dissenting:
I respectfully dissent. Although both sides stake out rather rigidly extreme positions, the more reasonable, middle ground is that the NRC’s decision, albeit reached under difficult time constraints, fails to pass muster under both the first and second steps of the familiar Chevron analysis.
I
Petitioners rely on National Cable Television Association, Inc. v. U.S., 415 U.S. 336, 94 S.Ct. 1146, 39 L.Ed.2d 370 (1974) (“NCTA ”), its companion case Federal Power Commission v. New England Power Co., 415 U.S. 345, 94 S.Ct. 1151, 39 L.Ed.2d 383 (1974), and the series of opinions from this court interpreting those two cases 1 for the proposition that, in order for the NRC fees to be sustained, the fees must be shown (1) to reflect the cost of providing specific benefits to identifiable power reactor licensees and (2) to exclude costs that serve independent public benefits.2
Petitioners make, in essence, two arguments for applying these criteria to the NRC fees at issue. First, they argue that NCTA and its progeny draw a distinction between “fees,” which agencies may assess (subject to the limitations developed in the decisional law), and “taxes,” which can be assessed only by Congress. Since the power to tax cannot be delegated to an administrative agency, the argument runs, Congress must have intended to authorize only a fee, which is therefore subject to the two criteria listed above.3 Second, petitioners *389argue that the language of COBRA tracks the language of both the IOAA and cases construing the latter statute. From this, petitioners conclude that Congress intended the criteria developed in that body of law to apply to COBRA. In petitioners’ view, the fee schedule must be invalidated and the assessments refunded, because the NRC did not apply these criteria in arriving at its fee assessments in the instant case.
NRC, not surprisingly, is of a decidedly different perspective. The Commission contends that the NCTA line of cases is irrelevant to COBRA’s provisions. Under COBRA, the applicable statutory standard requires only that the NRC impose charges that “reasonably relate[ ] to the regulatory services” it provides and that those charges “fairly reflect the cost ... of providing” such services. Since it has calculated the annual charge in accordance with those standards, NRC maintains, its assessment under the auspices of COBRA should be sustained.
These, then, are the competing positions. Both, as I indicated at the outset, seem rather extreme. If, as petitioners argue, Congress intended the same criteria for fee assessments developed under the IOAA to apply here, then enactment of COBRA would have been but a hollow exercise. Clearly, Congress intended NRC to go beyond the impositions of fees already assessable under IOAA. See FPL Brief at 28; maj. op. at 769. Nor can petitioners successfully argue that the NCTA line of cases (or indeed, NCTA itself) stakes out the outer limits of Congress’ power to delegate assessment authority. The specific criteria enunciated in those cases were developed in the context of construing a particular statute (the IOAA). The standards enumerated in Central & S. Motor Freight Tariff Ass’n v. ICC, 777 F.2d 722 (D.C.Cir. 1985), for example, are certainly not, as I understand them, of constitutional dimension.4
I thus agree with my colleagues that “Congress [did not] intend[] that COBRA incorporate the same limitations as IOAA.” Maj. op. at 768. I also agree that COBRA was intended to authorize the NRC to recover generic costs, such as those for research and rulemaking. Maj. op. at 769.5
On the other hand, NRC’s sweeping interpretation of COBRA is also extreme, as the Commission itself recognizes. From NRC’s calculation of the annual charge, it is apparent that the agency interprets the requirement that the fee be “reasonably related to the regulatory service provided” in a manner that reads out any requirement that the service provide specific benefits to identifiable beneficiaries. NRC makes no benefit-side calculation whatsoever, other than the hopelessly general determination that all services that form the basis for its cost calculations benefit the entire class of holders of power reactor operating licensees to a greater or lesser extent. NRC makes no attempt to allocate specific benefits to identifiable beneficiaries; nor does the Commission explain why a per capita assessment best makes any such allocation.
