FEINBERG, Chief Judge:
This is an appeal from an order of the United States District Court for the Southern District of New York, Kevin T. Duffy, J., enforcing an Internal Revenue Service (IRS) summons to Arthur Young & Co. (AY). The summons, issued pursuant to 26 U.S.C. § 7602,1 directed AY to produce all files related to its client, Amerada Hess Corp. (Amerada), for whom AY served as independent auditor. In response to an enforcement proceeding initiated by the IRS, Judge Duffy, in a decision reported at 496 F.Supp. 1152 (S.D.N.Y.1980), ordered the production of all the items requested by the IRS with two exceptions discussed below. The parties consented to a stay of production pending resolution of this appeal. In this court, AY challenges the district court’s order, objecting primarily to the request for [214]*214its audit workpapers file and for its tax pool (tax accrual workpapers) file. Amerada, utilizing the procedure available under 26 U.S.C. § 7609,2 has intervened to challenge production of the tax accrual workpapers file. Because we find that the audit workpapers are relevant to determining Amerada’s tax liability and thus fall within the scope of the IRS summons power, we affirm that part of Judge Duffy’s order requiring their production. However, we reverse that part of the district court order enforcing the summons of the tax accrual workpapers. Although we find that this material is also relevant to the determination of tax liability, we believe that these documents should remain confidential in order to protect the reliability of the independent audit process.
I.
The involvement of Arthur Young & Co., a firm of certified public accountants, with Amerada Hess Corp. began in November 1971, when AY was retained as Amerada’s independent auditor. Among its other duties, AY was responsible for reviewing the financial statements prepared by Amerada in satisfaction of the disclosure requirements of the federal securities law.3 An AY partner also signed Amerada’s tax returns, even though AY did not prepare the returns.
In April 1976, AY became further enmeshed in Amerada’s affairs. Around that time, it became known that many American companies had made illegal payments abroad. In response to this publicity, Amerada’s board of directors formed a special committee to investigate the possibility of Amerada wrongdoing. The law firm of Milbank, Tweed, Hadley & McCloy, and AY were engaged as independent investigators to assist the committee. This inquiry apparently, and despite the minor nature of the improprieties it uncovered, prompted the IRS in August 1977 to institute a criminal investigation of Amerada’s tax returns for the years 1972,1973, and 1974. Because a regular audit was then in progress, the IRS and Amerada designed a procedure in September 1977 to apprise the corporation of the progress of the criminal investigation. According to their agreement, Special Agent Kenneth Kalemba, who was in charge of the criminal investigation, would issue all summonses relating to that inquiry. In April 1978, a summons signed by Kalemba was issued to AY, requiring it to make all its Amerada files available to the [215]*215IRS.4 A quarter of a million pages of documents are involved in this request.
The IRS sought to enforce this summons by an order to show cause, dated October 9, 1979. Judge Duffy tested the summons against the four criteria set forth in United States v. Powell, 379 U.S. 48, 57-58, 85 S.Ct. 248, 254-255, 13 L.Ed.2d 112 (1964):
[The Commissioner] must show that the investigation will be conducted pursuant to a legitimate purpose, that the inquiry may be relevant to the purpose, that the information sought is not already within the Commissioner’s possession, and that the administrative steps required by the Code have been followed.. . .
He found that with the exception of two types of documents, the summons satisfied all the prongs of the Powell test and, accordingly, enforced it. The exceptions related to AY’s audit program and to the documents prepared by the special committee that investigated Amerada’s questionable payments. Judge Duffy held that the audit program was irrelevant because it “stands many steps removed from the question of the actual tax liability.” United States v. Arthur Young & Co., 496 F.Supp. at 1157. He found that the committee report was protected by the work-product privilege under the doctrine of Hickman v. Taylor, 329 U.S. 495, 67 S.Ct. 385, 91 L.Ed. 451 (1947), since the material was prepared under the auspices of Milbank, Tweed in anticipation of litigation, id. at 1157-58. The IRS does not appeal this portion of the district court’s order. Among the documents that Judge Duffy did order turned over are AY’s audit workpapers and tax accrual workpapers. The challenges advanced on appeal relate to these two categories.
