with whom Justice Marshall joins, concurring in part and dissenting in part.
I join Part V of the Court’s opinion, and I concur in the remainder to the extent it upholds the GAO’s right to inspect Merck’s “direct” cost records.1 I dissent to the extent the Court refuses to allow the GAO access to any of Merck’s “indirect” cost records.
The statutory provisions at issue, 10 U. S. C. § 2313(b) and 41 U. S. C. § 254(c), clearly were intended to allow the GAO the right to a reasonable degree of access to contractors’ records needed to determine whether prices charged to the Government were excessive. Of course, this right was not intended to be unlimited; the Court correctly identifies a congressional intent to protect private contractors from “officious governmental intermeddling.” Ante, at 835.
*846Unfortunately, for the conceded purpose of creating a bright-line test, ante, at 841, n. 18, the Court goes astray by adopting a rule that flatly bars the GAO from access to all indirect cost records pertaining to most fixed-price contracts, regardless of how urgent the need for them might be. The Court frankly admits that its rule may deny the GAO access to cost records critical to an assessment of the fairness of the contract price, thereby impeding the GAO’s ability to protect the public against wasteful Government expenditures. Ante, at 842-843. These undesirable consequences could be avoided, without sacrificing the contractors’ right to be free from unwarranted GAO “snooping,” by holding the Government to the burden of showing that requested records likely had a direct and substantial impact on the price charged to the Government and thus are “directly pertinent” to the contract.
In each of the four contracts involved here, the United States agreed to purchase certain pharmaceutical products from Merck at a fixed price. In each instance, Merck proposed a contract price based on its catalog or “market” price, and the Government contracting officer accepted the proposal without any “haggling” or other negotiation as to price. Each of the contracts contains the statutorily mandated provision allowing the GAO the right to inspect Merck’s books and records that are “directly pertinent” to the contract.
The GAO now seeks to examine those Merck records that indicate the cost to Merck of the goods sold to the Govern-' ment. The GAO deems such an examination necessary to carry out its statutory duty to “investigate ... all matters relating to the receipt, disbursement, and application of public funds,” and to “make recommendations looking to greater economy or efficiency in public expenditures.” 31 U. S. C. § 53(a).
By inspecting Merck’s cost records, the GAO hopes to be able to estimate whether the contract price paid by the Gov*847ernment was a fair one. The GAO has confirmed by experience the common-sense observation that the mere “fact that a product is listed in a manufacturer’s catalog and offered to any customer is no assurance that . . . the standard catalog price is reasonable. ”2 If the GAO’s inspection were to reveal that Merck’s prices were unreasonably high, the GAO presumably would recommend to the contracting agencies that they “negotiate prices more carefully or . . . obtain greater competition in future similar procurements,”3 or that they take other action “looking to greater economy or efficiency,” 31 U. S. C. § 53(a), in future expenditures.
The Court concludes, however, that, despite the inclusion of the access-to-records provision in the contracts, the GAO has no right to inspect any of Merck’s indirect cost records to determine how much it cost Merck to produce the products sold to the Government. This holding exalts the contractors’ privacy interest to such a degree that it displaces the GAO’s right to inspect records unquestionably needed for an accurate determination of the fairness of the contract price. Congress did not intend this order of priority of interests. What Congress did intend was that contractors be spared the burden of unwarranted intrusions. Congress did not want the GAO to be irresponsibly “snooping” into records lacking relevance to the question whether the Government paid a fair price for the products it purchased. But Congress did not wish to deny the GAO access to records legitimately needed to detect waste, extravagance, and ineffective procurement. To the extent the GAO can prove that some or all of a contractor’s indirect cost records fall into this latter category, the GAO’s right of access should be sustained.
