These petitions were brought by ALLTEL Corp. ("ALLTEL") and Telephone and Data Systems, Inc. ("TDS") for review of orders of the Federal Communications Commission ("FCC" or "Commission") regarding the modification of the Commission's rules on the eligibility of local exchange carriers to use an average schedule method rather than an actual cost method for determining costs attributable to the provision of interstate services.1 The orders we are asked to review represent the latest attempt by the FCC to amend its average schedule eligibility rules. The previous attempt was embodied in an order that was remanded, in relevant part, to the Commission in National Ass'n of Regulatory Util. Comm'rs v. FCC, 737 F.2d 1095 (D.C.Cir.1984) [hereinafter NARUC ], cert. denied, 469 U.S. 1227, 105 S. Ct. 1224, 84 L. Ed. 2d 364 (1985). Petitioners contend that the FCC's latest modification of these rules is the product of arbitrary and capricious decisionmaking for much the same reason that the FCC's previous attempt was found to be arbitrary and capricious by this court. We agree with petitioners and therefore grant the petitions for review.

ALLTEL and TDS are holding companies with subsidiaries that operate telephone companies providing long distance and local exchange service in nineteen and twenty-three states, respectively. Each company, with its subsidiaries, has operating revenues in excess of $40 million. Many of the subsidiaries of both companies are small companies, located in and serving isolated geographic areas with small numbers of access lines. Fourteen of ALLTEL's and nineteen of TDS's subsidiaries are average schedule companies. All of these companies would be required by the rule reviewed herein to convert to cost status.

In 1984, this court remanded to the FCC "for further, more careful, analysis" orders of the FCC involving the eligibility of certain independent non-Bell telephone companies ("exchange carriers") to obtain compensation for interstate message toll and access services under what are commonly known as "average schedules." NARUC, 737 F.2d at 1103, 1127-29. The part of the order that was remanded dealt with an aspect of the larger problem of determining "access charges," which is the subject of Part 69 of the FCC's rules. 47 C.F.R. Secs. 69.1 to 69.611 (1986). The rules of Part 69 establish a system for reimbursing exchange carriers for that portion of local plant and service expense that is dedicated to providing access to long-distance lines and consequently generates revenue for long distance carriers. The determination of access charges requires the separation of interstate costs from total (unseparated) inter- and intrastate costs. This in turn requires--ideally--accurate and detailed analyses of the elements of unseparated costs.2

If such analyses were costless, it would be a simple matter to require every exchange carrier to perform extensive, periodic cost studies in order to determine, as exactly as possible, interstate costs. Such studies are not costless, of course, and they have been understood to be relatively more burdensome for small exchange carriers than for large. See, Average Schedule Order, 103 F.C.C.2d at 1018-19; MTS and WATS Market Structure: Memorandum Opinion and Order, 97 F.C.C.2d 682, 760 (1983) [hereinafter Access Charges Reconsideration Order ]. "As a result, the Commission's rules have traditionally allowed smaller exchange carriers to estimate some or all of their costs through use of an 'average schedule' which adopts generalized industry data to reflect the costs of a hypothetical exchange company." NARUC, 737 F.2d at 1127. Those carriers that did not elect to use the average schedule method employed cost studies to determine their interstate costs. In the rule reviewed by the NARUC panel, the FCC sought to compel all average schedule companies that were affiliated with any cost company to convert to cost status.3 The court found that the FCC had "clearly expressed its intent 'to avoid imposing the burden of developing cost information upon companies' which may be too small to perform the necessary cost studies." NARUC, 737 F.2d at 1128 (quoting Access Charges Reconsideration Order, 97 F.C.C.2d at 760). But the court found "nothing in the record which suggests that parent, subsidiary, or affiliate status accurately distinguishes between those companies which would be prohibitively burdened by the cost, and those which could easily bear it." Id. The court found that "the mere inference that affiliation alone indicates ability to bear the cost burdens of the affiliate is not always reasonable," id. (emphasis in original), and illustrated the unreasonableness of the inference with the example of a small cost company forced to finance the conversion of a small average schedule company. "More importantly," the court continued, "the Commission imposed this 'compulsory pooling requirement' between affiliates without inquiring into the regulatory or corporate barriers which may prohibit such cross-pooling...." Id. (quoting Access Charges Reconsideration Order, 97 F.C.C.2d at 760).

