WILLIAM A. NORRIS, Circuit Judge:
Petitioners invoke our jurisdiction under 28 U.S.C. § 2342(1) and 47 U.S.C. § 402(a) to review orders of the Federal Communications Commission issued in a rulemaking proceeding known in the telecommunications industry as the Third Computer Inquiry or Computer III.1 Petitioners dispute two discrete rulings by the Commission. First, petitioners challenge the ruling that the divested Bell Operating Companies (BOCs) no longer be required to maintain corporate separation between their common carrier communications services, which are regulated under tariff pursuant to Title II of the Federal Communications Act,2 and the unregulated provision of enhanced or data processing services over the telecommunications network.3 Petitioners contend that the Commission’s decision to permit the BOCs to integrate their regulated and unregulated activities, violates section 10(e) of the Administrative Procedure Act, 5 U.S.C. § 706(2)(A), because it is arbitrary and capricious.
Second, petitioners challenge the Commission’s decision to preempt state regulation of communications common carriers’ provision of enhanced services on the ground that it violates section 2(b)(1) of the Communications Act, which denies the Commission jurisdiction over “charges, classifications, practices, services, facilities, or regulations for or in connection with intrastate communication service by wire or radio of any carrier_” 47 U.S.C. § 152(b)(1).
I
BACKGROUND
By the 1960s, the growing interdependence of the telephone and the computer created regulatory problems for the Federal Communications Commission. Increasingly, providers of data processing services, such as IBM, were using the transmission facilities of communications common carriers to deliver computer-based information to customers’ terminals. An entire new industry based upon the rapidly advancing technology of both the telephone and the computer took root and flourished. Because the provision of enhanced services — as the FCC has labeled telecommunications services combining both data process[1224]*1224ing and communications components4 — involves the use of the transmission facilities of communications common carriers, some argued that the Commission was statutorily required to regulate the new industry. The Commission rejected that argument in favor of a regulatory policy of promoting competition in the enhanced services industry notwithstanding the fact that enhanced services contained regulated communications as well as an unregulated data processing component.
The decision to foster a competitive enhanced services industry did not fully resolve the Commission’s regulatory problems. The Commission remained concerned that communications carriers would use their telephone exchange monopolies to obtain leverage in the competitive enhanced services market. First, because non-carrier providers of enhanced services were required to obtain access to the telecommunications network through local exchange “bottlenecks,”5 the Commission was concerned that carriers would gain an unfair competitive edge by discriminating in favor of their own enhanced service offerings in providing access. Second, the Commission was concerned that carriers would exploit their exchange monopolies by passing on to telephone ratepayers costs properly attributable to their unregulated enhanced services business. Such improper cost-shifting effectively subsidizes a carrier’s unregulated activities with monopoly profits from its regulated activities, to the detriment of both its monopoly ratepayers and its competitors in the enhanced services market. The two Computer III orders we review in this case represent a culmination of a long history of FCC attempts to guard against these potential abuses of communications carriers’ monopoly power.
A. The First and Second Computer Inquiries
In the First Computer Inquiry (Computer I),6 the FCC considered for the first time the appropriate regulatory treatment of telephone company participation in the newly emerging, competitive industry of delivering data processing services over telephone lines. The Commission required that any telephone carrier offering such enhanced services do so by means of a separate corporate subsidiary. The structural separation requirement was applied to all carriers with annual revenues exceeding $1,000,000, although the Commission never explained why it drew its regulatory line at the $1,000,000 mark. Structural separation was not initially regarded as applying to the American Telephone & Telegraph Company (AT & T) and its local exchange affiliates (the Bell System), because those companies were thought to be barred from offering data processing services by a 1956 antitrust consent decree. United States v. Western Elec. Co., 1956 Trade Cas. (CCH) H 68,246 (D.N.J. Jan. 24, 1956); see GTE Serv. Corp., 474 F.2d at 730 n. 7.
In its 1980 Second Computer Inquiry ('Computer II) decision,7 the FCC redefined regulated communications services and unregulated data processing. By creating a regulatory distinction between “basic” and “enhanced” services, the Commission sought to draw a bright line between activities that would be regulated as common carrier offerings and those that would not. See Computer II Final Decision, 77 [1225]*1225F.C.C.2d at 423, 428-30, 438-47. In Computer II, the FCC also preempted state regulation of the sale of both customer premises equipment (CPE)8 and enhanced services.
Although the FCC in Computer II continued to rely on structural separation as the principal means of preventing cross-subsidization and discriminatory access, it restricted the requirement to members of the Bell System and removed it from all other carriers regardless of whether their revenues exceeded $1,000,000.9 Thus only AT & T and its operating subsidiaries were required to form separate corporate subsidiaries to provide enhanced services.10 The Commission’s decision to switch from regulating all carriers with revenues in excess of $1,000,000 to regulating only AT & T apparently went unchallenged. The FCC attempted to predicate the need for regulation on a carrier’s national market power, which only AT & T had. However, the Commission neither defined the “national market” at issue nor explained why it drew its regulatory line between national market power and substantial regional market power such as that enjoyed by GTE Service Corporation (GTE) and Continental Telephone. The FCC again stressed the impor-tanee of a carrier’s monopoly control of local bottleneck facilities and the ability to abuse that control either by providing inferior access to competitors or by cross-subsidizing its own enhanced services with monopoly revenues derived from captive ratepayers. Computer II Final Decision, 77 F.C.C.2d at 466-68.-
B. The Breakup of AT & T
On January 1, 1984, AT & T divested itself of its 22 local exchange telephone companies — the Bell Operating Companies or “BOCs” — as part of a settlement of the government’s 1974 antitrust suit against AT & T. United States v. AT & T, 552 F.Supp. 131 (D.D.C.1982), aff'd, 460 U.S. 1001, 103 S.Ct. 1240, 75 L.Ed.2d 472 (1983); see also United States v. Western Elec. Co., 569 F.Supp. 990 (D.D.C.1983); United States v. Western Elec. Co., 569 F.Supp. 1057 (D.D.C.) aff'd, 464 U.S. 1013, 104 S.Ct. 542, 78 L.Ed.2d 719 (1983). Upon divestiture, the 22 BOCs were grouped into seven independent regional holding companies.11 The district court retained jurisdiction to approve implementation of the MFJ,12 to enforce its terms, and to entertain future [1226]*1226requests for removal or modification of the decree’s restrictions.
Under the MFJ, AT & T retained its long-distance telephone operations, its CPE manufacturing business, and Bell Laboratories. AT & T, now divested of the BOCs, was free to enter virtually all other facets of the marketplace, including data processing. The divested BOCs were restricted generally to providing local exchange telephone service and other “natural monopoly service[s] actually regulated by tariff.” 552 F.Supp. at 227-28. Under the MFJ, the BOCs could not provide inter-exchange telephone service, or offer “information services,” a category that substantially overlaps, but may not be identical to, the FCC’s “enhanced services.”13 Id. at 227.
The antitrust court initially left open the question whether the BOCs should conduct their unregulated activities through separate corporate subsidiaries. In accepting the MFJ, the court found that issue more appropriately a concern of regulatory agencies such as the FCC. 552 F.Supp. at 193 n. 251. In a subsequent order, the antitrust court imposed corporate separation as a condition on the grant of waivers that permitted a company to engage in non-communications-related activities. United States v. Western Elec. Co., 592 F.Supp. 846, 870-71 (D.D.C.1984), appeal dismissed, 777 F.2d 23 (D.C.Cir.1985). In 1988, the antitrust court permitted the BOCs to offer voice messaging services, voice storage and retrieval and electronic mail services — all of which are enhanced services under the FCC’s definitions. United States v. Western Elec. Co., 714 F.Supp. 1, 22 (D.D.C.1988), aff'd in part, rev’d on other grounds, 900 F.2d 283 (D.C. Cir.1990) (per curiam).
C. The BOC Separation Order
In preparation for the January 1984 restructuring of the Bell System, the FCC opened a proceeding to determine whether and how the Computer II rules should be applied to the divested BOCs.14 In its BOC Separation Order, the FCC stressed that the BOCs enjoyed the same monopoly power through their ownership of the local exchanges that AT & T had long enjoyed. The Commission noted that the seven regional BOCs, although smaller than AT & T, remained titans in the telecommunications industry.15 The divestiture of AT & T did not eliminate the Bell System’s control over local exchange facilities; it simply transferred that control from a single national company to seven regional BOC holding companies. Throughout its order, the FCC insisted that the BOCs’ monopoly power and control over the local exchange bottlenecks would give them both the incentive and the ability to cross-subsidize and to provide inferior network access to enhanced services competitors. BOC Separation Order, 95 F.C.C.2d at 1125-39.
