par. 16, J.A. 2032-33. Second, GTE itself is not the totally scattered entity envisioned by the district court. As this court has previously noted, GTE controls local exchange service in the entire state of Hawaii as well as in large portions of the Tampa and Los Angeles markets. Illinois Bell v. FCC, 883 F.2d 104, 112 n. 6 (D.C.Cir.1989); Fisher Aff. at par. 54, J.A. 2055-56. The evidence assembled by Professor Fisher shows that none of the anti-competitive consequences hypothesized by the district court have come to pass with GTE or other independent telephone companies, despite their entry into the information services market. Fisher Aff., pars. 53-63, J.A. at 2055-61.
Bafflingly, the district court characterized the BOCs’ evidence on the bottleneck issue as “half-hearted”, Remand Decision, 767 F.Supp. at 316, with barely a mention of the analyses we have just described. The affidavits in the record offer persuasive evidence that, despite their local monopoly power, the [276]*276BOCs will be unable to discriminate against competing information service providers.
The district court also alluded to a cross-subsidy theory that was by no means central to appellants’ objections — the possibility that BOCs might discourage others’ entry by “acquiring a reputation for strategic predatory pricing” (by responding to entry by lowering prices), financing the predation by shifting costs to its price-regulated monopoly operations in the local loop. See Remand Decision, 767 F.Supp. at 315 (citing Hall Aff., pp. 4-5, J.A. 2155-56). The Hall Affidavit is brief and conclusory on the point. Strikingly, one of the appellants’ affiants parried a BOC affidavit on predatory pricing by saying that predation was a “straw man” and that he was “inclined to agree” with the proposition that “the risk of classic predatory pricing alone” did not justify retention of the ban. Shapiro Aff., par. 62, J.A. at 2669. As we shall see, the pro-BOC affiants addressed the cost-shifting hazard in their treatment of regulatory constraints, to which we now turn.
(2) Efficacy of regulation. The BOCs and the Department of Justice pointed to regulatory safeguards that could limit anticompetitive effects even if the BOCs otherwise had the incentive and ability to discriminate or cross-subsidize. The district court dismissed these contentions. It believed instead that regulators had never been successful in stopping the Bell system from engaging in anticompetitive activities, and that there were no changes tending to enhance their ability to do so. It disposed of references to current FCC rules that require the BOCs to charge themselves the same access rates as they charge others, for instance, with the observation that a comparable rule did not work before the Bell system break-up. Remand Decision, 767 F.Supp. at 317-18. Indeed, the court believed that the increasing complexity and diversity of the industry, including BOC entry into other fields, might make it harder than ever to detect or stamp out discrimination and cross-subsidization. See generally id at 317-22.
There is a lot of evidence that the break-up and other recent developments have enhanced regulatory capability. The seven independent BOCs are not the old AT & T. As this court noted in the Triennial Review Opinion, the existence of seven BOCs increases the number of benchmarks that can be used by regulators to detect discriminatory pricing. See id, 900 F.2d at 299. Indeed, federal and state regulators have in fact used such benchmarks in evaluating compliance with equal access requirements (i.e., the requirements disparaged by the district court) and in comparing installation and maintenance practices for customer premises equipment. See Arrow/Rosenfield Aff., par. 43, J.A. 1764; Grossman Aff., par. 28, J.A. 2146; Carlton/Stigler Aff., par. 44-45, J.A. 1863-64. Moreover, information service giants operating throughout the country, such as IBM, AT & T and GE, will notice any discrepancies in treatment by the various BOCs and will have the capacity and incentive to bring anticompetitive conduct to the attention of regulatory agencies. Id. at par. 47, J.A. 1864-65. The BOCs themselves will be each other’s customers in the transmission of information services, with a unique capacity to spot misbehavior and notify regulators. Id. at par. 48, J.A. 1865. Finally, any effort to establish a reputation for predatory responses to new entry will require that potential new entrants be aware of such predation; a BOC may have great difficulty establishing such a reputation among potential entrants while keeping it a secret from regulators. Carlton/Stigler Aff., par. 62, J.A. 1872.
Changes independent of the AT & T break-up also support the likelihood of more effective regulation. For example, Professor Kahn pointed to the FCC move in the direction of price cap regulation, which the FCC has in fact extended to the BOCs since his affidavit. See National Rural Telecom Ass’n v. FCC, 988 F.2d 174, 179 (D.C.Cir.1993). This reduces any BOC’s ability to shift costs from unregulated to regulated activities, because the increase in costs for the regulated activity does not automatically cause an increase in the legal rate ceiling. Kahn Aff., par. 16, J.A. at 2393-94; Hausman Aff., pars. 32-33, J.A. 2260-61; National Rural Telecom Ass’n, slip op. at 8-9. In addition, the FCC has acted since the break-up to tighten its accounting rules, especially its treatment of joint costs, all tending to in[277]*277crease the chances of catching any attempts at cost-shifting. Farmer Aff., J.A. 1994— 2017.