On remand, the FCC published a notice of proposed rulemaking, "seek[ing] comments and data that [would] permit [it] to assess the actual costs of exchange carrier compliance with the average schedule portion of the Commission's access charges rules." MTS and WATS Market Structure: Notice of Proposed Rulemaking, 49 Fed.Reg. 50,413, 50,414 (1984) [hereinafter Notice ]. Noting that the NARUC court had questioned whether "affiliation, without more, is a reliable indicator of the ability of a company to bear the expenses of cost studies," Notice, 49 Fed.Reg. at 50,414, the FCC observed that in adopting its rule, it was

guided by the perception that affiliates of commonly owned exchange carriers collectively possessed sufficient resources to perform those cost studies that would be required for computing interstate access charges. In addition, permitting holding companies to select the companies in their corporate "families" that would be cost, and those that would be average schedule, would foster incentives that would run counter to our objective of developing cost-based access charges.

Notice, 49 Fed.Reg. at 50,414. This observation contains two strands--sufficiency of resources, and counterincentives (something the FCC will later call "abuse" of average schedules)--by which the FCC later attempted to justify the new rule. See infra pp. 558-61. Curiously, the FCC requested no comments on, and made no mention of, what the NARUC court found to be the "[m]ore important[ ]" lacuna in the FCC's inquiry and reasoning--the failure to explore regulatory or corporate barriers to compulsory pooling. NARUC, 737 F.2d at 1128.

Four interested parties, responding to the Notice, submitted comments concerning the average schedule eligibility rule. The Rural Telephone Coalition ("RTC") opposed the affiliation-based rule. RTC challenged the FCC's reliance on the resources of affiliates on a number of grounds. Such resources are not "automatically available between companies that are commonly-owned" and to the extent that they "may be available, they are not provided free of charge to affiliates." RTC Comments at 3, Joint Appendix ("J.A.") at 101.

The RTC provided an outline of the requirements for converting a company from an average schedule to a cost study basis, RTC Comments, Attachment A, J.A. at 111-14, and argued that affiliation of a small exchange carrier with cost carriers does not make conversion to cost of the small carrier significantly less expensive or easier. Id. at 10, J.A. at 108. It also argued that the FCC ignored the relatively great expense and complexity of conversion to cost for a small carrier, and ignored the complex interaction of the Commission's rules and state ratemaking when it inferred that incentives necessarily exist that lead affiliated carriers to manipulate the average schedule option. Id. at 5-8, J.A. at 103-06. The United States Telephone Association ("USTA"), the national trade association for local exchange carriers, also opposed the affiliation rule. Its comments included estimates of the cost of converting a small exchange carrier from average schedule to cost. USTA Comments at 2-3, J.A. at 116-17. The estimates of the USTA are the only dollar estimates to be found in the record. The comments of RTC and USTA indicate that the cost studies required for the conversion to and maintenance of cost status are not a one-time affair. RTC Comments, Attachment A, at 4, J.A. at 114; USTA Comments at 3, J.A. at 117. The cost of annual "true-up" studies was estimated by USTA to be about one third the cost of the initial study. USTA Comments at 3, J.A. at 117.

American Telephone and Telegraph Co. ("AT & T") and NYNEX argued for the retention of the FCC's eligibility rule. They reiterated, without supporting data or argument, the FCC's sugestion that affiliated companies must be selecting average schedule status to the detriment of the public. See AT & T Comments at 12, J.A. at 73; NYNEX Comments at 14, J.A. at 97. AT & T went on to mention potential problems with the rule, suggesting that they could be alleviated by the granting of waivers:

For certain small, rural exchange companies, the cost of preparing actual cost studies may be exorbitantly high because cost studies necessitate not only appropriate software but extensive data gathering, on site expertise, and accounting and administrative procedures. In some cases, the preparation of actual studies for all affiliates may, in fact, drive the total costs above what they would have been if both average and actual costs could be used, a result that is harmful to telephone customers.

AT & T Comments at 12, J.A. at 73.