*848I begin with the language of the statute. Jackson Transit Authority v. Transit Union, 457 U. S. 15, 23 (1982); Touche Ross & Co. v. Redington, 442 U. S. 560, 568 (1979). The legislation at issue requires unadvertised Government contracts to include a clause allowing the GAO to examine any of the contractor’s books, documents, papers, or records “that directly pertain to, and involve transactions relating to, the contract_” 10 U. S. C. § 2313(b). See 41 U. S. C. § 254(c).
“ ‘[I]t is hard to imagine anything more directly related to a contract than the cost of producing the items covered by it or the matters going into the makeup of the price.’” Eli Lilly & Co. v. Staats, 574 F. 2d 904, 913 (CA7), cert. denied, 439 U. S. 959 (1978). Accord, SmithKline Corp. v. Staats, 668 F. 2d 201, 208-209 (CA3 1981), cert. pending, Nos. 81-2082, 81-2268; Hewlett-Packard Co. v. United States, 385 F. 2d 1013, 1016 (CA91967), cert. denied, 390 U. S. 988 (1968). The Court does not contend otherwise. Indeed, all the Court has to say about the literal statutory wording is that it requires “some close connection between the type of records sought and the particular contract.” Ante, at 831. But there is, of course, no reason why the records of all indirect costs inherently lack the requisite “close connection. ” As the Court fully recognizes, ante, at 842-843, in some instances indirect costs have a critical bearing on the makeup of the contract price. The Government should at least be allowed an opportunity to prove that such is the case, and, to the extent it succeeds in this endeavor, the GAO should be allowed access.
H-1 f — I HH
Even if, contrary to my belief, the statutory language is somehow regarded as ambiguous, resort to the legislative history further refutes the Court’s position. The legislative history of the access-to-records provisions is relatively brief and to the point. It demonstrates beyond doubt that Con*849gress authorized the GAO to examine all of a contractor’s books legitimately needed to evaluate Government procurement techniques by ascertaining whether the Government had paid a reasonable price for the contractor’s goods or services.
Representative Hardy, the bill’s sponsor, indicated that the bill was intended to improve the adequacy of Government procurement techniques in various ways. He expressly remarked:
“The major purposes of this bill are twofold: One, to give the Comptroller General the proper tools to do the job the Congress has instructed him to do; and, two, to provide a deterrent to improprieties and wastefulness in the negotiation of contracts.” 97 Cong. Rec. 13198 (1951).
As noted supra, at 846, Congress has instructed the GAO to “investigate ... all matters relating to the receipt, disbursement, and application of public funds,” and to “make recommendations looking to greater economy or efficiency in public expenditures.” 31 U. S. C. § 53(a).4
Representative Hardy early explained to his colleagues that normal procurement procedures called for competitive bidding but that procurement by negotiation was sometimes necessary. In the latter context, where there is no competitive bidding to “operat[e] as a brake on the price which a contractor can demand from the Government,” Representative Hardy saw the need to establish “every reasonable safeguard against waste and extravagance in the spending” of Government funds. 97 Cong. Rec., supra, at 13198. He felt that, no matter how “conscientious and honest” the Government representatives might be, the contractor’s representatives would, in the great majority of cases, have a tremendous ad*850vantage from the standpoint of both training and experience. Ibid. Thus, there was “every chance in the world that the Government [would] come out on the short end of the deal,” and Representative Hardy deemed it necessary to “at least enable the [GAO] to check the transaction, both from the Government records and the contractors’ books.” Ibid.