*390The only virtue of NRC’s method of calculating the annual fees is its Spartan simplicity. In what Doestoevsky would doubtless characterize as an act of bureaucratic passion, the agency in classic Washington fashion aggregated all portions of the NRC’s budget that benefited all power reactor licensees as a class; that amount was then divided among all reactor licensees. NRC made no attempt to show that its expenditures reflect the same costs or the same benefits for each nuclear power reactor. Not only did the $950,000 annual fee apply to every power reactor licensee, without regard to whether (or to what extent) the licensee seeks the services of the Commission during the course of the year, but it likewise did not vary with any burden that the individual licensee places on the Commission.
NRC’s studied inattention to the “benefits” calculation represents a dramatic break with IOAA methodology, a cleavage unsupported by the limiting language that Congress fashioned in COBRA. And therein lies the rub; COBRA, in my reading of it, mandates that NRC consider this side of the calculation.
First, COBRA authorizes the NRC to collect fees only for a “regulatory service provided.” Whatever its precise meaning, this phrase, fairly read, conveys the need for some sort of “benefits” calculation. It requires the agency to make some assessment of which (and how much) licensees are benefitted by its “regulatory services.” In broadly defining the basis for NRC’s assessments, Congress must have meant for the NRC to take an expansive approach, one that would capture generic regulatory services. See supra p. 777. To this extent, I am in happy accord with my colleagues. But at the same time, I cannot agree that Congress meant that anything goes. This language, rather, is meant to require the NRC to weigh the benefits of its regulatory services to set fees for licensees in a fashion that takes those benefits into account.
Second, in enacting COBRA, Congress was not writing on a blank slate. It was legislating against the backdrop of a well-established body of judicial decisions construing fee statutes in general and the IOAA in particular. In adopting terms such as “regulatory service provided” and “reasonably related,” Congress can reasonably be assumed to have adopted at least some of the gloss placed on those terms through the decisional law.6 Our decisions under the IOAA have frequently referred to fee recovery for services provided by regulatory agencies, and we have often employed the phrase “reasonably related” to articulate the relationship that must obtain between costs to the regulator and benefits to the regulated. See National Cable Television Ass’n v. FCC, 554 F.2d 1094, 1106-07 (D.C.Cir.1976); Electronic Indus. Ass’n v. FCC, 554 F.2d 1109, 1114-15 (D.C.Cir.1976); National Ass’n of Broadcasters v. FCC, 554 F.2d 1118, 1133 (D.C.Cir.1976); Capital Cities Communications, Inc. v. FCC, 554 F.2d 1135, 1138 (D.C.Cir.1976).7
*391Third, although admittedly of highly limited relevance because of its post-enactment status,8 Senator Simpson, one of the Senate managers of COBRA, explained after the bill’s passage that “[COBRA] requires the Commission to demonstrate on a licensee-specific basis that the fee to be assessed is reasonably related to the actual cost of the regulatory service provided____” 132 Cong.Rec. S16.964 (daily ed. Oct. 17, 1986) (emphasis added). Indeed, Senator Simpson was vociferously critical of the Commission’s handling of the per capita fee assessment:
[B]y adopting a flat annual fee for all nuclear power reactors — other than those not authorized to operate — regardless of the variance in the regulatory service that would be provided to each individual licensee and the corresponding variance in the cost of providing such service — the Commission has virtually disregarded the directive contained in [COBRA]. Indeed, the “back calculation” that the Commission has undertaken in order to reach this flat fee ... is about as far from what was contemplated under [COBRA] as I can possibly imagine. If the fee reached through this bizarre calculation has any relation to the actual regulatory costs for individual licensees, it is only by coincidence, Mr. President — and certainly it is not the kind of determination that I think that Congress had in mind when it enacted [COBRA].
