concurring in part and dissenting in part:
This case concerns fourteen documents relating to the role of the Office of Management and Budget (“OMB”) in clearing executive (and “independent”) agencies’ legislative and budget proposals to Congress. The appellant, Public Citizen, contends that the Freedom of Information Act (“FOIA”) requires OMB to disclose these documents in their entirety. OMB argues that the undisclosed portions of the documents are exempt from disclosure under FOIA Exemptions 2 and 5. The district court granted summary judgment to OMB on the basis of Exemption 2. The majority today finds Exemption 2 inapplicable, and remands for further consideration of Exemption 5. I believe that OMB established the first of two requirements for withholding under Exemption 2, and that the case should be remanded for clarification as to the second requirement. I concur, however, in the court’s disposition of OMB’s assertion of Exemption 5.
Exemption 2 covers documents that are “related solely to the internal personnel rules and practices of an agency.” 5 U.S.C. § 552(b)(2). We have long interpreted the exemption more broadly than its language immediately suggests. As currently understood, the exemption’s threshold requirement is that the documents must be “used for predominantly internal purposes.” Schiller v. NLRB, 964 F.2d 1205, 1207 (D.C.Cir.1992) (quoting Crooker v. Bureau of Alcohol, Tobacco & Firearms, 670 F.2d 1051, 1073 (D.C.Cir.1981)). Documents satisfying that criterion may be withheld if they “deal with trivial administrative matters” (the “low 2” exemption) or if “disclosure ... would risk circumvention of agency statutes and regulations” (the “high 2” exemption). Id. As the documents clearly deal with non-trivial matters, the “high 2” exemption is the relevant one. I will discuss the two requirements — predominant internality and circumvention of law — -in turn.
Predominant internality. Whether these documents are predominantly internal depends on their content and function. Public Citizen understandably inferred from the titles of the documents — for example, “Agencies Exempt from the Legislative Clearance Process” — -that they contain OMB’s policies regarding which agencies may bypass the clearance process by which OMB reviews agency submissions to Congress. In addition, the documents were supplied by OMB in response to a request for records “listing agencies that may” directly submit legislative and budget proposals to Congress without OMB clearance. Plaintiffs inference, nonetheless, is a bit of an oversimplification.
*878First, the Assistant Director for Legislative Reference of the OMB stated flatly in his affidavit that the documents “do not represent or set forth OMB’s ‘official position’ (so to speak) regarding which agencies may, or may not, submit legislative materials directly to Congress.” Jukes Am. Decl. at 14. See also Office of Mgmt. & Budget, Executive Office of the President, OMB Circular No. A-19, Legislative Coordination and Clearance (1979), posted at http://www.whitehouse.gov/omb/ circulars/a019/a019.html (“Coverage. All executive branch agencies (as defined in section 5b) are subject to the provisions of this Circular, except those agencies that are specifically required by law to transmit their legislative proposals, reports, or testimony to the Congress without prior clearance.”).
Second, in camera examination shows that insofar as the documents guide the conduct of OMB personnel at all, they do so in relation not to the clearance process but to another aspect of OMB’s relation to agencies. The documents contain no instructions to OMB personnel to allow some agencies but not others to bypass the clearance process. Documents 1 and 3-14, which are memos from the OMB’s Assistant Director for Legislative Clearance to OMB staff, do contain other instructions making clear the documents’ real purpose. Specifically, the memos instruct the OMB staff not to automatically refer material to “bypass” agencies, so as to avoid letting the agencies forward such material to Congress. See, e.g., Jim Jukes, Memorandum for OMB Policy Officers and DADs [acronym unexplained] (Feb. 20, 2001) (“[The Legislative Reference Division’s] longstanding practice is to determine, on a case-by-case basis, whether or not to refer material for review to a ‘bypass’ agency.... In general, an OMB referral is not made to a ‘bypass’ agency unless the agency agrees to refrain from forwarding to Congress OMB-referred material or its response to an OMB referral.”) In other words, to the extent that the documents had a function beyond explaining what bypasses are and which agencies have asserted a right to bypass, it was to instruct employees on how those agencies should be treated in the referral process.