[1227]*1227In the BOC Separation Order rulemak-ing proceeding, the BOCs argued that divestiture had eliminated the need for Computer II’s structural separation requirements. The FCC rejected this argument because the BOCs had inherited AT & T’s monopoly power and control over access to the telecommunications network. In the Commission’s view, the critical fact was that each BOC would retain ownership of the local exchange monopolies in the region in which it operated. As the Seventh Circuit observed in approving the FCC’s order:
It is still the case, no less than before the divestiture, that if you want to have telephone service and are located in an area served by a Bell operating company you must get your access line from that company. Hence the operating companies have monopolies of basic telephone service in many important markets....
... The basic source of AT & T’s monopoly power was not the manufacture of telecommunications equipment or even the ownership of the nation’s long-distance lines; it was the operating companies’ control of access to the telecommunications network. The inheritors of the Bell monopoly are the [BOCs].... The worst bottlenecks in this industry are local.
Illinois Bell, 740 F.2d at 471, 473. Such bottleneck control, in the Commission’s view, created the same danger of anticom-petitive activity that existed before divestiture. BOC Separation Order, 95 F.C.C.2d at 1134-35.
The FCC also determined in its order that non-structural regulations on BOC cost-accounting and network-access practices would be ineffective safeguards against anticompetitive behavior by the BOCs. Absent a structural separation rule, the FCC found, the BOCs could install their own enhanced services equipment within the local networks, and would be free to market enhanced services through the same organizations used for basic telephone service. For example, the FCC found that
if joint marketing were permitted, the personnel who contact customers to market regulated services could be selling ... enhanced services at the same time. Difficult, sometimes impossible, problems of fairly allocating the costs of marketing between regulated and unregulated accounts would again occur.
BOC Separation Order, 95 F.C.C.2d at 1130. This integrated operation of regulated and unregulated activities would, in the FCC’s words, be “impossible” to monitor for cross-subsidization. Id. The Commission found that accounting regulations were insufficient to ensure that the BOCs were properly allocating the costs of providing each service in the charges passed on to consumers. Because the services were so closely related, the Commission reasoned, the BOCs could easily reclassify enhanced services costs or joint costs as basic telephone service costs, without being detected by the FCC. See BOC Separation Order, 95 F.C.C.2d at 1130-31. For the same reason, the Commission rejected the BOCs’ argument that state regulatory commissions, which set rates for intrastate basic telephone services, could effectively prevent improper cost-shifting to ratepayers. Because state ratemaking, like FCC ratemaking, is based on the cost of providing service, state regulators would encounter the same difficulties as the FCC in determining whether a BOC was properly allocating costs between enhanced services and basic telephone service. See id. at 1130-31.
Analogous enforcement problems, in the FCC’s view, would beset efforts to use network-access rules to assure that the BOCs would not provide inferior access to competitors. Because of the BOCs’ exclusive control over the connection and switching equipment of the local exchanges, it would be difficult for the Commission to know whether the BOCs were providing competitors with transmission services of comparable quality at comparable prices. By integrating their enhanced services equipment into the telephone network, the BOCs could favor their own enhanced services with superior signals and switching capacities without being detected. The BOCs could also engage in subtle tie-in practices in which they conditioned the [1228]*1228availability of transmission services on the purchase of their own enhanced services or equipment. See id. at 1135.
In contrast, the FCC found that structural separation continued to be an effective means of preventing cost-shifting and discrimination. It viewed structural separation as essentially a prophylactic measure. The separate subsidiary requirement forces the BOCs to produce and market enhanced services independent of basic telephone services, and maintain different books, different staffs, and different equipment premises for each service. This separation eliminates the problem of determining the proper allocation of joint costs, and makes it difficult for the BOCs to masquerade enhanced services costs as basic telephone service costs. BOC Separation Order, 95 F.C.C.2d at 1131. It also reduces the chance of discrimination by putting a BOC’s enhanced services subsidiary on substantially the same footing as the subsidiary’s competitors. Like its competitors, the subsidiary must deal with the BOC at arm’s length in purchasing transmission services, and must maintain its enhanced services equipment at a location separate from the BOC’s local exchange facilities. The subsidiary thus receives its transmission services along the same telephone lines, and through the same switching systems, as its competitors. In the FCC’s words, “[i]f a BOC’s separate entity is required to obtain access to the network in the same fashion as would a competing supplier, the provision of inferior access to a BOC rival would be much easier to detect.” Id. at 1136.
In mandating structural separation, the Commission made clear that its regulations were designed to protect the integrity of two distinct markets — the unregulated market for enhanced services and the regulated market for basic telephone service. In part, the separation requirements were designed to benefit participants in the enhanced services market by creating an even playing field for the BOCs’ competitors. If the BOCs were prevented from either cross-subsidizing their own enhanced services or discriminating against the enhanced service offerings of their competitors, then free and fair competition would flourish in the growing industry. BOC Separation Order, 95 F.C.C.2d at 1132-37. Structural separation also protected the basic telephone service market by preventing the BOCs from extracting, through cost-shifting, monopoly rents from captive telephone consumers. The “ratepayers” — that is, ordinary telephone customers — would not be forced to subsidize the BOCs’ expansion into the data processing business. See id. at 1129.
D. The Third Computer Inquiry
In August 1985, just 14 months after concluding the BOC Separation Proceeding, in which the FCC rejected the argument that divestiture had reduced the need for structural separation, the Commission reversed course16 and announced its intention to relieve the BOCs of the separation requirements.17 The FCC now reasoned that divestiture and increased competition in the enhanced services market had somehow diminished the value of structural separation as a safeguard against monopoly abuse. The FCC and the dissent attempt to justify this about-face by arguing that divestiture has drastically changed the [1229]*1229competitive structure of the telecommunications industry. Some aspects of the industry remain the same, however. For example, the FCC noted in Computer III that it could not conclude “that the BOCs’ motivation to cross-subsidize ha[d] decreased because of these developments.” Phase I Order, 104 F.C.C.2d at 1010. Furthermore, the FCC provided no record support that any of the so-called changes were relevant to its regulatory task of protecting captive ratepayers and competitors against the damaging effects of cross-subsidization. Nevertheless, the FCC tentatively concluded that the costs of separation now exceeded its public benefits and proposed to replace the requirement with accounting and other nonstructural regulations.18
In a Report and Order issued June 16, 1986 {Phase I Order), the FCC adopted new regulations permitting the BOCs to integrate their basic and enhanced services upon implementation of a plan of nonstructural safeguards approved by the FCC. Phase I Order, 104 F.C.C.2d at 1013. The required nonstructural safeguards are described in more detail below.
The FCC decided that its structural separation regulations had imposed costs in terms of the unavailability of certain services, lost economies and efficiencies, and the inability of customers to obtain complete telecommunications and data processing solutions from a single vendor. The FCC also determined that the BOCs’ ability to cross-subsidize had been restricted because of divestiture, the growth of competitive alternatives to the BOCs’ ordinary telephone service, and political and regulatory pressures at the state level to keep local phone rates down. The BOCs’ ability to discriminate by providing inferior network access had also diminished, according to the FCC, because of industry-wide coordination of network standards and the threat that enhanced services competitors could bypass the BOCs’ local exchanges. Phase I Order, 104 F.C.C.2d at 1007-12. However, the FCC did not address the question of whether divestiture had resulted in a diminution of the power of the surviving regional BOCs to subsidize their unregulated enhanced service operations by shifting costs to monopoly ratepayers.