In April 1986, the FCC issued the Average Schedule Order, which modified its average schedule eligibility rule. The rule as remanded to the Commission had required average schedule companies affiliated with a cost company to convert to cost status. The rule as modified (and as it now comes before this court) requires an average schedule company affiliated with a cost company to convert to cost status, only if the aggregate revenues of that company and its affiliates equal or exceed $40 million.4

The Average Schedule Order is surprisingly short of reasons for the modified affiliation-based eligibility rule. The unsupported suggestion that average schedule status has been elected by affiliated carriers as a means of inflating returns is reiterated as the evil the rule was designed to prevent, and the comments of RTC were said to suggest "the presence of a potential burden of sufficient magnitude to warrant softening the application of [the] rule." Average Schedule Order, 103 F.C.C.2d at 1028-29.

The question raised by the NARUC court as to the relevance of affiliate status to the determination of which companies should be required to change to cost status, NARUC, 737 F.2d at 1128, is not answered by the Average Schedule Order. The FCC also dismissed the NARUC court's concern with possible regulatory and corporate barriers by stating:

The relevance of [the NARUC court's] observation is not apparent in the context of distributions of interstate revenues from services that are not regulated by state commissions. This is particularly true in light of the fact that exchange carriers that choose to settle, or that are required to settle, on a cost basis for interstate services may include all reasonable expenses that are incident to the provision of those services (including the expenses that are associated with studies that determine the costs of providing interstate exchange access services), and receive compensation on that basis.

Average Schedule Order, 103 F.C.C.2d at 1021 n. 17 (citations omitted).

ALLTEL and TDS petitioned for reconsideration of those portions of the Average Schedule Order that dealt with average schedule eligibility. Among petitioners' contentions were that the FCC had still not adequately explained why affiliation could be taken as an indicator of the cost-effectivenessof, and a company's ability to bear, cost studies; that the $40 million cut-off, which was meant to "soften" the application of the rule, arbitrarily treats small companies that are affiliated with $40 million corporate families differently from small companies that are not; and that the FCC had not given any support for its assumption that average schedule status was being improperly manipulated by affiliated carriers.

The FCC denied the petitions for reconsideration insofar as they challenged the average schedule eligibility rule. The Commission again assumed the relevance of affiliate status to the determination of which exchange carriers would be required to convert to cost status. It simply asserted that its rule would require a company to convert to cost status "only when sufficient resources existed to provide for the studies necessary" for the conversion. "It is our view," stated the Commission, "that the expertise and resources of all affiliated companies may properly be called upon to perform cost studies for conversion of average study companies to interstate cost status." Reconsideration Order, slip op. at 7 & n. 25. In addition to relying upon the availability of resources, which the affiliation rule was said to assure, the FCC asserted that the reasonable costs of interstate cost studies would be fully recoverable through interstate rates. Id.

The FCC justified its choice of the particular number of $40 million by pointing out that $40 million in annual operating revenues was already used as the demarcation point for an unrelated purpose: the classification of companies into subsets in order to determine proportional representation on the Board of Directors of the National Exchange Carrier Association ("NECA"). See 47 C.F.R. Sec. 69.602 (1986). Forty million dollars was said to be "an adequate indicator of sufficient resources in virtually all instances" and would impose "no additional administrative costs" because it was already used to determine NECA Board representation. Reconsideration Order, slip op. at 7-8.

ALLTEL and TDS filed petitions for further reconsideration with the FCC on October 30, 1986. ALLTEL and TDS filed their petitions for review in this court, which were subsequently consolidated, on November 26, 1986, and December 1, 1986, respectively. The Commission moved to have this case held in abeyance until it issued its decision on the petitions for further reconsideration. On March 30, 1987, we granted the Commission's motion until "the earlier of the completion of [the FCC's] review of its orders or thirty days from the date of this order." Because the Commission failed to act within the 30-day limit, this case was calendared.

Prior to oral argument, the FCC dismissed the petitions for further reconsideration of ALLTEL and TDS. Dismissal Order, 2 FCC Rcd at 6642. The FCC characterized the petitions as repetitive. Although it dismissed the petitions, it did take the opportunity to repeat its own arguments in favor of the rule in extensive footnotes. On December 4, 1987, and December 21, 1987, respectively, ALLTEL and TDS brought petitions for review in this court of the Dismissal Order. By order filed January 13, 1988, we consolidated those petitions with the petitions for review of the Average Schedule Order and Reconsideration Order.