II.
AY alone objects to the production of its audit workpapers. These documents consist almost entirely of factual data generated from the books when accountants verify the financial statements prepared by Amerada’s own personnel by spot-checking selected bookkeeping entries and records. They include third-party confirmations of transactions, as well as the auditor’s own judgments about the implications of the company’s transactions. Some of the workpapers contain material learned during confidential discussions between Amerada and AY employees. This material is not used in preparing Amerada’s tax returns.
AY makes two objections to disclosure of these documents. First, it focuses on the second Powell criterion — relevance — and it claims that the IRS bears an enhanced burden of showing relevance when it seeks documents from a third-party, that is from an entity other than the taxpayer whose return it is auditing. Thus, AY attempts to distinguish our holding in United States v. Noall, 587 F.2d 123 (2d Cir. 1978), cert. denied, 441 U.S. 923, 99 S.Ct. 2031, 60 [216]*216L.Ed.2d 396 (1979), where a summons forcing a taxpayer to produce its own internal audit workpapers was enforced, on the ground that Noall did not involve a third-party. Where, as here, a stranger to the audit is involved, AY argues that this court’s opinion in United States v. Harrington, 388 F.2d 520, 524 (2d Cir. 1968), controls and it is insufficient for the IRS merely to allege that some chance of relevance exists.
In Harrington, the target of the summons was a lawyer who had as his client the ex-wife of the taxpayer in whom the IRS was interested. In the course of enforcing the IRS summons, the court did point out that it was “particularly appropriate” that the lawyer and his client receive “judicial protection against ... sweeping or irrelevant order[s] ...” because the demand was “directed not to the taxpayer but to a third-party who may have had some dealing with the person under investigation.” Id. at 523. But AY is not a stranger to Amerada’s concerns in the same way that lawyer Harrington was to the husband taxpayer in Harrington. AY was hired to investigate Amerada’s financial affairs, and to involve itself in matters that were likely to be the object of governmental inquiry. We do not believe that Harrington dictates the result sought by AY.
Moreover, we do not understand whose interests AY seeks to protect by this claim of an enhanced burden. Amerada does not need this protection: it can intervene in its own behalf under § 7609, see generally, Kenderine, The Internal Revenue Service Summons to Produce Documents: Powers, Procedures, and Taxpayer Defenses, 64 Minn.L.Rev. 73 (1979). And, in fact, it has done so. Since Amerada does not object to production of the audit workpapers, it would be anomalous to deny enforcement on the strength of its accountant’s objection. Nor can AY assert any particular burden on it. Judge Duffy carefully protected AY from such possibilities by requiring the IRS to shoulder the costs and to inspect on site, United States v. Arthur Young & Co., 496 F.Supp. at 1160.
Once it is accepted that the usual threshold of relevance applies, it is clear that the audit workpapers pass the Powell test. In Noall, we rejected a simplistic equation of relevance with use in tax return preparation, and recognized that:
the purposes of the internal audit include the detection of overstatements or understatements of revenues or expenses, and of identifying accounting procedures that would lead to these. If the internal auditors have ascertained an understatement of revenues or an overstatement of expenses, this plainly might throw light on the correctness of the returns.
587 F.2d at 126. AY has advanced no persuasive argument to distinguish between the usefulness of the internal audit procedure at issue in Noall and the external audit at issue here. The intervening decision in Thor Power Tool Co. v. Commissioner, 439 U.S. 522, 99 S.Ct. 773, 58 L.Ed.2d 785 (1979), does not change this result. There the Supreme Court held that the accounting methods that a taxpayer uses for his own financial records cannot be dispositive of the accounting method used for tax purposes. But that does not mean that the reliability of the taxpayer’s records is not legitimately of interest to the IRS. Tax liability depends on a taxpayer’s actual revenue and expenses, not on his bookkeeping. The IRS has an appropriate interest in ensuring that the latter accurately reflects the former.