The debate continued two days later when Representative Hardy proposed an amendment that would have allowed agency heads the discretion to omit the access-to-records clause from contracts with foreign contractors, id., at 13371, and Representative Harvey proposed an amendment that would have exempted “a manufacturer or processor who is a supplier of material to a primary contractor and who is not a subcontractor” from the scope of the bill’s coverage. Id., at 13376. Both of these proposals were ultimately defeated, but, during the lively debate on the proposed amendments, several Members of Congress stated without contradiction that the bill would allow the GAO extremely broad authority to examine records. For example, Representative Harvey asked whether, if the bill became law, a subcontractor of a primary Government contractor “would be subject to having all of his books opened up for inspection by Government officials.” Id., at 13372 (emphasis added). Representative Hardy replied that it would, unless the subcontractor only supplied some “casual item” in connection with the performance of the contract. Ibid. Based on this understanding, Representative Harvey later argued that his limiting amendment was needed, because otherwise “every manufacturer . . . of. . . goods that eventually find their way into defense production ... is going to have to supply all the answers to the GAO on everything he manufactures.” Id., at 13376. “[E]very section of his books will have to come under the complete scrutiny of the GAO.” Ibid. In response, Representative Hardy did not dispute this characterization of the scope of the GAO’s authority, but he nevertheless opposed the Harvey amendment, because it “would make it impossi*851ble frequently to obtain information which would be vital in the study of a contract.” Ibid.
Representative Hoffman, a strong opponent of the bill, several times during the debate observed that the “GAO under this bill can go into the books of [contractors] and ask and get from them anything and everything they want.” Id., at 13373. He indicated his belief that the bill would allow the GAO “to snoop into [a contractor’s] books and find out what [the goods or services] cos[t] or what will be a fair price or what profit we make.” Ibid. See also id., at 13375, 13377. Standing alone, of course, the statements of an opponent of the bill, such as Representative Hoffman, would not carry much weight,5 but here, even though all comments pro and con were made in the midst of a free-wheeling debate, the proponent of the bill, Representative Hardy, in no way took issue with Representative Hoffman’s view of the scope of the GAO’s authority. Representative Hardy’s essential response was that it was necessary to require contractors to afford the GAO this broad authority, and that Representative Hoffman’s fears of excessive GAO “snooping” were groundless, because the GAO would have neither the inclination nor the manpower to examine the records of every individual supplier. Id., at 13376, 13377.
The original Hardy bill required the inclusion, in negotiated contracts, of a clause allowing the GAO the right to examine any records that were “pertinent” to the contract. At the very end of the debate, Representative Hoffman proposed the amendment that added the word “directly” before the word “pertinent.” Id., at 13377. Representative Hoffman explained that he had discussed his amendment with Representative Hardy, the bill’s sponsor, and that, although the amendment was “not all that it should be,” it was the *852most that Representative Hardy would agree to. Ibid.6 Representative Hoffman then stated that the purpose of his amendment was “to limit the ‘snooping’ that may be carried on under this bill which we do not have the votes to defeat.” Ibid. At that point, Representative Hardy remarked that he had no objection to the amendment, and it was accepted without further discussion. Ibid.7
In light of Representative Hardy’s consistent position throughout the debate, it cannot plausibly be argued that he agreed to the Hoffman amendment with the understanding that it effected a drastic reduction in the scope of the bill’s coverage or purpose. As outlined above, Representative Hardy continuously spoke of the need to provide a mechanism to combat waste and extravagance in federal procurement, and he vigorously and successfully opposed the Harvey amendment, which would have significantly limited the bill’s *853scope. His acceptance of the addition of the word “directly” may have been largely a sop to the bill’s opponents. The most that can be said is that Representative Hardy accepted the amendment to allay concerns that the legislation “would let the GAO go into everybody’s business and look it over if they just wanted to take a look at it.” Id., at 13373 (Rep. Hoffman). The amendment gave assurance that the bill would not be used as a basis for inspection of books and records having no substantial connection with Government procurement. But the amendment definitely was not intended to bar the GAO’s access to records legitimately needed to assess the reasonableness of prices charged to the Government, and thereby to protect the Government against waste, excessive prices, and ineffective procurement.8
> HH
Despite the plain statutory language and legislative history, the Court refuses to uphold the GAO’s right of access to all cost records that are essential to an accurate determination of whether the Government wasted money by entering into these contracts. Adopting the so-called “Bristol test,”9 the Court affirms a judgment that limits the GAO’s access to records pertaining to a contractor’s
“manufacturing costs (including raw and packaging materials, labor and fringe benefits, quality control and supervision), manufacturing overhead (including plant *854administration, production planning, warehousing, utilities and security), royalty expenses, and delivery costs.”10
The Bristol test excludes
“books, documents, papers, or records with respect to research and development, marketing and promotion, distribution and administration (except to the extent such data may be included in the cost items listed above).”11
The Court describes this test, perhaps inaccurately,12 as being based on a “direct costs v. indirect costs” or an “allocated costs v. unallocated costs” dichotomy.