The question before us is whether the average schedule eligibility rule is arbitrary and capricious. 5 U.S.C. Sec. 706(2)(A) (1982). The scope of our review under that standard is, of course, narrow; we are "not empowered to substitute [our] judgment for that of the agency." Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 416, 91 S. Ct. 814, 824, 28 L. Ed. 2d 136 (1971). Nevertheless, as a reviewing court we must search for a reasoned explanation for agency action, and where, as here, that reasoned explanation is not articulated by the agency in the orders under review and cannot be reasonably discerned by this court from the actions of the agency, we are constrained to grant the petition for review and remand the case to the agency. See, e.g., Motor Vehicle Mfrs. Ass'n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43, 103 S. Ct. 2856, 2866, 77 L. Ed. 2d 443 (1983).

The Commission is authorized to prescribe just and reasonable rates. We have repeatedly recognized that the FCC has broad discretion in its choice of methods for the determination of rates. See, e.g., MCI Telecommunications Corp. v. FCC, 675 F.2d 408, 413 (D.C.Cir.1982). "A basic principle used to ensure that rates are 'just and reasonable' is that rates are determined on the basis of cost." Id. at 410 (quoting 47 U.S.C. Sec. 201(b) (1976)). Like the NARUC panel, "[w]e do not find unreasoned the Commission's determination that, to the extent practical, telephone prices should be 'based upon the true cost characteristics of telephone company plant.' " NARUC, 737 F.2d at 1147 (quoting MTS and WATS Market Structure, 93 F.C.C.2d 241, 251 (released 1983) (Access Charges Order)). In spite of this interest in determining "true cost characteristics," however, the FCC has not required all exchange carriers to perform cost studies. The FCC has permitted some exchange carriers to forego cost studies and to rely instead on less accurate average schedules. The reasons that the FCC has offered this exception from its true cost approach are important to this case.

The NARUC panel noted that the Commission had "clearly expressed its intent 'to avoid imposing the burden of developing cost information upon companies' which may be too small to perform the necessary cost studies." NARUC, 737 F.2d at 1128 (quoting Access Charges Reconsideration Order, 97 F.C.C.2d at 760). AT & T, in its response to the Commission's request for comments after the NARUC remand (a response that argued in support of the affiliate-based eligibility rule), observed that

[f]or certain small, rural exchange companies, the cost of preparing actual cost studies may be exorbitantly high.... In some cases, the preparation of actual studies for all affiliates may, in fact, drive the total costs above what they would have been if both average and actual costs could be used, a result that is harmful to telephone customers.

AT & T Comments at 12, J.A. at 73.

The Commission, in the Average Schedule Order, commented upon the historical practice of allowing small carriers to use average schedules:

This procedure had the advantage of substantially reducing the costs that were imposed upon small exchange carriers in receiving compensation for their interstate services, with the concomitant benefit to the public of reducing the expenses of accounting and separations studies that otherwise would have to be borne, in part, by interstate ratepayers.

Average Schedule Order, 103 F.C.C.2d at 1019.

It is clear from these descriptions of average schedule eligibility that the cost of performing cost studies is relatively more burdensome for small carriers than for large. The cost of conducting a cost study on certain small carriers is apparently too large to justify incurring it. There is some economy of scale here: as progressively larger carriers are considered, the cost of performing cost studies decreases relative to the benefit provided--however that benefit may be measured. This understanding of the average schedule option is apparently shared by all of the parties.

The economy of scale appears to arise as a function of the size of the exchange carrier rather than as a function of the size of the affiliated group of carriers. That is a crucial point in this case. The NARUC court mentioned the absence in the record of anything that would indicate that "affiliate status accurately distinguishes between those companies which would be prohibitively burdened by the cost, and those which could easily bear it." NARUC, 737 F.2d at 1128. The comments of RTC raise the point directly. RTC submitted an outline of the tasks and special studies an average schedule company would have to complete in order to convert to cost status, and stated "that placing an average schedule company affiliated with a cost company on cost is not significantly less expensive or easier by reason of such affiliation." RTC Comments at 10, J.A. at 108. The Petition for Reconsideration of TDS also challenged directly the idea that the crucial economy of scale would arise as a function of aggregate affiliate size: "The burdensome cost activities described by RTC apply equally to individual small rural companies, small rural companies with only one or a few affiliates and small rural companies affiliated with any number of other small rural companies." Petition for Reconsideration of TDS at 4, J.A. at 162.