AY next maintains that the broad, generic document request issued here cannot be tolerated under the Fourth Amendment. The federal courts have repeatedly dealt with similar contentions, and have consistently rejected them. Before the IRS knows where the issues lie, it has no choice but to utilize a general summons. Similarly, before the Service knows what the documents contain, it cannot describe them with any specificity. If the target’s own opinion of relevance controlled, the entire audit process would be eviscerated. The judges in this circuit have therefore construed the summons power liberally, and allowed it to extend to widely-sweeping document re[217]*217quests. See, e.g., United States v. Noall, 587 F.2d at 127 (rejecting even in camera review as being overly cumbersome); Foster v. United States, 265 F.2d 183, 186-187 (2d Cir.), cert. denied, 360 U.S. 912, 79 S.Ct. 1297, 3 L.Ed.2d 1261 (1959); United States v. Acker, 325 F.Supp. 857, 862-63 (S.D.N.Y. 1971) (Frankel, J.). AY argues that cases such as these do not apply here because it is a third party, not the taxpayer, but we have already rejected that premise on these facts. We therefore affirm that portion of the district court’s order requiring AY to turn over the audit workpapers.
III.
AY is joined by Amerada and a host of amici curiae in challenging the power of the IRS to compel the disclosure of tax accrual workpapers. These documents are generated when an auditor verifies whether the taxpayer has accurately determined its contingent tax liability. Under the federal securities laws, a registrant must file a financial statement of its potential liabilities. Furthermore, the amici tell us that generally accepted auditing standards require an auditor to determine whether his client has put aside enough reserves to cover the contingency that, upon audit, it will owe the government more taxes than originally remitted. To make this assessment, the auditor must not only determine how the taxpayer treated his income and expenses in his tax return; he must also decide whether that treatment comports favorably with the Internal Revenue Code, the Regulations, and the case law. In areas where the law is unclear, he must predict the chances that the taxpayer’s position will be upheld by the courts — a judgment based on his knowledge of the law and his opinion of where the law is headed. The auditor must also take into account the likelihood that the client will settle the dispute — a judgment based on the auditor’s confidential and intimate knowledge of the client. This process frequently requires the auditor to elicit and engage in speculation as to positions that might be taken by the IRS and taxpayer, theoretical analysis, and opinions bearing on the fairness and reasonableness of the parties’ positions (as distinguished from factual transactional data), which, if revealed to the IRS, could seriously prejudice the taxpayer in negotiations with it. If the auditor were required to disclose this material, which is not used in preparation of the taxpayer’s return and is not needed by the IRS to determine the correctness of the taxpayer’s return, the auditor would probably be inhibited from creating or maintaining such papers as part of a full and frank exchange useful in complying with the taxpayer’s obligations under federal securities laws.
The Service’s emerging practice of requesting tax accrual workpapers5 has aroused considerable controversy in the accounting and legal professions,6 as well as sharp disagreement among the courts that have considered the issue.7 In this appeal, appellants advance three arguments for denying enforcement to this facet of the summons. First, Amerada vehemently claims that since the summons was signed by Special Agent Kalemba, the relevance of the tax accrual workpapers must be measured [218]*218with respect to the criminal investigation that Kalemba was conducting. In other words, Amerada claims that the September 1977 agreement restricts the IRS’s right to obtain material to the task of the agent who summoned it.