The FCC concluded that the regulatory goals served by structural separation could be achieved by two nonstructural safeguards.19 First, the FCC would develop cost allocation methods to minimize the BOCs’ ability to shift costs from their unregulated to regulated activities. Id. at 1010-11. Second, the FCC adopted regulations specifically designed to prevent the BOCs from using their “substantial market power in providing network access” to discriminate against competing providers of enhanced services. Phase I Order, 104 F.C.C.2d at 1026. These antidiscrimination regulations contained three prongs. First, the FCC endorsed an open-network policy of requiring the BOCs to make the telephone networks as accessible to competitors as they are to the BOCs themselves.20 Second, each BOC must notify its competitors in the enhanced services industry of changes in the network that may affect the provision of enhanced services so as to permit competitors to take advantage of the changes. Third, each BOC must pro[1230]*1230vide its competitors with information about customer use of the telephone network so that the competitors may design their services to suit customer needs.
II
THE APA ISSUE
Petitioners claim that it was irrational for the FCC to abandon structural safeguards so soon after imposing them on AT & T in Computer II and reimposing them on the divested BOCs in the BOC Separation Order. Structural safeguards, argue petitioners, have been the Commission’s primary weapon for preventing anticompet-itive conduct such as cross-subsidization and discriminatory access, and these safeguards, which were originally imposed in Computer I on all communications carriers with assets of over $1,000,000, are a longstanding FCC rule which was abandoned without a reasoned agency decision.
The FCC counters by characterizing structural safeguards as a transitional measure imposed on the BOCs to cushion the effects of AT & T’s divestiture. FCC Brief at 97 n. 148. Furthermore, the Commission argues that it announced at the time of the BOC Separation Order that it would refine its separation requirements as it gained experience with BOC offerings. Id. (citing BOC Separation Order, 95 F.C. C.2d at 1140, 1127-28). The FCC vigorously disputes petitioners’ claim that it has abandoned any past policy and insists that it is required to reevaluate its policies in light of changed circumstances. WWHT, Inc. v. FCC, 656 F.2d 807, 819 (D.C.Cir.1981); Geller v. FCC, 610 F.2d 973, 980 n. 59 (D.C.Cir.1979).
We recognize that the FCC is obligated to reevaluate its policies when circumstances affecting its rulemaking proceedings change. We also recognize that the Commission has broad discretionary authority to change its regulatory mind. But the Commission cannot expect us simply to rubberstamp its change in policy. Although we respect the Commission’s exercise of discretion in an area in which it has expertise, we cannot, as a reviewing court, accept an agency’s change of course uncritically. Section 10(e) of the APA requires us to set aside agency action if it is “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 5 U.S.C. § 706(2)(A).
The FCC explains its lifting of the structural separation requirements by arguing that certain developments in telecommunications markets and technologies have materially changed circumstances in the industry. Specifically, the FCC asserts that technological advances, political and regulatory pressure by the states, divestiture, and a generally competitive enhanced services market have reduced the risks of both cross-subsidization and discriminatory access.
Section 10(e) of the APA requires us to determine whether the Commission’s decision was a reasonable exercise of its discretion, based on consideration of relevant factors, and supported by the record. See, e.g., Motor Vehicle Mfrs. Ass’n v. State Farm Mutual Auto Ins. Co., 463 U.S. 29, 43-44, 103 S.Ct. 2856, 2866-67, 77 L.Ed.2d 443 (1983) (State Farm). The scope of review under this standard is a narrow one: we are not to substitute our judgment for that of the Commission in making policy choices as to how to carry out its statutory mission. While our standard of judicial review is highly deferential, it may not be uncritical. Under the APA, an agency’s discretion is not boundless, and we must satisfy ourselves that the agency examined the relevant data and articulated a satisfactory explanation for its action based upon the record. We must find agency action to be in violation of the APA if the agency has “failed to consider an important aspect of the problem” or has “offered an explanation for its decision that runs counter to the evidence before the agency.” Id. at 43, 103 S.Ct. at 2866; see also Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 416, 91 S.Ct. 814, 823, 28 L.Ed.2d 136 (1971).
We stress that petitioners’ contention that the FCC violated § 10(e) of the APA when it allowed the BOCs to integrate their regulated and unregulated ac[1231]*1231tivities requires judicial scrutiny of the cost/benefit analysis that is at the heart of the FCC decision. Especially in a rulemak-ing proceeding as economically and technologically complex as Computer III, this scrutiny is a daunting task. Nonetheless, because the FCC relied on a cost/benefit analysis, we must decide whether the FCC’s evaluation of the costs and benefits it has identified as relevant finds support in the administrative record. While we scrutinize the FCC’s evaluation of each individual cost and benefit, we make no attempt to exercise the judgment required in striking the balance between the competing costs and benefits. That balancing is a matter that must be left to the discretion of the FCC. But if the FCC’s evaluation of any significant element in the cost/benefit analysis lacks record support, we cannot uphold the agency action under § 10(e). Thus, we must be satisfied that the Commission’s assessment of the various costs and benefits is reasonable in light of the administrative record.
We note further that we are limited in our evaluation to the reasons the FCC gave; we cannot supply our own rationale. According to the Commission, technological advances and market changes have made it possible for nonstructural safeguards, such as cost accounting, to protect ratepayers from cost-shifting and competitors from both cost-shifting and inferior access. We thus address the question of whether the record supports the FCC’s finding that the individual costs and benefits of structural separation have been materially affected by changed circumstances.
A. The Costs of Structural Separation
Throughout the various Computer rulemaking proceedings, the FCC has recognized that the protective benefits of structural separation do not come free of cost. See, e.g., BOC Separation Order, 95 F.C.C.2d at 1137-38. In appraising the social costs of structural separation, the FCC found in Computer III, as it had found in prior proceedings, that separation has discouraged innovation in developing and marketing new enhanced services technologies, has prevented the BOCs from providing customers with efficient packages of basic and enhanced services, and has generally created inefficiencies by forcing the BOCs to maintain duplicate organizations and facilities. The FCC made no finding that the costs of these inefficiencies were any greater at the time of Computer III than they were at the time of Computer II and the BOC Separation Order. Nor do we see any basis in the record for such a finding. Nonetheless, the FCC contends that its experience under the Computer II regime has provided it with more evidence that these social costs are significant. Even though the record support is marginal, we hold that it suffices to support an administrative finding that the BOCs have had to raise the price of certain services, withhold others from the market, and perhaps never develop still others because of the inefficiencies of structural separation.
The Commission makes a fair case that the integration of regulated and unregulated telecommunications services by the BOCs would yield some economies of scale. On the technological side, the BOCs could integrate enhanced services equipment more closely into the telephone network. On the marketing side, the BOCs could offer customers — particularly business customers — packages of basic and enhanced services. These packages would minimize transaction costs, and reduce delays and coordination problems in satisfying consumer demands. Moreover, the BOCs would not have to maintain, and customers would not have to pay for, separate organizations and facilities for operating and repairing basic and enhanced services. See Phase I Order, 104 F.C.C.2d at 987-89.
While the Commission concludes that these costs are “potentially very significant,” it makes no attempt to quantify them. Id. at 1008. The Commission asserts that it has had several years since Computer II in which to study the costs of structural separation. We recognize that the Commission now enjoys the benefit of additional experience with structural separation; however, the Commission makes no showing that this experience proves that the costs of separation have increased. In [1232]*1232the absence of concrete data, the Commission has relied on vague generalizations offered by the BOCs, see id. at 987-89, which makes us hesitate to give great weight to the FCC’s “potentially very significant” assessment. Nonetheless, the BOCs’ submissions provide an adequate, if marginal, basis for concluding that experience shows that some economies of scale would result from integrating basic and enhanced operations.
The FCC makes a plausible case that its separation requirements have impeded the marketing and even development of some new forms of enhanced services. The Commission invokes the example of the Custom Calling II VMS (voice message system), which AT & T developed in the early 1980s. AT & T unsuccessfully sought a waiver from the Computer II structural separation rules so it could market the system through its local exchange operating companies. The result, according to the Commission, is that neither this system nor any substantial equivalent has been available to the public. Phase I Order, 104 F.C.C.2d at 1008. The record contains other examples of this problem, although the examples are neither detailed nor systematic. The Computer III record is replete with BOC assertions that because their ownership of the local exchanges gives them control of access to the telecommunications network, they are best situated to develop and market products that take full advantage of the network’s capabilities. Yet the only concrete example in the record is the voice message system cited by AT & T.