The FCC failed to answer these challenges. Neither the Average Schedule Order nor the Reconsideration Order argued that the subject economy of scale arises as a function of aggregate affiliate size. The FCC does eventually utter such an argument--in its brief, at oral argument, and in the Dismissal Order. Why we must regard those utterances as too little, too late, will be discussed below. Infra p. 562. First we will consider the rationale that was presented by the Commission when it promulgated the rule. That rationale--presented in the Average Schedule Order and the Reconsideration Order--was woven of four strands. The FCC, referring to those four strands, argues before us that its affiliate-based rule is not arbitrary and capricious because (1) it assures adequate resources for conversion to cost, (2) the costs of conversion are compensable from the interstate revenue pool, (3) the rule remedies carrier abuses, and (4) carriers may apply for waivers of the rule. We will take up each of these arguments in turn.

1. The FCC contends that the affiliation-based average schedule eligibility rule should survive scrutiny because it ensures that any average schedule company required to convert to cost status will have adequate resources for the conversion. "It is our view," wrote the FCC, "that the expertise and resources of all affiliated companies may properly be called upon to perform cost studies for conversion of average study companies to interstate cost status." Reconsideration Order, slip op. at 7 n. 25. This does nothing to explain what affiliation with a large corporate family has to do with the question of whether a small rural exchange carrier, for example, should be required to convert to cost status. If average schedule status is permitted because the cost of performing cost studies on a small geographical study area overwhelms whatever benefits would derive from such studies, then affiliation, which does nothing to alter the characteristics of the study area, is not a relevant criterion for determining eligibility for that staus. The fact that other companies in an affiliated group pay for the study does not cure the problem that customers pay for a study that costs more than it is worth. That the customers so amerced are those of affiliated carriers rather than of the small carrier required to do the study would not seem to alter the analysis.

It is difficult, therefore, to understand the Commission's reliance on the resources of an affiliated group in determining whether an average schedule company should change to cost. We are apparently to assume, in the face of comments to the contrary, and without supporting arguments from the Commission, that the wealth of a holding company somehow reduces the high cost of cost studies relative to the benefits derived from those studies. We do not hold that no such showing could be made, but the Commission must do more than simply ignore comments that challenge its assumptions and must come forward with some explanation that its view is based on some reasonable analysis.

Without some showing that aggregate revenues of affiliates is relevant to the decision of whether an average schedule company should change to cost, limiting the application of the rule to companies with aggregate revenues of $40 million does nothing to cure the rule struck down by the NARUC court. The Commission observed that the affiliation rule might impose certain burdens, and that a "softening" of the application of the rule was warranted. The softening consisted of the $40 million threshold. The inadequacy of this response to the remand is perhaps best illustrated by an analogy. It is as if the FCC, asked to explain a rule that people with brown hair will be denied broadcast licenses, responded that it recognized that its rule imposed a burden on people with brown hair and in order to "soften" the application of the rule it would deny licenses only to people with brown hair who were born in odd numbered years. Softening an arbitrary and capricious rule does not necessarily cure it.

The Commission contends that its choice of where to draw the line in a case like this "where there is no mathematically 'correct' answer, is entitled to deference." Brief for Respondents at 18 n. 29 (citing Home Box Office, Inc. v. FCC, 567 F.2d 9, 60 (D.C.Cir.), cert. denied, 434 U.S. 829, 98 S. Ct. 111, 54 L. Ed. 2d 89 (1977)). The Commission misses the mark. If the Commission could demonstrate that affiliation were a relevant criterion, then the Commission would be entitled to deference on its selection of the precise point on the scale of aggregate revenues. But the Commission has failed to make any such showing. We cannot defer to the Commission's selection of a precise point on a scale when the scale itself has "no relationship to the underlying regulatory problem." Home Box Office, 567 F.2d at 60.