The short answer to this argument is that Amerada has failed to prove the nature of its agreement with the IRS. We cannot place limits on the broad mandate that § 7602 gives to the IRS with no more than taxpayer’s contested assertion that the Service agreed to that limit. A longer answer is that § 7602 is only a power to compel production for the purpose of determining civil liability. Had an iron-clad agreement been forged to restrict the summons to documents pertaining to the criminal investigation, the summons would probably not have been enforceable, see United States v. LaSalle National Bank, 437 U.S. 298, 98 S.Ct. 2357, 57 L.Ed.2d 221 (1978). But as that case points out, the agent’s personal motivation would not be dispositive in any case. In deciding enforcement issues, the court’s mandate is to look to the motive of the “Service in an institutional sense .... ” Id. at 319, 98 S.Ct. at 2368. Besides, considerations of judicial economy compel us to decide this question on broader grounds: If we denied enforcement only because the documents were irrelevant to Agent Kalemba’s purpose, the IRS could simply issue a new summons, and the case would come before us again. Cf. Hickman v. Taylor, 329 U.S. at 505, 67 S.Ct. at 390 (declining to insist on “the empty formality of pursuing the right procedural device only to reestablish precisely the same basic problem now confronting us.”).
This brings us to a more fundamental inquiry, namely, whether the summoned documents are relevant to the civil audit. Appellants derive considerable support for the proposition that this material is irrelevant from the Tenth Circuit’s decision in Coopers & Lybrand, supra, and from United States v. Matras, 487 F.2d 1271 (1973), where the Eighth Circuit refused to enforce a § 7602 summons for taxpayer’s proposed budgets. The reasoning in both cases can be summed up as follows: Tax liability depends on the taxpayer’s actual transactions, not on what the taxpayer thinks and plans. Tax accrual workpapers reflect what the taxpayer — and his accountant— think about transactions that have already taken place; budgets reveal how the taxpayer hopes to conduct his affairs in the future. They are thus both irrelevant to the question of actual transactions, and therefore to the issue of liability for taxes. These courts conceded that both sorts of documents would be convenient to the IRS because they give the Service a “road map” for following the management of the taxpayer’s business. They went on to hold, however, that “[t]he term ‘relevant’ connotes and encompasses more than ‘convenience.’ ” Id. at 1275. Consequently, they refused to require that the documents requested be turned over to the government.
We cannot agree with such a restrictive definition of relevance. The summons power has consistently been interpreted as a broad mandate, designed to give the Service the “authority .. . necessary for the effective enforcement of the revenue laws .. . . ” United States v. Euge, 444 U.S. 707, 715-16 & n.9, 100 S.Ct. 874, 880 & n.9, 63 L.Ed.2d 141 (1980) (enforcing a summons for an examplar of taxpayer’s handwriting). See also United States v. Powell, supra (refusing to require a showing of probable cause in an enforcement proceeding); Donaldson v. United States, 400 U.S. 517, 91 S.Ct. 534, 27 L.Ed.2d 580 (1971) (refusing to deny enforcement of civil summons even though potential for criminal prosecution exists); United States v. Bisceglia, 420 U.S. 141, 95 S.Ct. 915, 43 L.Ed.2d 88 (1975) (upholding IRS power to issue a John Doe summons). Consequently, we have consistently used as our test of relevance whether the documents requested “might have thrown light upon” the correctness of a return, see, e.g., United States v. Shlom, 420 F.2d 263, 265 (2d Cir. 1969), cert. denied, 397 U.S. 1074, 90 S.Ct. 1521, 25 L.Ed.2d 809 (1970); Foster v. United States, 265 F.2d at 187; United States v. Noall, 587 F.2d at 125; United States v. Acker, 325 F.Supp. at 862. The documents at issue here certainly [219]*219pass this low threshold of relevance. Different tax positions lead to different amounts of liability. It is difficult to say that the assessment by the independent auditor of the correctness of positions taken by the taxpayer in his return would not throw “light upon” the correctness of the return.
Finally, the Coopers & Lybrand decision, upon which appellants have relied so heavily, emphasized the fact that the documents requested were not used in preparing the audited returns. This test has been rejected in this circuit, Noall, 587 F.2d at 126-27,8 and accordingly, much of what was persuasive to the Tenth Circuit leaves us unimpressed.