Although we are concerned that the rule-making record fails to indicate the magnitude of the costs of structural separation, we recognize that the FCC necessarily must speculate about the extent to which releasing the BOCs from the handicap of separate operations will create a more vigorous enhanced services market. Because the BOCs have never had an opportunity to offer enhanced services in a deregulated environment, the real benefits, if any, of integrated operations may not be quantifiable at this time. We therefore conclude that the FCC could reasonably decide on this record that structural separation impairs the BOCs’ ability to achieve certain economies of scale and has impeded the development and marketing of some new forms of enhanced services. The record, however, yields no evidence that the costs of separation have increased as a result of changed circumstances following Computer II and the BOC Separation Order. The regulatory experience has merely confirmed that these costs are “potentially very significant.” Phase I Order, 104 F.C. C.2d at 1008.
B. The Benefits of Structural Separation
The FCC has consistently identified the social benefits of structural separation as safeguarding against the twin evils of discriminatory access and cross-subsidization. Guarding against these improper and anti-competitive practices has always been the Commission’s chief concern in promoting its goal of protecting both captive ratepayers and enhanced services competitors of communications carriers. In Computer III, the Commission found that changed technological and market circumstances had reduced the need for structural separation as a safeguard against access discrimination and cross-subsidization of the BOCs’ own enhanced services with monopoly revenues derived from ratepayers. The FCC cites as changed circumstances a more competitive enhanced services market, technological advances that include an increased ability to detect discriminatory access and the growth of opportunities to bypass local exchanges, increased state political and regulatory pressures, and the AT & T divestiture of the Bell Operating Companies. We therefore consider whether the record supports the FCC’s conclusion that these changed circumstances have reduced the value of structural separation, thereby justifying the substitution of nonstructural for structural safeguards.
(1) Prevention of Discriminatory Access
The FCC found that technological and market developments in the last few years have reduced the danger that the BOCs will discriminate against their competitors [1233]*1233in providing access to the telephone network. We agree that the Commission could reasonably decide that changed circumstances have made it more difficult for the BOCs to provide competitors with inferior access.
The FCC could reasonably conclude that the emergence of powerful competitors such as IBM, which have the resources and expertise to monitor the quality of access to the network, reduces the BOCs’ ability to discriminate in providing access to their competitors. The record also supports the proposition that these large corporate competitors — unlike the average telephone user — have a growing capability of bypassing local exchanges by using microwave systems or cable networks for linking their computers with their customers’ terminals.
Moreover, the network-access policies that the FCC proposes to substitute for structural separation involve new technologies for detecting inferior access that have not previously been available. The FCC’s policy of Comparably Efficient Interconnection (CEI) is designed to ensure that each BOC will provide competitors with connections to the local exchange that equal the connections available to the BOCs’ own enhanced services offerings. Moreover, the Open Network Architecture (ONA) policy requires each BOC to incorporate CEI concepts into the overall design of its basic service network. Thus, the record supports the Commission’s finding that technologies for ensuring equal access have improved, and may be effective in preventing discrimination in ways not feasible in the past.
(2) Prevention of Cross-Subsidization
We cannot agree, however, that there is adequate record support for the Commission’s conclusion that changed circumstances have reduced the danger that the BOCs will effectively subsidize their competitive activities with monopoly revenues improperly derived from captive ratepayers. The FCC points to four changes in the telecommunications market that allegedly have reduced this danger following Computer II and the BOC Separation Order: (1) the enhanced services market has become “extremely competitive”; (2) the growth of “bypass” and other alternatives to local exchange facilities has eroded the BOCs’ monopoly power; (3) political and regulatory forces at the state level exert pressure on the BOCs to keep the price of basic telephone service down; and (4) divestiture has dissipated the monolithic power of the Bell System. Given these changes, the FCC asserts that nonstructural safeguards such as accounting controls may be substituted for structural separation requirements without increasing the risk of cross-subsidization.
From our review of the administrative record, we hold that the evidence fails to support the FCC’s finding that changed market and technological circumstances have significantly diminished the risk of undetected cost-shifting by the BOCs. Nor do we find any record support for the FCC’s abandonment of its original position that cost-accounting regulations are relatively ineffective means of preventing improper cost-shifting. We note in this regard that the FCC mischaracterizes the record when it argues that its reliance on cost accounting to detect cost-shifting does not depart from past policy. FCC Brief at 104. The Commission concedes that it has “concluded [in past proceedings] that cost allocation alone would be insufficient to protect against cost-shifting.” Id. The FCC denies, however, that in Computer III it is relying solely on cost accounting regulations as a safeguard against cross-subsidization.
[The FCC] is relying on cost accounting as one element of its overall scheme, the efficacy of which must be viewed with other Computer III requirements. As the Justice Department noted in its comments, ‘regulatory mechanisms, that by themselves may be insufficient to prevent a regulated bottleneck monopolist from impeding competition in related markets, may, in conjunction with an open architecture network, significantly reduce the risk of anticompetitive harm.’
Id. at 105 (citation omitted). This argument is to no avail because it obscures the distinction between regulatory protections [1234]*1234against discriminatory access and regulatory protections against cost-shifting. Open Network Architecture, for example, does nothing to detect or deter cost-shifting; it only guards against inferior access. Thus, we must reject the Commission’s assertion that it did not depart from past policy in Computer III when it relied heavily on cost accounting regulations to detect cost-shifting. We can only conclude that this position taken in Computer III represents an unexplained change from the FCC’s original position. We agree with the FCC that it has broad discretion to implement reasonable policy changes. But as the Supreme Court has said, “ ‘[A]n agency’s view of what is in the public interest may change.... But an agency changing its course must supply a reasoned analysis.’ ” State Farm, 463 U.S. at 57, 103 S.Ct. at 2874 (quoting Greater Boston Television Corp, v. FCC, 444 F.2d 841, 852 (D.C.Cir.1970)).
The Commission does not explain how the fact that the enhanced services market is “extremely competitive” reduces the BOCs’ ability to cross-subsidize their unregulated enhanced services businesses by misallocating costs to their regulated activities. The BOCs’ ability to shift costs derives from the fact that they are monopoly providers of basic telephone service. Because competition in the unregulated enhanced services market does nothing to decrease the BOCs’ monopoly power in the basic services market, we fail to see how it can dimmish the BOCs’ ability to shift costs to their regulated services without detection in ratemaking proceedings. If anything, increased competition in the enhanced services market simply increases the BOCs’ incentive to shift costs so they can engage in predatory price-cutting as a means of maintaining or increasing their share of the market for enhanced services.
The Commission also asserts that the BOCs’ monopoly power over the local networks has been “eroded” by new technologies permitting some telephone users to bypass the BOCs’ local exchange facilities. These new bypass technologies include “Teleport” fiber networks, cable networks, and microwave systems, all of which enable some circumvention of the local exchanges. The record indicates that such services are available in some metropolitan areas to large business subscribers.
Nynex, for example, listed in considerable detail the bypass technologies currently being employed by major New York banks, and described how other corporate subscribers are interested in developing their own private networks. However, Nynex offered no evidence that these bypass opportunities were widely available to ordinary residential subscribers. Nynex did suggest that some large apartment and condominium buildings in certain metropolitan areas might purchase “shared tenant services” which would enable telephone users to avoid the BOCs’ local exchanges, but gave no indication as to the present existence or potential scope of such residential services. Nynex also suggested that corporations maintaining private networks might sell their excess capacity to smaller users who could not afford to purchase microwave systems; but again, it offered no evidence that any such “trickle down” effect actually existed, nor any evidence of how widely available it would be. Moreover, Nynex offered no evidence that the availability of bypass as a theoretical alternative to BOC local exchange service has in any way restricted the ability of the BOCs to inflate basic service prices and cut enhanced service prices through cost-shifting.
Nynex’s submission was fairly representative of other commenters’ submissions on the question of local bypass. Cincinnati Bell, for example, described how Procter & Gamble and General Electric have constructed their own private telecommunications networks, but offered no evidence of bypass opportunities for residential subscribers.21 Joint Appendix at 220-22 (Comments of Cincinnati Bell Telephone Compa[1235]*1235ny). Southwestern Bell extolled the fact that “customers can now install a point-to-point microwave system for less than $20,-000, excluding multiplexing equipment or tower construction costs,” but again offered no evidence that ordinary telephone users were able to bypass the BOCs’ local exchanges. Joint Appendix at 424-25 (Comments of Southwestern Bell Corporation). Finally, Ameritech described the range of fiber-optics and microwave networks available to occupants of several large office buildings in Chicago, but gave no examples of existing bypass opportunities for residential customers. Joint Appendix at 565-75 (Reply Comments of the Ameritech Operating Companies). None of these commenters submitted any evidence that bypass has significantly diluted the BOCs’ monopoly power in the local exchange markets.