The courts that have adopted this test have no doubt been influenced by a need to come up with some form of reasonable limitation on the broad access demand the GAO has made. The GAO claims it has the right to examine records pertaining to every cost “defrayed from commingled general revenues that include the Government’s payments under the contract.” I agree that the GAO’s demand is somewhat over-broad; the Court correctly observes, ante, at 836, that it would require Merck to allow inspection of cost records totally unrelated to the Government contracts, such as records of expenditures for raw materials used to manufacture products other than those sold to the Government under the contracts.13 However, I am not convinced that the proper conclusion is to limit the GAO to the cost records allowable under the Bristol test.
The Bristol court adopted its standard solely on the basis of the cost records that the contractor was willing to disclose *855to the GAO. The court felt that the contractor’s offer “reflected a responsible and reasonable effort to distinguish ‘directly pertinent’ matter within the meaning of the access to records clause.”14 The court thus accepted the contractor’s contention that the cost records it was not willing to disclose had “only the most general relation, if any, to the prices charged.”15
Although cost records having, at most, only an insubstantial relation to the price charged are not “directly pertinent” to the contract, it is apparent that many of the records deemed unexaminable under Bristol relate to costs that may have had a critical bearing on the prices charged, and that would be of central importance to a GAO inquiry into the fairness of these prices. In the pharmaceutical industry, it has been estimated that “direct” or “allocated” costs constitute only about nine percent of the sale price of individual products. The so-called “indirect” or “unallocated” costs— primarily research and development, advertising and other promotion, general administrative expenses, and taxes — and profit are much larger and economically more significant.16 Yet, under the Bristol test, the GAO is denied access to all records in this category, thus making it impossible for the GAO to make an accurate assessment of the fairness of the prices and thus the adequacy of the Government’s procurement technique.17
*856In my view, the correct rule in a case of this nature is that any books or records that bear directly on the question whether the Government paid a fair price for the goods or services it purchased are “directly pertinent” to the contract of purchase. Under this test, for example, the cost records of an advertising campaign to promote only the particular products sold to the Government, or a research project designed specifically to develop or improve these products, would clearly be “directly pertinent.”18 On the other hand, *857records of advertising campaigns and research projects involving only unrelated products would lack the requisite degree of pertinence. Of course, in many instances a commercial advertisement or a research project will be designed to promote or develop both products sold to the Government and other, unrelated products. With respect to cost records of efforts such as these, there might be some close questions as to whether such records are “directly pertinent” to the Government contracts. If, however, the GAO could bear the burden of proving that the records are of costs that likely had a direct and substantial impact on the price charged to the Government under the contract, I would allow the GAO access to the records.
V
The inquiry does not necessarily come to an end once the GAO establishes that it has a statutory and contractual right to inspect particular records. In addition to the statutory *858“directly pertinent” limitation, the GAO’s right of inspection is further circumscribed, and the contractors’ right to privacy is further protected, by constitutional standards, such as the Fourth Amendment reasonableness requirement.