132 Cong.Rec. S16963 (daily ed. Oct. 17, 1986) (emphasis added).
Finally, it cannot reasonably be thought that the Commission is somehow unable to attend to the benefit side of the equation. NRC’s recent efforts to link more closely regulatory costs incurred to benefits received by individual licensees belie any “impossibility” defense. In response to an inquiry from the Committee on Environment and Public Works, the NRC Chairman wrote:
*392The Commission believes that it can develop a revised [fee rule under COBRA] that provides for tying fees more closely to the costs of providing specific services rendered to specific licensees.... The Commission, after further reflection, now recognizes that it would be preferable to base its annual fee on the principle that those licensees who require the greatest expenditure of NRC resources should pay the greatest annual fee.
Letter of Apr. 10, 1987, WEP Brief, Appendix C, Question 12.9
In sum, traditional tools of statutory construction lead me to conclude that, in the absence of indications to the contrary, Congress could not reasonably have intended to work such a complete break from judicial decisions interpreting the pivotal terms that Congress deliberately employed. Although Congress intended to go beyond the IOAA, it just cannot be that the Article I branch meant to permit the Commission to take such grossly nontailored action.
II
But there is another, and highly fundamental, reason for insisting that the agency move more cautiously than it did in interpreting COBRA. Whatever the boundaries of the constitutional limits on Congress’ ability to delegate its taxing power, the fact that a constitutional issue looms menacingly in the background counsels in favor of a narrower construction of COBRA than that embraced by the Commission.10 As the FPL petitioners aptly put it, “it is ... wholly inappropriate to assume that, without having made its intention express, Congress empowered the NRC to enter into what, at best, must be regarded as a constitutionally unchartered area.” FPL Brief at 26.
Fortunately, there are some familiar markers to assist us in finding our way. The identical situation confronted the Court in NCTA when it first addressed the IOAA. In that case, the Supreme Court supplied a narrowing construction in order to avoid constitutional problems arising out of the lack of intelligible standards contained in the text of the Act itself.11
The difficult question in such eases is where to draw the lines. If Congress had enacted a statute directing the NRC to collect fees up to the constitutional limits of its authority, then we would be forced to define those limits. But that is not what Congress enacted; rather, as I have al*393ready tried to show, Congress saw fit to enact a statute expressly containing (by reference to decisional law under the IOAA) limiting language. In particular, Congress’ use of a two-pronged articulation of the standard is interpretively significant. It suggests that Congress meant for the NRC, in making its assessment calculation, to weigh both benefit to the regulated entity and cost to the NRC. NRC’s self-confessed inattention to the benefit side of the equation is fatal to its interpretation.12
Once again, I am not suggesting that COBRA incorporates precisely the standards of the IOAA.13 Nor am I suggesting that scientific precision is required. To the contrary, a fair approximation of the benefit inuring to the licensee would have sufficed.14
NRC has done nothing in this respect. It has failed entirely to seek to allocate among the various individual recipients the costs of regulation that it seeks to recover through COBRA, or to explain why it cannot make such an allocation. The Commission failed to consider any of a number of methods by which it might have drawn a closer fit between benefit and burden. By way of modest suggestion, these include (1) basing the annual fee upon the level of a licensee’s fee assessments under the IOAA as an approximation of that licensee’s use of regulatory services; or (2) basing the fee upon the amount of time spent by the NRC on each licensee or the number of regulatory matters pending. See WEP Brief at 35, 43.15
In failing to consider such linkages, the NRC effectively reads out of the statute standards requiring its fees to be “reasonably related” to “regulatory services provided.” It is singularly unlikely that each licensee — regardless of differences in demand for regulatory services, plant size, plant age and other factors — derives roughly the same benefits as every other *394licensee.16 But that remarkable, egalitarianism-with-a-vengeance perspective is precisely the assumption, undergirding the NRC’s fee regime. The agency’s statutory interpretation thus lacks the “reasoned explanation” required by Rettig v. Pension Benefit Guaranty Corp., 744 F.2d 133, 151 (D.C.Cir.1984) (“a reviewing court must determine both whether the interpretation is arguably consistent with the underlying statutory scheme in a substantive sense and whether “the agency considered the matter in a detailed and reasoned fashion” (citing Chevron, 467 U.S. at 865, 104 S.Ct. at 2793)). See also Continental Air Lines v. Department of Transportation, 843 F.2d 1444, 1451-53 (D.C.Cir.1988) (discussing the application of a “reasoned decisionmaking” requirement to agencies’ statutory interpretations).