The foregoing reasoning does not mean that the IRS summons in dispute here must now be mechanically enforced, for other considerations also bear on our decision. The Supreme Court has recognized that “contrary legislative purposes” can undercut the “broad latitude” otherwise provided to the IRS, United States v. Euge, 444 U.S. at 716 & n.9, 100 S.Ct. at 880 & n.9. It has also recognized that common law privileges serve to limit the scope of the IRS’s power, Upjohn v. United States, 449 U.S. 383, 101 S.Ct. 677, 687, 66 L.Ed.2d 584 (1981); United States v. Euge, 444 U.S. at 714, 100 S.Ct. at 879. We think that the countervailing policies at issue in the case before us require us to fashion protection for the work that independent auditors, retained by publicly owned companies to comply with the federal securities laws, put into preparation of tax accrual workpapers.
Our starting point is Hickman v. Taylor, supra. In that case, the Supreme Court announced a policy of shielding “written statements, private memoranda and personal recollections prepared or formed by an adverse party’s counsel in the course of his legal duties ... ”, 329 U.S. at 510, 67 S.Ct. at 393, not because the documents were “privileged or irrelevant,” id. at 509, 67 S.Ct. at 392, but because their “[pjroper preparation ... is the historical and necessary way in which lawyers act within the framework of our system of jurisprudence to promote justice and to protect their clients’ interests.” Id. at 511, 67 S.Ct. at 393. Hickman was decided in the context of discovery under the Federal Rules of Civil Procedure, and the doctrine enunciated there has since been incorporated into Fed.R.Civ.P. 26(b)(3). But neither the Court nor Congress expressed an intent to limit the doctrine to the discovery context, Upjohn Co. v. United States, 449 U.S. at 398, 101 S.Ct. at 689. Rather, we read both Hickman and Rule 26 as requiring us to balance strong public policies against a party’s need for information whenever a conflict between the two arises.
The conflict in this case is between “the legitimate interest of society in enforcement of its laws and collection of the revenues,” Couch v. United States, 409 U.S. 322, 336, 93 S.Ct. 611, 619, 34 L.Ed.2d 548 (1973), and the “national public interest [in] in-surfing] the maintenance of fair and honest markets in [securities] transactions,” 15 U.S.C. § 78b. Congress has protected the former interest by vesting the Service with “the extensive powers granted ... by the Internal Revenue Code,” United States v. Powell, 379 U.S. at 56, 85 S.Ct. at 254, including the summons power of § 7602. The latter interest has been vindicated by extensively regulating the securities industry. Foremost among these regulations is the Securities Exchange Act of 1934, 15 U.S.C. § 78a et seq., which requires public companies such as Amerada Hess to file financial statements, 15 U.S.C. § 781, verified by independent accountants “in accord-, anee with generally accepted auditing standards,” 17 C.F.R. § 210.1-02(d). The verification procedure envisioned by the Act requires, in turn, that management feel free to cooperate with their auditors, and to disclose to them confidential information, such as the questionable positions taken on tax returns, and willingness to settle rather [220]*220than litígate when these positions are challenged by the IRS. The procedure recognizes that since tax laws must be general enough to treat complex and diverse factual situations, they are necessarily drafted with a great deal of flexibility. The Code’s requirements are therefore not always clear, and people of good faith can often take different positions as to the correct answer in a particular circumstance. While it is quite possible that a corporation motivated to save its shareholders’ money might take a favorable — i.e. “minimum tax” — position on a debatable item, it is equally true that the Service, which is also composed of mortals, might, in its zeal to maximize revenues, take an equally incorrect “maximum tax” position. We do not agree with the dissent’s statement that in the end, the corporation “pay[s] only what the tax laws required it to pay in the first place.” This would be true if every difference of opinion were litigated. But the fact is that very few controversies reach that stage. In reality, at the end of an audit, the IRS and the taxpayers frequently compromise out their differences, each giving in a little. Such negotiations strike an appropriate balance between collecting revenue and conserving judicial resources, but requiring the taxpayer alone to disclose settlement positions would improperly weight the balance.