In short, although the record contains an impressive array of evidence demonstrating the technical feasibility of bypass, the record contains no evidence that bypass has become a realistic option for any appreciable number of ordinary telephone users. More importantly, the record is devoid of evidence that the potential for bypass is significant enough to reduce the BOCs’ ability, as a practical matter, to exact monopoly rents from basic service customers by burdening them with costs from unregulated activities. On the basis of the record before us, therefore, the FCC’s contention that the availability of bypass will hamper BOC attempts at cost-shifting must be characterized as speculative at best. As the court administering the AT & T antitrust MFJ concluded in an opinion contemporaneous with the Phase I Order, “ ‘bottleneck’ monopolies continue to exist as before,” and “the ability to exploit the bottlenecks anticompetitively has remained precisely the same.” United States v. Western Elec. Co., 627 F.Supp. 1090, 1095 (D.D.C.) aff'd in part, rev’d on other grounds, 797 F.2d 1082 (D.C.Cir.1986), cert. denied, 480 U.S. 922, 107 S.Ct. 1384, 94 L.Ed.2d 698 (1987); see United States v. Western Elec. Co., 673 F.Supp. 525, 536-40 (D.D.C.1987) (specifically rejecting the contention that bypass has reduced the BOCs’ monopoly control over local exchange service) aff'd in part, rev’d on other grounds, 900 F.2d 283 (D.C.Cir.1990) (per curiam). Thus, the danger persists that the BOCs can effectively subsidize their unregulated enhanced services business by misallocat-ing costs to ordinary telephone service.22
We stress that the relevance of bypass to the issue of cross-subsidization lies in the potential for bypass to become so widespread that monopoly power over local exchange service is materially reduced. Nothing in the record supports such a conclusion. In other words, the record yields no evidence that the BOCs’ power to extract monopoly rents from local telephone customers has diminished. In fact, the FCC has never attempted to quantify even in the most broad and general way any such change. To provide anecdotes about specific bypass opportunities without discussing their impact on the BOCs’ ability to misallocate costs to ratepayers’ revenues confuses the issue. Once again, the Commission fails to distinguish the role of bypass in reducing the risk of discriminatory access from the role of bypass in reducing the risk of cross-subsidization. The record evidence about bypass goes only to a possible reduction in the risk of discriminatory access. The FCC cites to nothing and we know of nothing in the record that even [1236]*1236begins to show that bypass reduces the BOCs’ potential for cost-shifting in any significant way. It is the ability to shift costs to ratepayers that enables communications carriers to subsidize their unregulated activities. We therefore conclude that the evidence of increased bypass opportunities provides no support for a finding that the BOCs’ ability to cross-subsidize has been significantly reduced.
Nor does the record yield evidence of any reduction in the BOCs’ ability to increase monopoly profits by misallocating costs in seeking rate increases for basic telephone service. The FCC provides us with no basis for deciding when and whether a communications common carrier has the requisite monopoly revenues to engage in such improper cost shifting as a means of subsidizing its unregulated enhanced services business. As a reviewing court, we cannot judge for ourselves whether the BOCs generate sufficient revenues from their local exchange monopolies to pose a serious threat of harm to competition in the enhanced services market. We lack the capacity to make such a determination. Not only has the Commission failed to address the issue of the BOCs’ monopoly power, but its decisions in past proceedings about when monopoly revenues are of significant magnitude to merit regulation have been confusingly erratic. In Computer I, the FCC originally imposed structural separation requirements on all carriers with annual gross revenues exceeding $1,000,000. In Computer II, the Commission relieved all carriers of the separation requirements except one, AT & T.23 The conversion from a policy of including small carriers with only $1,000,000 in revenues to a policy of excluding all carriers but AT & T has gone both unexplained and unsupported. Frankly, we find no coherence in the Commission’s policy shifts in deciding which carriers to subject to separation requirements.24 The Commission’s attempt in Computer II to explain its decision to exempt all carriers save AT & T after originally exempting only carriers with revenues under $1,000,-000 lacks any persuasive power. The FCC simply recited that AT & T had “national market power” as though those words had talismanic significance. We are left with no explanation of why AT & T’s power in some undefined, perhaps regulated, perhaps unregulated “national market” provides a rational basis for drawing the regulatory line between AT & T and all other communications carriers. Why suddenly did carriers such as GTE, with revenues of $3,894,000,000, and United Telecommunications, with revenues of $1,084,956,000,25 lose their ability to cross-subsidize? Computer III, like Computer II before it, fails to answer this question. Because the FCC has never explained how “national market power” relates to the threat of cross-subsidization, we must conclude that we have no basis on which to discern whether the FCC’s current policy choice is rational.
We are also unpersuaded by the Commission’s new-found faith in Computer III that political and regulatory forces in the states will exert pressure on the BOCs to “minimize rural, residential, and small business local exchange rates, even to levels below cost,” thereby “limit[ing] the BOCs’ ability to shift costs to regulated [basic] services.” Phase I Order, 104 F.C.C.2d at 1010; Phase I Reconsideration, 2 FCC Red at 3039. As we note above, the BOCs made the same claim five years ago when the FCC originally imposed structural separation requirements on them. The FCC rejected the argument on the ground that given the difficulty of determining the proper allocation of costs for basic and enhanced services, state regulators had no effective means of ensuring that ordinary [1237]*1237telephone users were protected against cost-shifting. See BOC Separation Order, 95 F.C.C.2d at 1130-31. In Computer III, the FCC does an about-face, and insists that state regulation will now be more effective, but provides neither record support nor analysis for its new position. It offers no evidence that conditions at the state level have changed, or that it erred in the BOC Separation Order. Indeed, the Commission refers to nothing in the record; it simply relies upon the bald assertion that state regulators have become more effective in protecting ratepayers against improper cost-shifting.
Finally, the FCC states that “[djivesti-ture eliminated AT & T’s control over the most significant operations it could use for cross-subsidization — its former local exchange companies.” Phase I Order, 104 F.C.C.2d at 1005; see also Phase I Reconsideration, 2 F.C.C.Rcd at 3038. The Commission, however, ignores its finding in the BOC Separation Order that those local exchange companies and the monopoly revenues they generate, now in the hands of the BOCs, continue to provide significant opportunity and incentive for subsidizing unregulated activities to the detriment of both ratepayers and enhanced services competitors. The FCC has pointed to no change in circumstances which indicates that the danger of cross-subsidization by the BOCs has diminished since the BOC Separation Order. Rather than identifying changed circumstances, the Commission asserts that because it announced in the BOC Separation Order that it would eventually reevaluate the need for separation requirements, it has broad authority to do so. We do not dispute that the FCC has the authority to rethink its post-divestiture policies in light of changed circumstances. It must, however, provide reasoned explanations for its policy changes. Here, the Commission’s explanation is that circumstances have changed; yet, it fails to cite any changes since the BOC Separation Order that logically reduce the danger of cross-subsidization by the BOCs.
The only information in the voluminous Computer III record that even purports to explain the FCC’s decision is its statement that the divested BOCs now resemble GTE. But this so-called resemblance tells us nothing, for the FCC has lacked any consistency in fitting GTE into its structural separation policies. Under the Computer I regime, GTE was required to offer enhanced services through separate subsidiaries because it had annual revenues exceeding $1,000,000. Under the original Computer II order, GTE, along with AT & T, was required to continue to maintain structural separation. However, on reconsideration, the FCC lifted the requirement from GTE. At the time of the BOC Separation Order, the FCC determined that the divested BOCs did not compare with GTE. Now in Computer III, the FCC has decided that, in fact, they do. The FCC’s classification of GTE has been at best inconsistent and at worst totally random as has been its comparison of GTE with the individual BOCs.