Where, as here, the GAO wishes to see a contractor’s records and the contractor declines to accede voluntarily to a GAO request, the GAO must issue an administrative subpoena. If the contractor refuses to comply with the subpoena, the GAO must apply to a district court for enforcement of the subpoena. 31 U. S. C. §54(c) (1976 ed., Supp. V).19
Once in the district court, a contractor such as Merck has the benefit of all of this Court’s jurisprudence limiting the bounds of an agency’s right to demand the production of a private entity’s records. Essentially, in assessing an agency’s application for enforcement of an administrative subpoena, we have insisted that the agency’s demand be reasonable. The general rule is that “when an administrative agency subpoenas corporate books or records, the Fourth Amendment requires that the subpoena be sufficiently limited in scope, relevant in purpose, and specific in directive so that compliance will not be unduly burdensome.” See v. City of Seattle, 387 U. S. 541, 544 (1967). See United States v. Morton Salt Co., 338 U. S. 632, 652-653 (1950); Oklahoma Press Publishing Co. v. Walling, 327 U. S. 186, 208 (1946); 1 K. Davis, Administrative Law Treatise §4:15 (2d ed. 1978). This standard is a flexible one that takes into account the extent to *859which the public interest will be served if the subpoena is enforced. See v. City of Seattle, supra, at 545.
In the present case, Merck has claimed that compliance with the GAO’s demand would entail substantial expense and disruption of its operations. This claim is based on evidence that the proposed GAO inspection would require Merck to allow an entire team of GAO auditors to remain on site at Merck for over two years.20 This, of course, is a matter for first-instance determination by the District Court, but, if the proposed GAO inspection would in fact cause such a high degree of interference with Merck’s business, a credible argument could be made that compliance would be unreasonable and unduly burdensome and that the GAO’s access should therefore be limited in some way.
h — I >
In view of the foregoing, I would remand these cases to the District Court, with instructions to uphold the GAO’s request for access to Merck’s “direct” and “indirect” cost records, but only to the extent that: (1) the records sought by GAO related to costs that likely had a direct and substantial impact on the prices charged to the Government under the contracts; and (2) the request is reasonable in scope and would not unduly burden Merck. To the extent the GAO’s demand conforms to these statutory and constitutional standards, Merck should be required to allow the GAO’s examination to proceed.21
The Court correctly rejects Merck’s contention that none of its cost records are subject to inspection by the GAO. Merck’s theory has been emphatically rejected by every Court of Appeals that has considered it. In addition to the opinion below in the present case, Merck & Co. v. Staats, 214 U. S. App. D. C. 418, 665 F. 2d 1236 (1981), see SmithKline Corp. v. Staats, 668 F. 2d 201 (CA3 1981), cert. pending, Nos. 81-2082, 81 — 2268; United States v. Abbott Laboratories, 597 F. 2d 672 (CA7 1979); Eli Lilly & Co. v. Staats, 574 F. 2d 904 (CA7), cert. denied, 439 U. S. 959 (1978); and Hewlett-Packard Co. v. United States, 385 F. 2d 1013 (CA9 1967), cert. denied, 390 U. S. 988 (1968). Apparently recognizing the untenability of the argument advanced here by Merck, the pharmaceutical manufacturer in Bristol Laboratories Division of Bristol-Myers Co. v. Staats, 428 F. Supp. 1388 (SDNY 1977), aff’d, 620 F. 2d 17 (CA2 1980), aff’d by an equally divided Court, 451 U. S. 400 (1981), did not even dispute that it was “obliged by the terms of the contracts to provide access to records of its manufacturing costs, records which relate to the pricing of the products delivered and records required to verify all data obtained during the course of the review.” 428 F. Supp., at 1389.
Hearings on Relation of Cost Data to Military Procurement before the Subcommittee for Special Investigations of the House Committee on Armed Services, 88th Cong., 1st Sess., 8 (1963) (remarks of Robert F. Keller, General Counsel of the GAO).
App. 23a (affidavit of Paul G. Dembling, General Counsel of the GAO).
In addition, “[t]he Comptroller General is authorized and directed to make an expenditure analysis of each agency in the executive branch of the Government . . . which . . . will enable Congress to determine whether public funds have been economically and efficiently administered and expended.” 31 U. S. C. § 60.
See, e. g., National Woodwork Manufacturers Assn. v. NLRB, 386 U. S. 612, 639-640 (1967); Schwegmann Bros. v. Calvert Distillers Corp., 341 U. S. 384, 394 (1951).