Similarly, Kahn argued that it was by no means valid to assume the prevalence of regulatory “capture”. Noting state agencies’ overwhelming concern for keeping the rates for local residential service low, he reasoned that this would inhibit any effort by the BOCs to shift costs into those (monopoly) services. He also noted that nearly every state allowed, competitive resale of interLATA service — a rule clearly inconvenient to the BOCs and incompatible with an assumption of successful capture. Kahn Aff., pars. 10-11, J.A. 2391. Given Professor Kahn’s careful analysis, not to mention his distinguished experience as a regulatory economist, it would take far more than a reference to the district judge’s own experience in the AT & T matter to vindicate his disparagement of Kahn’s testimony as “completely in error”. See Remand Decision, 767 F.Supp. at 318 n. 49.
Of course, proponents of the entry ban submitted affidavits in its favor. The district court attached preeminent importance to that of Professor Noll, who argued that increasing product diversification would make strategic cost accounting and discriminatory practices hard to detect. Noll Aff., pars. 74-75, J.A. 2597-98; see Remand Decision, 767 F.Supp. at 320. The Noll affidavit is striking in that it focuses more on what BOCs might want to do than on their actual capacity for achieving it, and it does not truly address the post-break-up environment or developments such as price caps and enhanced FCC accounting rules, see Noll Aff., par. 72, J.A. 2597 (discounting these as either antedating the break-up or not yet adopted).
It is not, however, for us to choose among the opposing positions of distinguished economists. The issue is whether the Department of Justice’s position had substantial factual support and was grounded in reasonable analysis. On the proposition that regulation would be a substantial obstacle to anti-competitive behavior, the Department passed this test.
(3) Competitive benefits of BOC entry. Parties favoring removal of the ban pointed to substantial benefits that they believed likely to follow. For example, they pointed to several segments of the information services market that are highly concentrated (e.g., alarm services, financial transactions, and airline reservations), arguing that BOC entry would make the market more competitive. Arrow/Rosenfield Aff., par. 23, J.A. 1753. Further, there is reason to believe that BOCs entering the information services market will have some economies of scope (the capacity to produce related goods or services at an aggregate cost lower than the total for each produced separately), with resulting consumer benefits. See Carlton/Stigler Aff., pars. 17-22, J.A. at 1846-50.
The district court brushed these contentions aside as “preposterous”, Remand Decision, 767 F.Supp. at 326, partly on the theory that BOC competition could contribute no consumer benefit because the BOCs had no prior experience in the content or substance of information services. Id. This is a curious idea, as businesses are capable of acquiring expertise and experienced talent. Under the court's view, new entry could rarely create competitive benefits; even a development so close to home as recent entries into long-distance telephone service could not be expected to generate consumer benefits.
Instead of consumer benefit, the court envisioned a nightmare in which the BOCs would drive out the current competitors and perhaps “extinguish[ ]” competition altogether. Remand Decision, 767 F.Supp. at 326. The latter scenario is of course internally contradictory; because there are seven BOCs, competition would exist even if they drove out all the incumbents. More generally, the district court’s scenario necessarily depends upon successful use of some combination of discrimination and cross-subsidization; we have already reviewed the obstacles that such efforts would confront.
It is worth noting here the character of some of the firms that the BOCs would have to drive out. They include GE (with annual revenues about five times those of a BOC), AT & T itself (revenues three times those of a BOC), IBM and Sears with their Prodigy service, Merrill-Lynch, ITT, Mead Corpora[278]*278tion, American Express, Citicorp, Chase Manhattan Bank, and a variety of foreign and independent telephone companies. Hausman aff., pars. 11-14, J.A. 2225-26. These firms are not pushovers. Also, we have already noted evidence that local exchange costs represent only a tiny share of information service providers’ total costs, so that the BOCs’ bottlenecks do not, in perspective, seem like invincible weapons.
If appellants are right, BOC entry will raise prices and reduce output in the information services industry; if the Department is right, BOC entry will lower prices and expand output. Firms that sell goods and services that are inputs to the production and use of information services stand to gain an expanding market if the Department’s prediction is right, and have the incentive to make a completely unbiased judgment on the matter. It is something of a vindication of the Department’s view that representatives of those firms, the Computer and Communications Industry Association and the United States Telephone Association (which includes about 1000 independent local telephone companies), filed briefs with the district court supporting removal of the ban. See CCIA Brief 17-18, J.A. 78081; USTA Brief 1, 5-7, J.A. 784-87.
We stress again that the judicial role here is not to determine whether removal of the information services ban is an optimizing move. Under the arrangements made by Congress, that task (in the present context) is for the Department of Justice, subject to judicial review. The distinguished experts marshalled by the appellants may, in the eyes of an omniscient being, be “right”. Conversely, the fact that removal of the ban is supported by an array of prominent economists (including two Nobel laureates, Stigler and Arrow) does not in itself make that action right. The quality of their presentations is enough, however, to establish an ample factual foundation for the judgment call made by the Department of Justice and to make its conclusion reasonable. Insofar as the district court may be considered to have found the contrary, the finding was clearly erroneous.
We affirm the judgment removing the information services line of business restriction from the Modified Final Judgment. •
So ordered.