An average schedule eligibility rule based on revenues (or number of access lines) of an exchange carrier would not pose the same difficulties as the rule under review. There is a certain surface similarity between a rule based on revenues of an exchange carrier and a rule based on revenues of an affiliate group. But while revenues of exchange carriers might be a relevant scale upon which to choose the eligibility point because of its correlation to the apposite characteristics of study areas (we cannot say on this record), the scale of aggregate revenues has not shown to be relevant.

2. In the rulemaking and review proceedings following the NARUC remand, the FCC emphasized a rather peculiar argument. In a purported attempt to explain its assertion that the resources of affiliates are relevant to the determination of which carriers should be eligible for average schedule status, the Commission completely undercut that assertion. In the Reconsideration Order the Commission "decided that $40 million in annual operating revenues for all affiliated companies would adequately address the concerns of the [NARUC] Court, especially because the reasonable costs of interstate studies for average schedule companies that convert to cost company treatment are fully compensable from interstate ratepayers as allowable expense." Reconsideration Order, slip op. at 7-8 (emphasis added). Rather than adequately addressing our concerns, this explanation aggravates them. If the reasonable costs of cost studies are fully compensable out of the interstate revenue pool, then any and all average schedule companies have adequate resources to convert to cost status. The Commission's argument is difficult to comprehend. We are asked to believe that resources of affiliates is a relevant criterion especially because--since the reimbursing of carriers is guaranteed--it is an irrelevant criterion. The compensability rationale does not support the resources rationale; it eviscerates it.

The FCC's argument contains two obviously contradictory positions. Either costs are fully compensable and resources of affiliates are irrelevant, or resources of affiliates are relevant and some costs must be borne by customers of or investors in affiliates. It does not matter which is the true position of the FCC; either way we fail to find a reasoned basis for the rule.

It is not at all clear what the Commission means to gain by arguing full compensability. Perhaps the Commission wishes to justify its failure to explore the concerns with corporate and regulatory cross-pooling raised by the NARUC court. It may be true that corporate and regulatory cross-pooling do not present problems for the rule if costs are fully compensable, but that does nothing to explain the rationality of the affiliation criterion.

If, however, we take seriously the implications of the Commission's reliance on the resources of affiliates, then the rule presumably would necessitate interaffiliate cross-pooling and would have some impact on the decisions made by intrastate ratemakers. In that case, issues raised by the NARUC panel and dismissed as irrelevant by the Commission, stand unanswered. The NARUC court noted that the Commission imposed its " 'compulsory pooling requirement' between affiliates without inquiring into the regulatory or corporate barriers which may prohibit such cross-pooling." NARUC, 737 F.2d at 1128 (citation omitted). The Commission argues that state ratemaking and inter-corporate barriers are not its concern because state and corporate authorities must arrange their affairs to comply with valid FCC rules. But what is at issue here is the validity of an FCC rule. We will not speculate here about possible financial covenants and other corporate barriers under which local exchange carriers may operate that would prevent the transfer of corporate funds to affiliates in other jurisdictions. The FCC informs us that the record is barren of evidence of such impediments, yet another searches the notice of proposed rulemaking in vain for any request for comments on this issue which was specifically highlighted as a concern of the NARUC court. Evidence of regulatory impediments to interaffiliate transfers may be found in state statute books. See, e.g., N.Y.Pub.Serv.Law Secs. 106, 107, 110 (McKinney 1955 & Supp.1988). If resources of affiliates is indeed a relevant criterion as the FCC sometimes argues, then the FCC must come forward with some rationale for taking actions that will displace state regulatory and corporate law arrangements.

3. The FCC argues that companies have abused the average schedule option in order to inflate their interstate costs, and that the eligibility rule is a remedy for this abuse. This facially plausible justification fails on closer analysis. There has been no showing that such abuse exists and no showing that the rule targets companies engaged in such abuse.