The result of working with a tax code that is not cast in black and white is that the independent auditor is faced with a very sensitive inquiry. In order to assess whether the corporation has set aside enough reserves to pay contingent liabilities, the auditor must pinpoint possibly vulnerable areas in his client’s tax return and then predict how the corporation will handle the Service and vice versa, if these areas are questioned on audit. The inquiry necessarily lays bare not only the auditor’s thoughts but also the taxpayer’s basic thinking, including decisions not to litigate that are based on considerations wholly apart from the inherent legality of what the taxpayer has done, e.g., the cost of litigation and the possibility that confidential information may be disclosed to competitors. Divulging this information necessarily puts the taxpayer-corporation at a substantial disadvantage when it is audited. The prejudice involved in exposing to the Service appraisals of a taxpayer’s weaknesses and settlement positions on audit is of such proportions that a prudent organization might not be perfectly candid with independent auditors once it knew that the information revealed would be reachable under § 7602.9
This response ignores the fact that the depth and quality of any investigations . .. would suffer .... The response also proves too much, since it applies to all communications covered by the privilege: an individual trying to comply with the law or faced with a legal problem also has strong incentive to disclose information to his lawyer, yet the common law has recognized the value of the privilege in further facilitating communications.
The case at bar therefore involves much more than delineating the scope of the IRS’s authority under § 7602. When the IRS chooses to regularly summon tax accrual workpapers, it creates a clash between two important congressional policies. No matter how we decide this issue, one policy has to bend a bit. Giving the complete latitude to the summons power that the IRS seeks here compromises the procedure designed by Congress to protect the investing public from inaccurate financial information. Protecting the accuracy of the securities laws dilutes the power of the Service to summon a roadmap of the thoughts and theories of a taxpayer and its independent auditor. However, so long as a case does not involve allegations of fraud, the Service does not need to know taxpayer’s thoughts. The corporation’s own books and the audit workpapers furnish the IRS with all the raw data that it needs to calculate the taxpayer’s tax liability. A road-map would merely save the IRS some time in finding the best arguments for asserting a deficiency. The investing public, on the [221]*221other hand, relies most exclusively on the data generated as a result of the SEC laws. It therefore seems to us that this collision requires that some form of privilege be carved out to protect the independent auditing process. A work-product privilege, similar to the privilege fashioned in Hickman, seems to us appropriate. It protects those who benefit from enforcement of the securities laws from the “grave dangers of inaccuracy and untrustworthiness,” Hickman, 329 U.S. at 513, 67 S.Ct. at 394, that would result if tax accrual workpapers were routinely revealed to the Service. At the same time, it allows the IRS to procure these documents when the rare situation10 arises where it can make a sufficient showing of need to adequately justify invading the integrity of the auditing process, cf. Hickman, 329 U.S. at 512-13, 67 S.Ct. at 394, Fed.R.Civ.P. 26(b)(3). In reaching this conclusion, we are comforted by the knowledge that if Congress feels that this privilege unduly restricts the IRS’s authority, it can require that this material be made available, or that taxpayers flag their questionable positions directly on their returns, in line with the suggestions of former Commissioner Kurtz, 32 Tax Law. at 15-16; Cooper, The Avoidance Dynamic: A Tale of Tax Planning, Tax Ethics, and Tax Reform, 80 Colum.L.Rev. 1553, 1612 (1980). Procuring tax accrual workpapers under § 7602 seems to us to effectuate Commissioner Kurtz’s proposals through the back door.
In the case at hand, the IRS has not made a sufficient showing of need to override the privilege. Presumably, Amerada has made its own books available pursuant to the Third Circuit’s decision in United States v. Amerada Hess Corp., supra. In addition, this decision directs AY to release its audit workpapers, along with the other papers that the district court required it to turn over. Since the IRS is not attempting to prove that Amerada is guilty of fraud, the documents released should be sufficient for its purpose of determining the accuracy of Amerada’s returns.
For the foregoing reasons, the judgment of the district court is affirmed in part and reversed in part.