Thus constrained to conclude that the NRC’s interpretation of the statute was unreasonable within the meaning of Chevron and its progeny, I would grant the petition for review.
. In 1976, this court decided a series of four companion cases providing substantial gloss on the Supreme Court decisions noted in the text: National Cable Television Ass'n v. FCC, 554 F.2d 1094, Electronic Indus. Ass'n v. FCC, 554 F.2d 1109, National Ass’n of Broadcasters v. FCC, 554 F.2d 1118, Capital Cities Communications, Inc. v. FCC, 554 F.2d 1135.
. See, e.g. APL Brief at 19 ("In order to be sustained by this Court, the annual user fees that the NRC adopted pursuant to COBRA must be shown to reflect the cost to the government of providing specific benefits to identifiable power reactor licensees, and exclude costs that serve independent public benefits.”).
.Alternatively, petitioners argue that if Congress can delegate its taxing power, it must provide intelligible standards to guide agency discretion. Since Congress failed to provide such standards in this instance, the argument *389runs, Congress could not have intended a delegation of taxing authority — or if it did, the attempted delegation is invalid.
. Central & S. Motor Freight Tariff Ass'n is the most recent word in this circuit on the criteria to which an administrative agency must adhere in assessing fees under the IOAA. There, the court enumerated five criteria that fee assessments must meet under the IOAA:
(1) the agency must identify the specific agency activity or activities for which the fee is being assessed;
(2) the service must produce a special, private benefit;
(3) the value of that private benefit must be reasonably related to the fee;
(4) the benefit must accrue at least in part to an identifiable private beneficiary and not merely to the industry as a whole; and
(5) the service in question must produce no independent public benefit.
777 F.2d at 730.
. COBRA was enacted against a backdrop of decisional law that definitively constrained fees levied under the authority of IOAA. Under that law, the NRC could not recover, for example, research and rulemaking costs. See, e.g., Mississippi Power & Light Co. v. United States Nuclear Regulatory Comm'n, 601 F.2d 223 (5th Cir.1979), cert. denied, 444 U.S. 1102, 100 S.Ct. 1066, 62 L.Ed.2d 787 (1980). COBRA, it seems to me, was designed to break out of those constraints on IOAA fee assessment.
. See, e.g., Blitz v. Donovan, 740 F.2d 1241, 1245 (D.C.Cir.1984) (Tamm, J.):
In construing statutes, the Supreme Court has found it "always appropriate to assume that our elected representatives ... know the law____” Cannon v. University of Chicago, 441 U.S. 677, 696-97 [99 S.Ct. 1946, 1957-58, 60 L.Ed.2d 560], Furthermore, "Congress is deemed to know the ... judicial gloss given to certain language and thus adopts the existing interpretation unless it affirmatively acts to change the meaning.” Florida National Guard v. FLRA, 699 F.2d 1082, 1087 (11th Cir.), cert. denied, 464 U.S. 1007 [104 S.Ct. 524, 78 L.Ed.2d 708] (1983).
See also 2A Sutherland, Statutory Construction § 49.09 at 400 (4th ed. 1984).
. The NRC argues that construction of the word “fee” as used in COBRA should not be controlled by cases construing the IOAA. (As I have already stated, with that much I agree. But that is not to say that the decisional law under the IOAA is entirely irrelevant to determining what Congress meant by terms, such as “regulatory service," that seem inherently to be terms of art.) Instead, the NRC cites us to a very different judicial gloss of the word “fee” found in the context of intergovernmental tax immunity. See NRC Brief at 40 n. 29. There, the NRC suggests, “where the distinction between a fee and a tax is all-important, the Courts have applied a different standard for fees, one much closer to that embodied in COBRA than the IOAA standard." Id. But the Supreme Court case to which the NRC cites us, Massachusetts v. United States, 435 U.S. 444, 98 *391S.Ct. 1153, 55 L.Ed.2d 403 (1978), lays down a three-pronged test that includes as one of its elements the requirement that a fee be "based on a fair approximation of use of the system." Id. at 466-67, 98 S.Ct. at 1166-67. That part of the analysis is, as we have seen, lacking in the interpretation that the NRC now defends before us.