In fact, it appears that it was Representative Hardy, not Representative Hoffman, who proposed the language of the amendment that limited the Government to “directly pertinent” records. Representative Hoffman stated that this language was “the best he [Representative Hardy] could think of.” 97 Cong. Rec. 13877 (1951) (emphasis added). Apparently what happened is that Representative Hoffman and Representative Hardy had a private meeting, during which Representative Hoffman suggested the need for a limiting amendment. Representative Hoffman had stronger language in mind, which Representative Hardy refused to accept. Representative Hardy must have then proposed the “directly pertinent” language to placate his colleague, and Representative Hoffman, though not satisfied, “was rather forced to accept it and to agree with him.” Ibid. Thus, although the Court is correct in stating, ante, at 833, that the amendment “circumscribe[d] the inquiry the Comptroller General was authorized to undertake,” it is clear that this circumscription was only to the extent agreeable to Representative Hardy, who, as shown by his opposition to the Harvey amendment, and his other remarks throughout the debate, clearly would not have agreed to a sharp reduction in the scope of his bill.
Immediately after passage of the Hoffman amendment, the House passed the bill, as amended, 97 Cong. Rec. 13378 (1951). On the following day, the Senate passed it without any debate whatsoever. Id., at 13411.
Accord, Merck & Co. v. Staats, 214 U. S. App. D. C., at 431, 665 F. 2d, at 1249 (Mikva, J., concurring in part and dissenting in part); Eli Lilly & Co. v. Staats, 574 F. 2d, at 916; Comment, 92 Harv. L. Rev. 1148, 1157-1158 (1979) (“The court [in Eli Lilly] correctly rejected Lilly’s argument that the insertion of the word ‘directly’ into the ‘directly pertinent’ formula sharply narrowed the scope of inquiry to be allowed the GAO”).
The test was first adopted by the District Court in Bristol Laboratories Division of Bristol-Myers Co. v. Staats, 428 F. Supp. 1388 (SDNY 1977), aff’d, 620 F. 2d 17 (CA2 1980), aff’d by an equally divided Court, 451 U. S. 400 (1981).
App. to Pet. for Cert, in No. 81-1273, p. 39a. See ante, at 829-830.
App. to Pet. for Cert, in No. 81-1273, p. 40a. See ante, at 830.
The Bristol court itself stated that the costs records held examinable under its test are “by no means . . . limit[ed]” to direct costs. 428 F. Supp., at 1391.
The Court is also correct in concluding, ante, at 837-839, that, because the GAO’s interpretation of the statute has not been consistent, it is not entitled to particular deference.
428 F. Supp., at 1391.
Id., at 1390.
Eli Lilly & Co., supra, at 913; Merck & Co. v. Staats, supra, at 429, 665 F. 2d, at 1247 (Mikva, J., concurring in part and dissenting in part); Reekie, Price & Quality Competition in the United States Drug Industry, 26 J. Indus. Econ. 223, 235 (1978); Rucker, Public Policy Considerations in the Pricing of Prescription Drugs in the United States, 4 Int’l J. Health Services 171, 173 (1974).
For example, the Bristol test might not allow the GAO to examine a contractor’s records of advertising costs. One would imagine that, if the GAO were aware that a great percentage of the cost of the products of a certain company went to support a large advertising campaign, rather *856than, say, to maintain quality control, the GAO might recommend to the contracting agency that it not deal with that company in the future. Yet, under the Bristol test, the GAO might not be able to obtain the information needed to make such a recommendation.
On this point, the Court and I appear to be in agreement, although this fact is far from clear from the Court’s opinion. At one point, the Court defines indirect costs as “costs incurred in the areas of research and development, marketing and promotion, distribution, and administration, which are not directly attributable to a particular product.” Ante, at 836, n. 11 (emphasis added). See also ante, at 841 (indirect costs are those requiring “somewhat arbitrary accounting allocations”). This definition indicates the Court agrees with me that records of costs in these areas that are directly attributable to the products sold to the Government — such as the cost records of an advertising campaign or a research project designed to promote or develop only the specific products sold to the Government— should be considered “direct costs” that are subject to examination by the GAO.