In the Notice, the Commission stated that "[permitting carriers to elect average schedule treatment ("AST"), when AST results in returns that exceed costs, and costs when cost treatment yields returns that exceed AST, would tend to create a revenue requirement that would diverge from costs...." Notice, 49 Fed.Reg. at 50,414. NYNEX repeated the Commission's statement in its comments: "Absent [this] rule[ ], a company would be free to choose to rely on AST or actual cost studies depending on which method produces the higher costs." NYNEX Comments at 14, J.A. at 97. AT & T also picked up the Commission's statement and, in its comments, asserted that the eligibility rule "is designed to prevent a parent company from performing cost studies for its high cost subsidiaries while relying on average schedule costs for its low-cost subsidiaries, thus obtaining an overall return exceeding what it would have realized if it treated all of its subsidiaries' costs on the same basis." AT & T Comments at 12, J.A. at 73. In the Average Schedule Order, the FCC completes the circle by quoting from this AT & T comment as to the purpose of the eligibility rule. See Average Schedule Order, 103 F.C.C.2d at 153.

The Commission has done little more than hypothesize that companies are selecting average schedule status in order to overstate interstate costs. Neither the Commission nor any party to the rulemaking or to this case has made any showing that such manipulation has taken place. We are told that it must be so, because it would be " 'irrational to surmise that the carrier would have declined to try to maximize its allowable profits.' " Brief for Respondents at 17 n. 27 (quoting New England Telephone and Telegraph Co. v. FCC, 826 F.2d 1101, 1110 (D.C.Cir.1987)).

We accept the plausibility of the profit-maximizing hypothesis as an incentive for this sort of manipulation of regulation. But it does not follow, and there is no showing, that eliminating the possibility of some unknown amount of suspected abuse outweighs the other disadvantages of the affiliation rule which we have discussed. Moreover, the commission has failed to present a believable argument that, in the case before us, the necessary conditions exist for the profit motive actually to lead to such manipulation.

The FCC's argument relies on too many questionable assumptions. First we must assume that companies know, without performing cost studies, that their interstate costs would be lower after cost studies than under the average schedule. Second, we must assume that the lower interstate recovery that companies are supposed to be avoiding would not, in any event, be offset by the correspondingly higher intrastate revenues that petitioners contend would result from residual state ratemaking.5 Third, we must assume companies are not choosing average schedules here in many cases simply because the cost of the marginal improvement in accuracy is too high--in which case actions desirable from a profit perspective and from the FCC's public good perspective may coincide. And finally, we must assume that the conditions leading to this "abuse" under the old average schedules will continue to exist under the updated average schedules that have been adopted by the Commission. See Average Schedule Order, 103 F.C.C.2d at 1024. The argument that relies on this chain of assumptions, and which is made in the Commission's brief, cannot substitute for the reasoning that is lacking in the FCC's orders. See, e.g., SEC v. Chenery Corp., 318 U.S. 80, 87-88, 63 S. Ct. 454, 459, 87 L. Ed. 626 (1943); City of Brookings, 822 F.2d at 1165.

Furthermore, if we assume that companies are "abusing" the current system, the Commission has given us no reason to believe that the companies engaged in such abuse are those that are affiliated with cost carriers. We do, of course, accord deference to a determination by the Commission that a problem exists within its regulatory domain, but deference is not a blank check. As we have noted before, "a 'regulation perfectly reasonable and appropriate in the face of a given problem may be highly capricious if that problem does not exist.' " Home Box Office, 567 F.2d at 36 (quoting City of Chicago v. FPC, 458 F.2d 731, 742 (D.C.Cir.1971), cert. denied, 405 U.S. 1074, 92 S. Ct. 1495, 31 L. Ed. 2d 808 (1972)).

4. The Commission would have us uphold this rule because "the Commission stated that it would entertain waiver requests from holding companies subject to the conversion requirement if they could demonstrate that application of the rule would impose an unreasonable burden." Brief for Respondents at 11.

The FCC cannot save an irrational rule by tacking on a waiver procedure. "The very essence of waiver is the assumed validity of the general rule...." WAIT Radio v. FCC, 418 F.2d 1153, 1158 (D.C.Cir.1969). The deference that we accord administrative action on waiver applications depends upon this assumption. Only because of this assumption does "[a]n applicant for waiver face[ ] a high hurdle even at the starting gate." WAIT Radio, 418 F.2d at 1157. It is true "that a rule which otherwise might be impermissibly broad can be saved by the 'safety valve' of waiver or exemption procedures." KCST-TV, Inc. v. FCC, 699 F.2d 1185, 1200 (D.C.Cir.1983) (Scalia, J., dissenting) (emphasis in original). If the Commission's argument were accepted, no rule, no matter how irrational, could be struck down, provided only that a waiver provision was attached. A rule with no rational basis--at least none that the FCC has so far been able to reveal--cannot be saved in this fashion.