Thus, the requirement that fees be tied to specific benefits is a requirement of fee statutes generally; it is not an IOAA idiosyncracy. Indeed, I am aware of no case (and we are cited to none) interpreting a fee statute in which the requirement of a nexus between costs and benefits has not been imposed. See, e.g., In re Jenny Lynn Mining Co., 780 F.2d 585 (6th Cir.), cert. denied, 477 U.S. 905, 106 S.Ct. 3276, 91 L.Ed.2d 566 (1986); City of Vanceburg v. FERC, 571 F.2d 630 (D.C.Cir.1977), cert. denied, 439 U.S. 818, 99 S.Ct. 79, 58 L.Ed.2d 108 (1978); see also Gillette & Hopkins, Federal User Fees: A Legal and Economic Analysis, 67 Boston U.L.Rev. 795, 843 (1987) ("Insofar as authorizations such as those conferred on the NRC disavow any linkage to particular benefits ... any relationship between fees imposed and appropriate levels of services provided may be wholly coincidental. The user fee in this situation becomes largely a revenue-raising device imposed on particular recipients of government services. Such a device, although denominated a charge or user fee, is in essence a redistributional tax.”).
Metaphysical arguments as to the distinction between a tax and a fee aside, it is critically important as a matter of statutory interpretation that Congress chose to enact a user fee (rather than purport to delegate authority to tax).
. The Supreme Court has warned time and again about the manifest dangers of resorting to legislative history. (Perhaps most noteworthy of late are recent Court opinions authored by Justices Marshall and Scalia suggesting the irrelevance of legislative history. See Burlington N. R.R. v. Oklahoma Tax Comm., 481 U.S. 454, 107 S.Ct. 1855, 1860, 95 L.Ed.2d 404 (1987); United Sav. Ass'n v. Timbers of Inwood Forest Assocs., Ltd., — U.S. -, 108 S.Ct. 626, 634, 98 L.Ed.2d 740 (1988).) And those dangers are, of course, at their height when repair is had to post-enactment statements by legislators.
Without abandoning those well-established concerns, I note by way of modest defense of my invocation of Senator Simpson's statements that Congress has of late been passing smorgasbord measures of distinctly substantive bite in omnibus pieces of appropriations legislation. After the dust settles, questions then have to be sorted out, sometimes in court. The recent action of the political branches with respect to the future of the ownership of the Boston Herald provides an instructive case in point. It may well be, then, that if legislative history is to be used (inconsistently with the sound practice of our English counterparts), then the courts should not close their eyes to what may well be, in practical terms, the only legislative history available if Congress proceeds to conduct its business in accord with more recent procedural innovations.
. Like my colleagues, I sympathize with the short time within which the NRC was required to promulgate its rule, maj. op. at 770. As NRC Chairman Zech said with respect to the proposed rule, it was "the best the staff can do at this time." 51 Fed.Reg. at 24,087; J.A. 401. Nonetheless, those constraints are insufficient to justify adoption of a rule that is contrary to law. And, with the inestimable benefit of hindsight, we now know that the agency can do better.
. The Commission itself expressed doubts about the constitutionality of COBRA. In its July 198Ó report to Congress, the Commission stated:
[Additional clarifying legislation should be proposed because the standards set forth in [COBRA] are unclear. ... Because it appears that the lack of clear statutory language that definitively distinguishes [COBRA] from the IOAA, the legal problems identified by the Supreme Court in distinguishing the IOAA fees from taxes may be also present in [COBRA].