In sum, the four purported “changes” in the telecommunications market identified by the FCC lend no support to its conclusion that the risk of cross-subsidization by the BOCs has decreased. We are left with the Commission’s assurance that cost-accounting regulations will be sufficient to “minimize” the BOCs’ ability to misallocate costs, Phase I Order, 104 F.C.C.2d at 1010, along with its profession that it has other unidentified methods of preventing cost-shifting.26 See supra p. 1233; see also FCC Brief at 104-05. But as we have already pointed out, the Commission’s consistent position before Computer III has always been that monitoring and enforcement problems make cost-accounting regulations an ineffective tool in detecting cost-shifting.27 Should the BOCs be free to [1238]*1238integrate their basic and enhanced operations, nothing in the record suggests that the FCC (or state regulators) will have any less difficulty than before in determining whether costs have been misallocated. Indeed, the only justification the Commission has offered for its heavy reliance on cost-accounting regulations in Computer III is that the risk of cost-shifting has been reduced by the four so-called “market changes.” Because, as we have discussed, the record fails to show that these purported market changes have demonstrably reduced either cost-shifting opportunities or incentives, the Commission’s justification for its new policy change lacks record support. In sum, the Commission has failed to explain satisfactorily how changed circumstances justify its substitution of nonstructural for structural safeguards to protect telephone ratepayers and enhanced services competitors from cross-subsidization.
C. The Reasonableness of the FCC’s Decision
The Computer III nonstructural regulations are designed to protect both ordinary telephone customers and the competitive enhanced services market. Phase I Order, 104 F.C.C.2d at 1010. We agree that the Commission has made a plausible case that lifting the structural separation requirements will benefit consumers of enhanced services by permitting the BOCs to operate more efficiently in the enhanced services market. We also agree that the Commission has made a plausible case that ONA, CEI, and the growth of bypass technology will be effective in reducing the risk of BOC access discrimination. Thus, the record supports the FCC’s determination that Computer Ill’s substitution of nonstructural for structural safeguards will benefit the enhanced service industry.28 On the other hand, the record yields no support for the Commission’s position that market and technological changes since Computer II and the BOC Separation Order have reduced the danger of cross-subsidization by the BOCs. We therefore hold that it was arbitrary and capricious within the meaning of section 10(e) of the APA for the Commission to abandon structural separation and rely on cost accounting regulations to provide regulatory protection for ratepayers and competitors against the harmful effects of cross-subsidization.
In conclusion, we emphasize that it is not our role to decide whether a rationale exists that would justify the FCC's decision to abandon the Computer II and BOC Separation Order regulatory framework. See State Farm, 463 U.S. at 43, 103 S.Ct. at 2866 (“we may not supply a reasoned basis for the agency’s action that the agency itself has not given”). We can only review the rationale used by the Commission, itself — that changed circumstances have reduced the danger of cross-subsidization.29 We cannot review a rationale that the Commission could conceivably have used, but did not. Unlike “minimum rationality” review under the due process and equal protection clauses, “arbitrary and capricious” review under the APA does not permit us to impute reasons to the agency and uphold its action if it has any conceivable rational basis. Although the Commission could have decided that the national interest in allowing the BOCs to compete more effi[1239]*1239ciently in the enhanced services industry justified reduced regulatory protection against cross-subsidization, that is not the case before us. As stated above, we are limited in our review to the reasons the FCC actually gave. Because those reasons are not supported by the record, we must hold that the Commission’s decision to release the BOCs from the structural separation requirements is arbitrary and capricious and therefore in violation of § 10(e) of the Administrative Procedure Act.
Ill
THE PREEMPTION ISSUE
In Computer III the FCC issued orders preempting nearly all state regulation of the sale of enhanced services by communications common carriers.30 Specifically, the FCC orders preclude state regulators from: (1) tariffing enhanced services sold by communications carriers; (2) requiring communications carriers to maintain structural separation between their basic and enhanced service operations; and (3) requiring nonstructural safeguards that are inconsistent with or more stringent than the FCC’s nonstructural safeguards.31 Petitioners argue that the FCC preemption orders violate § 2(b)(1) of the Communications Act, which denies the FCC jurisdiction “with respect to (1) charges, classifications, practices, services, facilities, or regulations for or in connection with intrastate communication service by wire or radio of any carrier.” 47 U.S.C. § 152(b)(1) (Supp. V 1982).
A. The Applicability of§ 2(b)(1) to Enhanced Services
The question of statutory interpretation presented for decision is whether § 2(b)(1) distinguishes between basic and enhanced telecommunications services. The FCC concedes that § 2(b)(1) reserves to the states the sole authority to regulate intrastate basic telephone service, but argues that the statute does not bar the FCC from regulating enhanced services to the exclusion of state regulation of intrastate enhanced services. Petitioners challenge the Computer III preemption orders on the ground that the plain language of § 2(b)(1) forecloses the FCC’s attempt to distinguish between basic and enhanced services for preemption purposes.
Simply put, petitioners contend that enhanced services are services “for or in connection with intrastate communication service by wire or radio of any carrier,” 47 U.S.C. § 152(b)(1), because enhanced services consist of nothing more than the transmission of electronic signals over the telephone network. In this fundamental respect, petitioners assert, enhanced services are indistinguishable from basic services. Petitioners rely particularly on the broad language of § 2(b)(1) reserving to the states the authority to regulate services “for or in connection with” communication services provided by telephone carriers. This broad language, claim the petitioners, explicitly “fences off from FCC reach or regulation,” Louisiana Pub. Serv. Comm’n v. FCC, 476 U.S. 355, 370, 106 S.Ct. 1890, 1899, 90 L.Ed.2d 369 (1986) (Louisiana PSC), both intrastate basic and enhanced services.
Given the fact that both basic and enhanced services are delivered by communications carriers over the same telephone network, and the broad “in connection with” language of § 2(b)(1), the Commission faces an uphill battle in defending its position that § 2(b)(1) distinguishes between basic and enhanced services. The gist of the Commission’s argument is that enhanced services fall beyond the reach of [1240]*1240§ 2(b)(1) because enhanced services, unlike basic telephone services, are not offered on a “common carrier” basis.32 In the Commission’s view, the key words in § 2(b)(1) are “for or in connection with intrastate communication service by wire or radio of any carrier." Because the term “carrier” is synonymous with “common carrier” for purposes of the Act,33 the Commission reasons, the words “of any carrier” must be interpreted to mean that § 2(b)(1) applies only to telecommunications services that are offered on a common carrier basis.
We reject the Commission’s interpretation of § 2(b)(1).34 We find nothing in the language of § 2(b)(1) to support the cramped reading advanced by the Commission. To the contrary, the broad language of § 2(b)(1) makes clear that the sphere of state authority which the statute “fences off from FCC reach or regulation,” Louisiana PSC, 476 U.S. at 370, 106 S.Ct. at 1899, includes, at a minimum, services that are delivered by a telephone carrier “in connection with” its intrastate common carrier telephone services. When telecommunications services are delivered on an intrastate basis by telephone carriers over the telephone lines, they at the very least qualify as services “in connection with intrastate communication service by wire ... of any carrier.” 47 U.S.C. § 152(b)(1). That these enhanced services are not themselves provided on a common carrier basis is beside the point. As long as enhanced services are provided by communications carriers over the intrastate telephone network, the broad “in connection with” language of § 2(b)(1) places them squarely within the regulatory domain of the states.
We are unimpressed by the Commission’s argument that the words “of any carrier” in § 2(b)(1) must mean that Congress intended the reach of § 2(b)(1) to be limited to common carrier services. The plain meaning of the language “of any carrier” is that the statute applies to communications services provided by common carriers such as AT & T and the BOCs as distinguished from communications services provided by non-common carriers such as IBM. Thus, the distinction made by the statute is between providers of communications services, i.e., between carriers and non-carriers. When services are provided by communications carriers, the statute makes no distinction based upon the terms and conditions on which the services are offered, i.e., whether on a common carrier or private contract basis.35
[1241]*1241We are equally unimpressed with the Commission’s argument that the legislative history of the Communications Act demonstrates that § 2(b)(1) was intended only to protect states’ jurisdiction to regulate intrastate common carrier telephone services. Before the Act was passed in 1934, state regulation had focused on common carrier services. During consideration of the 1934 Act, concern with maintaining states’ regulatory authority over common carriers emerged as a prominent theme.36 Congress’ purpose in writing § 2(b)(1) into the Act, the Commission asserts, was to preserve the states’ traditional role in protecting consumers against monopoly abuses by telephone carriers providing ordinary telephone services. “Nowhere,” says the Commission, “does the legislative history reveal any intention to insulate from federal regulation new types of communications services that might develop in the future and that the states never had regulated.” FCC Brief at 65.