In its brief the Commission introduces a new rationale. That is, of course, unavailing since we sit to review the agency's explanations rather than its counsel's. See, e.g., SEC v. Chenery Corp., 318 U.S. 80, 87-88, 63 S. Ct. 454, 459, 87 L. Ed. 626 (1943); City of Brookings, 822 F.2d at 1165. The FCC contends here for the first time that the cost estimates submitted by USTA in its comments, see supra p. 555, are inapposite. "Economies of scale likely would reduce the costs per company when the expertise of affiliates are made available and the studies are conducted by in-house experts. The USTA estimate does not specifically take into account these economies of scale." Brief for Respondents at 21 n. 37. The record gives no indication, however, that the FCC considered this kind of efficiency. To the contrary, the record is scattered with data and remarks that cry out for explanation or response if this is indeed the rationale for the rule. The comments of RTC state directly that such affiliation-related efficiencies do not exist. If the FCC's rule was based on a view of the industry that squarely contradicted this prominent assertion of the RTC, we think that the FCC would have given some indication that it noticed that the cornerstone of its rule had been challenged. Instead, the orders do not mention this crucial disagreement. Indeed, they do not even mention this rationale that depends on a tacit assumption that was contradicted in the record. Counsel's claim that the rule is based on these economies of scale comes too late in the day and flies in the face of too many contrary indications and conspicuous silences in the record to be believed.

An agency's claim that certain economies of scale exist in the industry that it regulates is the kind of claim that we would ordinarily accord deference because of presumed agency expertise. Cf. Office of Communication of United Church of Christ v. FCC, 707 F.2d 1413, 1440 (D.C.Cir.1983) (discussing deference to agency's cost benefit analysis). In this case, the agency has done a remarkable job of rebutting the presumption of its own expertise. In its brief the FCC claimed that the average schedule eligibility rule would apply only to ALLTEL and TDS. Since the FCC's brief was in many respects a defense of the rule as applied to petitioners rather than a defense of the rule in general, this point was not trivial. Later, in a letter to this court, the FCC admitted that it had erred. "[T]here are several other holding companies who are required to convert average schedule affiliates to cost status." Letter from Laurel Bergold, Counsel, to George A. Fisher, Clerk (Nov. 19, 1987). We are left with the impression that the FCC is unsure of the effect its rule would have. Under these circumstances we receive agency pronouncements with some skepticism. "While agency expertise deserves deference, it deserves deference only when it is exercised...." Cities of Carlisle and Neola v. FERC, 741 F.2d 429, 433 (D.C.Cir.1984).

In the Dismissal Order, the FCC dismissed petitions for reconsideration filed by ALLTEL and TDS because "petitioners have not demonstrated any special circumstances that would warrant consideration of repetitive petitions." Dismissal Order, 2 FCC Rcd at 6642. Dismissal notwithstanding, the FCC argues extensively in support of the average schedule eligibility rule in the footnotes of the Dismissal Order. We need not decide here whether substantive arguments presented in an order dismissing, on procedural grounds, reconsideration of agency action may be considered part of the "reasoned explanation" of that action by a reviewing court in determining whether the agency acted arbitrarily and capriciously. The only "new" arguments presented in the dismissal order are the same unsubstantiated assertion about economies of scale and the same erroneous claim about which companies are affected by the rule that were found in the Commission's brief. See supra pp. 561-562. It does not matter whether or not we look to these arguments. They do not supply the necessary reasoned explanation that is missing from the FCC's promulgation of this rule.6

We find that the Commission has failed to supply a reasoned basis for its average schedule eligibility rule and that, absent such a basis, the average schedule rule is arbitrary with or without the exception for companies affiliated with groups of a certain size. We also find that the Commission cannot escape judicial review of wholly arbitrary action by instituting a waiver procedure that would allow it to correct in the future at its discretion the arbitrary results of that action.

We grant the petitions for review and remand for further proceedings consistent with this opinion.

So ordered.