Potential confusion arises, however, in view of the fact that the Court affirms the judgment below in its entirety, even though that judgment can perhaps be read as denying the GAO access to product-specific records in areas such as advertising and research. The judgment denies access to all records in areas such as research and advertising, “except to the extent such data may be included in the cost items [to which the GAO is allowed access].” Merck & Co. v. Staats, 214 U. S. App. D. C., at 428, 665 F. 2d, at 1246. Product-specific records in areas such as advertising and research do not appear to be included among any of the access-permitted cost items mentioned in the judgment, unless it can be said that they constitute records of “manufacturing costs.” See ibid.
The Court adds to the uncertainty with its statement, ante, at 840, that the GAO should be allowed access to records in such areas as research and advertising “to the extent the contractor has allocated these costs as attrib*857utable to the particular contract.” This language suggests that the contractor can avoid a GAO examination of even product-specific records in these areas by arbitrarily refusing to attribute these costs to the specific products.
In light of the Court’s emphasis on the need for “[b]right-line rules upon which the parties’ expectations may be firmly established,” ante, at 841, n. 18, it is surprising that the Court is not precise on this point. My “objection” is not, as the Court indicates, ante, at 840, n. 16, that the Court ties the GAO’s ability to inspect indirect cost records to the contractor’s accounting practices. Rather, my concern is with the tension between the Court’s holding and the judgment it affirms without modification. Under the Court’s rationale, product-specific research and advertising costs are “direct costs” and are subject to GAO inspection regardless of the contractor’s accounting practices. It is not clear that the judgment below permits this result. Nor is it clear that the judgment would allow the GAO to conduct a “windfall” inspection of contractor-allocated indirect cost records, which the Court expressly sanctions. Ibid. The potential problem arises mainly because the opinion speaks generally of a direct-indirect dichotomy, although it affirms a judgment that may not break down neatly along these lines. In any event, I read the Court’s opinion as allowing access to all records that are or must fairly be attributed to the specific products sold to the Government.
The GAO has had this subpoena power only since 1980. See Pub. L. 96-226, § 102(c), 94 Stat. 312 (codified as 31 U. S. C. § 54(c) (1976 ed., Supp. V)). Prior to 1980, the GAO could only sue for specific performance of its contractual right of access. See, e. g., Hewlett-Packard Co. v. United States, 385 F. 2d 1013 (CA9 1967), cert. denied, 390 U. S. 988 (1968). The GAO can still opt in favor of this latter, slower course, but the Government does not make any claim that its rights in a specific performance action would be any different from those in an action for judicial enforcement of an administrative subpoena. See Tr. of Oral Arg. 29-30.
At the time the GAO issued its demand for access to Merck’s records, it made identical demands to five other drug companies. Apparently, only one of the six companies, Hoffman-LaRoche, Inc. (Hoffman), voluntarily acceded to the GAO’s request. According to Merck, a team of GAO auditors remained on site at Hoffman from July 1975 to July 1977 without completing the review, and, in July 1977, Hoffman terminated its voluntary participation. Of course, even if Merck’s claim as to the Hoffman precedent is accurate, Merck would have to prove that a GAO review of its records would require similar or greater disruption.
The Court does not explicitly discuss the degree of access permissible to the GAO in instances where the Government enters into a fixed-price *860contract in reliance on cost representations by the contractor. ' There can be little doubt, however, that, in such an instance, the GAO is entitled to inspect all of the contractor’s cost records, both direct and indirect. Under these circumstances, the contract is “cost-based” and “the public interest served by permitting the GAO to inspect records supporting these claims clearly outweighs any privacy interests the contractor possesses in those records.” Ante, at 841, n. 17.