Accordingly, the Commission recommends that the Congress enact clarifying legislation____
J.A. at 445.
When Congress then turned a deaf ear to the Commission's entreaties, the NRC responded by, in effect, throwing up its hands and leaving it to the courts to strike down the regulations.
.See also Industrial Union Department, AFL-CIO v. American Petroleum Institute, 448 U.S. 607, 100 S.Ct. 2844, 65 L.Ed.2d 1010 (1980) (adopting a narrowing construction in the course of holding invalid the agency’s promulgation of standards); L. Tribe, American Constitutional Law, § 5-17 at 366 n. 15 (1988):
Exercises of general taxing power are, as a practical matter, not susceptible to ordinary methods of judicial review. See generally Ely, Legislative and Administrative Motivation in Constitutional Law, 79 Yale L.J. 1205, 1235 (1970). The same discretionary character of taxation that ordinarily shields it from effective judicial review would also make legislative oversight of a delegate difficult. Therefore, it seems likely that the taxation power, if it is to be exercised legitimately, may be exercised only by Congress itself.
. Indeed, the court’s treatment of the benefit analysis is troubling in another way as well. The court holds that there is "no requirement that these generic costs must be reduced by a portion artifically allocated to public benefit, so long as the fees charged are ‘reasonably related’ to services provided the feepayers and do not exact payment for an independent public benefit. See Central & S. Motor Freight, 777 F.2d at 729." Maj. op. at 769-70 (emphasis in original). But there is nothing in the record to indicate that the agency made any effort to isolate fees that amounted to a payment for an independent public benefit.
. APL interprets our cases under the IOAA to adopt the following rule for analyzing the benefit conferred on the utility:
The NRC's annua! fee is improper if it is not a "reasonable approximation" of costs to provide special benefits to individual recipients, and is invalid to the extent the NRC’s fee "unreasonably exceeds the value of the special services for which it is charged.” NCTA II, 554 F.2d at 1106. The fee is also improper unless it is equivalent, "or at least proportionate to," the value of the benefit conferred, City of Vanceburg v. FERC, 571 F.2d 630, 643 (D.C.Cir.1977), cert. denied, 439 U.S. 818, 99 S.Ct. 79, 58 L.Ed.2d 108 (1978); the fee must be reasonably related to the individual costs of services and to the total costs for particular segments of recipients. NCTA II, 554 F.2d at 1108.
APL Brief at 31.
. The Court characterizes petitioners’ conten- ' tion that there must be a nexus between fees charged and benefits received as a “‘fairness’ argument” and notes with strong approval Commission attempts to revise the fee schedule. Maj. op. at 770. Similarly Commissioner Zech, in his statement of separate views published with the Notice of Proposed Rulemaking, urged the Commission to "move toward an annual charge which would take into account, first, the amount of electricity produced by each licensee and, second, the resources which are expended by the Commission which fairly reflect the cost to the Commission of providing inspection and other services to the licensee.” 51 Fed.Reg. 24078, 24087 (Jul. 1, 1986); J.A. at 401. What my colleagues and Commissioner Zech characterize as a matter of "fairness” or "sound policy,” however, I believe is statutorily mandated.
.NRC protests that it duly considered and rejected in reasoned fashion comments suggesting that the fee should be based on the power rating in thermal megawatts for each reactor. But the NRC’s response to that suggestion was strictly in terms of cost to the NRC and the ability of licensees to pay; there was no analysis with respect to the benefit to licensees. J.A. at 702.
In choosing the per capita approach rather than the megawatt approach in its initial Notice of Proposed Rulemaking, the Commission stated that "the Commission has selected [the per capita approach] for the proposed rule because there is no clear correlation between reactor size and NRC regulatory costs.” J.A. at 397.
. In 1985, the 90 nuclear power reactors in the United States ranged in size from 165 to 3,833 thermal megawatts. Inspection, licensing, and individual operator fees for these reactors ranged from $27,997 to $523,172. J.A. at 643-46.