This exegesis of the Act’s legislative history does not persuade us to abandon what we believe to be the unambiguous command of § 2(b)(1). As the Supreme Court noted in Louisiana PSC, even if it were true that Congress’ dominant concern was to protect states’ ability to control common carrier charges for intrastate telephone service, the language Congress chose reserved to the states expansive authority to regulate intrastate telecommunications. 476 U.S. at 372-73, 106 S.Ct. at 1900-01. “[W]here a statute is clear on its face, it is unnecessary to look to its legislative history to discern its meaning and scope.” In re Hudson, 859 F.2d 1418, 1421 (9th Cir.1988) (citing Tennessee Valley Auth. v. Hill, 437 U.S. 153, 184 n. 29, 98 S.Ct. 2279, 2296 n. 29, 57 L.Ed.2d 117 (1978)).
Our rejection of the Commission’s interpretation of § 2(b)(1) is in accord with the holding of the D.C. Circuit in National Ass’n of Regulatory Util. Comm’rs v. FCC, 880 F.2d 422 (D.C.Cir.1989). There, the D.C. Circuit held that § 2(b)(1) denies the FCC power to preempt state regulation of “inside wiring”37 that is used for both interstate and intrastate telephone service, despite the FCC’s classification of the installation and maintenance of inside wiring as non-common carrier services. See id. at 426. The court reasoned as follows:
Louisiana PSC establishes the governing principles for interpreting section 152(b). There the Court made clear that the Act, through section 152(b), establishes a system by which the states exercise the same authority over intrastate wire [1242]*1242communication as the FCC exercises over interstate wire communication:
By its terms, this provision fences off from FCC reach or regulation intrastate matters-indeed, including matters `in connection with' intrastate service. Moreover, the language with which it does so is certainly as sweeping as the wording of the provision [section 151] declaring the purpose of the Act and the role of the FCC.... We agree with petitioners that ... sections [151 and 152(b)] are naturally reconciled ... to enact a dual regulatory system....
476 U.S. at 370, 106 S.Ct. at 1899 (first emphasis added, second in original).
This passage and the language of section 152(b) compel us to reject the FCC's attempt to limit the reach of section 152(b) to "intrastate common carrier communication services." Recon. Order, 1 F.C.C.Rcd. at 1192, paras. 14-is. As the Court has made clear, section 152(b) `not only imposes jurisdictional limits on the power of a federal agency, but also by stating that nothing in the Act shall be construed to extend FCC jurisdiction to intrastate service, provides its own rule of statutory construction.' Louisiana PSC, 476 U.S. at 376-77 n. 5, 106 S.Ct. at 1902-03 n. 5. Indeed, even if the statute could be interpreted to read `intrastate common carrier communication service,' inside wiring would still fall within it as a facility or service offered `for or in connection with' a common carrier communication service, namely, intrastate telephone service. Either way, we cannot countenance the Commission's attempt to rewrite the statute. See id. at 376, 106 S.Ct. at 1902.
Id. at 428.
We agree with the D.C. Circuit. The extent of the authority to regulate intrastate communications services reserved to the states by § 2(b)(i) does not turn on whether the services are provided on a common carrier or non-common carrier basis. The Commission's interpretation of § 2(b)(1) would require us to qualify the statute's sweeping language "for or in connection with intrastate communication service by wire or radio of any carrier" by adding the words "which the carrier provides on a common carrier basis," If the Commission advocates such an amendment to the statute, then it must make its case to Congress, not to the courts.
B. The "Impossibility" Exception to S 2(A)(1)'s 1?e.striction on the FCC'S Preemption Authority
Having rejected the Commission's argument that § 2(b)(1) does not safeguard state authority to regulate intrastate enhanced services, we turn to the Commission's alternative argument that a portion of its Computer III preemption order is valid under an "impossibility" exception carved out of § 2(b)(1) by the Supreme Court in Louisiana PSC. In essence, the Commission argues that its preemption of state-imposed structural separation requirements and some state-imposed nonstructural safeguards is valid because such state regulations cannot feasibly coexist with the Computer III scheme.38
In Louisiana PSC, the Commission advanced this same argument in defending an order establishing uniform depreciation rates for telephone equipment used jointly for interstate and intrastate service. The Supreme Court rejected the argument, holding that § 2(b)(1) reserved to the states the right to set equipment depreciation rates for equipment used in providing intrastate telephone service. 476 U.S. at 373, 106 S.Ct. at 1901. Part of the Court's rationale was that separate FCC and state depreciation rates could coexist because it was feasible to allocate the cost of jointly used telephone equipment between interstate and intrastate use. Id. at 375, 106 S.Ct. at 1902.
The Court went on to suggest, however, that the Commission's uniform depreciation [1243]*1243schedule for equipment used jointly might have been valid if “it [were] not possible to separate the interstate and intrastate components of the asserted FCC regulation.” 476 U.S. at 375 n. 4, 106 S.Ct. at 1902 n. 4. The Commission finds support in this language for the recognition of an “impossibility” exception to § 2(b)(1). The Commission also finds comfort in the Supreme Court’s apparent approval of the Fourth Circuit’s decisions in North Carolina Utils. Comm’n v. FCC, 537 F.2d 787 (4th Cir.) (NCUC I), cert. denied, 429 U.S. 1027, 97 S.Ct. 651, 652, 50 L.Ed.2d 631 (1976), and North Carolina Utils. Comm’n v. FCC, 552 F.2d 1036 (4th Cir.) (NCUC II), cert. denied, 434 U.S. 874, 98 S.Ct. 222, 223, 54 L.Ed.2d 154 (1977). See Louisiana PSC, 476 U.S. at 375 n. 4, 106 S.Ct. at 1902 n. 4. In NCUC I and NCUC II, the Fourth Circuit upheld an FCC order preempting state regulations forbidding telephone customers from connecting their own terminal equipment to the telephone network. The Fourth Circuit upheld the order on the ground that FCC regulations permitting the connection of customer-owned terminal equipment to the telephone network for interstate communications could not feasibly coexist with state regulations prohibiting the use of customer-owned equipment for local service. In other words, the state regulations would effectively trump the FCC regulations because customers would not buy their own terminal equipment for interstate phone service if they were forbidden by state regulations from also using it for local service. The Fourth Circuit, after noting that “separation of terminal equipment used exclusively for local communication is a practical and economic impossibility,” concluded that “if state commissions could lawfully prohibit interconnection unless terminal equipment is used exclusively for interstate communications, federal [regulations] authorizing interconnection would be made nugatory.” NCUC II, 552 F.2d at 1043 (referring to conclusions reached in NCUC I).
We agree with the Commission that the language of Louisiana PSC requires us to recognize an “impossibility” exception to § 2(b)(1). The D.C. Circuit recently came to the same conclusion in NARUC III, 880 F.2d at 429, and Illinois Bell Tel. Co. v. FCC, 883 F.2d 104, 113-15 (D.C.Cir.1989). “In sum, the only limit that the Supreme Court has recognized on a state’s authority over intrastate telephone service occurs when the state’s exercise of that authority negates the exercise by the FCC of its own lawful authority over interstate communication.” NARUC III, 880 F.2d at 429.
The impossibility exception, however, is a limited one. The FCC may not justify a preemption order merely by showing that some of the preempted state regulation would, if not preempted, frustrate FCC regulatory goals. Rather, the FCC bears the burden of justifying its entire preemption order by demonstrating that the order is narrowly tailored to preempt only such state regulations as would negate valid FCC regulatory goals. As the D.C. Circuit held in NARUC III, “a valid FCC preemption order must be limited to [state regulation] that would necessarily thwart or impede” the FCC’s goals. Id. at 430 (emphasis added). “The FCC has the burden ... of showing with some specificity that [state regulation] ... would negate the federal policy_” Id. (emphasis added). We are therefore faced with the task of deciding whether the FCC’s regulation of interstate enhanced services would necessarily be frustrated by all possible forms of state-imposed structural separation requirements and by all state-imposed nonstructural safeguards that are inconsistent with Computer III requirements.
Our review of the record fails to persuade us that the Computer III preemption orders are “limited to [state regulation] that would necessarily thwart or impede” valid FCC goals.
(1) Preemption of State Structural Separation Requirements
First, the FCC has failed to carry its burden of demonstrating that all state-imposed separation requirements would negate its policy of permitting the structural integration of basic and enhanced services offered on an interstate basis. The Commission argues that any requirement that a [1244]*1244carrier separate its intrastate enhanced and basic services will necessarily force carriers to separate their interstate services as well, because the intra- and interstate components of enhanced services are structurally inseverable. See Phase I Reconsideration, 2 FCC Red at 3062. The Commission’s argument fails because it neglects to address the possibility that enhanced services may be offered on a purely intrastate basis. For example, the state petitioners call our attention to voice mail services that are offered to discrete locales within a state. Other examples of purely intrastate enhanced services include alarm services and database services for schools. The FCC does not explain how the structural separation of such purely intrastate enhanced services from basic telephone service would interfere in any way with a carrier’s ability to provide interstate enhanced services (or enhanced services with mixed intra- and interstate components) on an integrated basis.
In addition, the FCC appears to base its inseverability argument on the assumption that state structural separation regulations would necessarily require separation of physical facilities. See Phase I Reconsideration, 2 FCC Red at 3061-62. The record, however, does not support this assumption. State regulations might require only that carriers establish separate corporate organizations for providing intrastate basic telephone and enhanced services, while allowing the same facilities to be used for both types of services. The Commission has failed to explain why requiring communications carriers to offer intrastate enhanced services through a separate corporation would frustrate the Commission’s goal of giving communications carriers the freedom to choose whether to integrate or separate their interstate operations.
The Commission has made a plausible argument that some forms of state structural separation requirements would negate its policy of permitting the integration of basic and enhanced services offered on an interstate basis. For example, a state-imposed requirement that carriers use separate physical facilities for all basic telephone and enhanced services offered on an intrastate basis would almost certainly force carriers to separate their interstate services as well. However, the fact that state separation requirements could conceivably take forms that would frustrate valid federal goals is insufficient under the narrow “impossibility” exception to justify the Commission’s preemption of all state regulations requiring some form of structural separation.
(2) Preemption of State Nonstructural Safeguards
The record also fails to support the Commission’s preemption of (1) all state nonstructural safeguards applicable to AT & T and the BOCs that are inconsistent with the nonstructural safeguards imposed on AT & T and the BOCs by the FCC, and (2) all state nonstructural safeguards applicable to the independent communications carriers that are more stringent than those imposed by the FCC on AT & T and the BOCs.
As with its arguments in defense of its preemption of all state structural separation requirements, the Commission's arguments in defense of its preemption of inconsistent nonstructural safeguards fail to support the full breadth of the FCC’s preemption order. The FCC argues, for example, that “[m]any of the Computer III Open Network Architecture (“ONA”) and Comparably Efficient Interconnection (“CEI”) guidelines will require technical modifications to existing networks.... [Sjuch technical requirements cannot be separated jurisdictionally.” FCC Brief at 87 (emphasis added). As explained above, an argument that state regulation will negate valid federal purposes in “many” eases does not suffice to justify preemption of all state regulation in an area. The “impossibility” exception to § 2(b)(1) is a narrow one that may be invoked only when state and federal regulation cannot feasibly coexist. As the Supreme Court made clear in Louisiana PSC, the FCC must do more than pay lip service to Congress’ intent “to enact a dual regulatory system.” 476 U.S. at 370, 106 S.Ct. at 1899. “[T]he Commis[1245]*1245sion may take appropriate measures in pursuit of [its] goal[s], but only to the degree necessary to achieve [them].” NARUC III, 880 F.2d at 430.
The Commission offers even less compelling justification for that part of its order preempting state nonstructural regulations applicable to the independents that are more stringent than those imposed by the Commission on AT & T and the BOCs. The Commission notes only that it is desirable “that [the independents] not be subject to a greater degree of regulation than are the BOC[s],” Phase II Order, 2 FCC Red at 3102, and fails to explain what valid FCC goal such regulation would negate. In the absence of even an attempt to demonstrate that state regulation would frustrate valid FCC goals, we cannot uphold this part of the FCC’s preemption order.
IV
THE PRECLUSION ISSUE
The Commission contends that the doctrines of administrative finality and res judicata bar judicial review of the state petitioners’ challenge to the portion of the Computer III order preempting state tariffing of enhanced services. This part of the Computer III order simply reaffirms the Commission’s previous action in Computer II, which was “affirm[ed] ... in its entirety” in Computer & Communications Indus. Ass’n v. FCC, 693 F.2d 198, 203 (D.C.Cir.1982). The Commission now argues that because the validity of its preemption of state tariffing of enhanced services was determined definitively in the Computer II and CCIA proceedings, administrative finality and res judicata preclude the states from challenging Computer Ill’s continuation of the FCC preemption policy. We find the Commission’s reliance on these preclusion doctrines to be misplaced.
The doctrine of administrative finality generally bars a litigant from seeking review of administrative rulemaking after the statutory deadline for appeal has lapsed. See, e.g., Eagle-Picker Indus. Inc. v. EPA, 759 F.2d 905, 911-12 (D.C.Cir.1985). This doctrine does not apply, however, when an agency itself initiates a new rulemaking proceeding which reopens, and seeks public comment on, issues decided in the previous proceedings. In such circumstances, a litigant may seek review of the new proceeding even if the agency ultimately reaffirms its earlier decision. See Montana v. Clark, 749 F.2d 740, 744 (D.C.Cir.1984), cert. denied, 474 U.S. 919, 106 S.Ct. 246, 88 L.Ed.2d 255 (1985). Because Computer III was issued pursuant to a new notice-and-comment rulemaking proceeding which fully reopened the preemption issues decided in Computer II, the administrative finality doctrine does not bar judicial review of the order preempting state tariffing of intrastate enhanced services.
Nor is review of the tariffing preemption order barred by the doctrine of res judicata. Under res judicata, a final judgment on the merits ordinarily bars further claims by the parties based upon the same cause of action.39 See Montana v. United States, 440 U.S. 147, 153, 99 S.Ct. 970, 973, 59 L.Ed.2d 210 (1979). Res judicata does not act as a bar, however, “when the party against whom the earlier decision is asserted did not have a ‘full and fair opportunity’ to litigate the claim.” Kremer v. Chemical Constr. Corp., 456 U.S. 461, 480-81 & n. 22, 102 S.Ct. 1883, 1896-97 & n. 22, 72 L.Ed.2d 262 (1982).
We reject the Commission’s res judicata claim. We hold that the petitioners in CCIA did not have a full and fair opportunity to challenge the Commission’s preemption of state tariffing of enhanced services. The Commission’s preemption order was buried in a footnote, see Computer II Further Reconsidered Decision, 88 F.C.C.2d 512, 541 n. 34 (1981), and was issued after the CCIA petitioners had already filed their [1246]*1246opening briefs in the D.C. Circuit challenging the Computer II regulatory scheme set forth in the Commission’s Computer II Final Decision, 11 F.C.C.2d 384 (1980), and Computer II Reconsidered Decision, 84 F.C.C.2d 50 (1980).40
It is conceivable that the petitioners in CCIA could have amended their petition to the D.C. Circuit to respond to the Commission's footnote. However, we are unwilling to conclude that this limited opportunity to respond to the FCC’s footnote gave the state petitioners in CCIA a sufficiently full and fair opportunity to litigate. We are particularly unwilling to reach that conclusion in light of the public interest in the tariffing issue raised by petitioners and the public function that these state regulatory bodies serve. Cf. United States v. Mendoza, 464 U.S. 154, 162, 104 S.Ct. 568, 573, 78 L.Ed.2d 379 (1984) (public interest in government litigation relevant to applicability of preclusion rules). Accordingly, we hold that res judicata does not bar petitioners’ present challenge to the FCC’s order preempting state tariffing of enhanced services.
V
CONCLUSION
In summary, we hold (1) that the FCC’s substitution of nonstructural safeguards for the federal structural separation requirements to which the BOCs were subject was unlawfully arbitrary and capricious; and (2) that the FCC has failed to carry its burden of showing that its preemption orders are necessary to avoid frustrating its regulatory goals. We therefore grant the petition for review, vacate the Computer III Orders, and remand to the Commission for further proceedings consistent with this opinion.