The majority opinion hints that the memos contain a policy concerning which agencies may bypass the clearance process. It relies on Document l’s statement that “[f]orty four Federal agencies currently have some form of ... ‘bypass.’ ” Maj. Op. at 871 (quoting Jim Jukes, Memorandum for OMB Policy Officers and DADs (Feb. 20, 2001)). In context, though, it is clear that this is a statement of fact concerning which agencies have asserted bypass authority, not an endorsement of their claims or even a statement of OMB assessment of or response to those claims. See Bypass Agencies 5 (Feb.2001) (explaining that informal bypasses are asserted by several agencies without explicit authority, and that although Circular A-19 does cover these agencies they “generally do not comply with its provisions.”).
The panel’s next argument on this point similarly confuses a factual statement with a policy. It relies on the Assistant Director’s statement that the documents “seek to summarize the currently-held internal-OMB perspectives and views regarding which Federal agencies have a basis — in statute or in prior agency and OMB practice — for not submitting [materials for OMB clearance].” Jukes Am. Decl. at 13. All this means, however, is that the documents list all the agencies which may view themselves as exempt from the clearance process, whether this is because of a statute, the agency’s own prior practice, or some combination of the agency’s practice and OMB’s response to it. Consider, for example, Document l’s treatment of the Federal Trade Commission “FTC”: “The *879FTC has no statutory legislative bypass but acts as if it does.... OMB has made no effort in recent memory to subject the FTC to the requirements of Circular A-19.” Bypass Agencies 19 (Feb.2001). The document explains why the FTC might have a “basis” in its own practice, as well as OMB’s recent actions, to think of itself as a bypass agency. It does not, however, endorse the FTC’s claim to bypass authority or adopt a policy of not attempting to get the FTC to comply with Circular A-19.
The panel further argues that if the documents really just recorded past OMB practice' — rather than setting out a policy of OMB acquiescence in these agencies’ behavior — OMB would not worry about sharing information with them. As my colleagues put it, “If the listed agencies had no bypass authority, OMB would have no reason to avoid sharing other agency proposals with them.” Maj. Op. at 873. Again the analysis confuses a pattern of successful agency bypass with OMB policy. Even if OMB made every possible effort to subject the FTC or some other informal bypass agency to Circular A-19, it would still sensibly worry about sharing information with that agency — unless and until its efforts proved successful. The panel’s theory simply assumes that OMB has enforcement power; but we have no basis for inferring such power’s existence.
In support of its view the panel invokes a couple of passages from oral argument. I hesitate to draw serious conclusions from a muddled colloquy in which the judges more than once declared that they could not understand counsel’s answers. Oral Arg. at 35:30-35:35; id. at 36:05-36:15. As the panel’s conviction that the materials disclose “secret law” turns on an understanding of OMB enforcement power, we should at the very least remand for a determination that such power exists rather than rely on inferences from highly ambiguous statements at oral argument by a counsel to whom our questions appeared extremely unclear.
In any event, counsel’s answers fall far short of a clear claim to effective power to insist that all agencies submit proposals for OMB review before submission to Congress. The statement of OMB counsel at oral argument that OMB “always [has] the power as granted by the Executive Order, by the Circular, by the memo of February 15, 2001 to say ‘we want to look at this,’ and [it] can always do that,” see Maj. Op. at 873 (quoting Oral Arg. at 37:24-37:33), does nothing to establish enforcement power. Quite literally, counsel observed that OMB had the power to “say” to agencies that they should turn over the specified type of document. It brings to mind Hotspur’s famous rejoinder to Glendower:
Glendower: I can call spirits from the vasty deep.
Hotspur: Why, so can I, or so can any man; But will they come when you do call for them?
At most it suggests that OMB has a legal right to require certain agencies to participate in the clearance process, quite distinct from an effective enforcement power.
The majority also points to an exchange in which a judge asked counsel whether OMB “even with respect to those agencies not statutorily exempt will for particular pieces of legislation exercise its review function,” and counsel responded, “if it chooses that it should at a particular time, it will.” Maj. Op. at 873 (quoting Oral Arg. at 35:35-36:00). Counsel was primarily asserting that OMB can’t make “any decisions about agencies and their ability to go around OMB,” Oral Arg. at 34:30-34:38. But it does, he argued, have some discretion with regard to specific legislative proposals: “It can make a decision, for example, when a [proposed?] statute comes to it, that this particular statute does not need to go through a review process.” Id. *880at 34:59-35:05. Counsel’s answer to the question, then, merely reaffirms his point that OMB has discretion over “particular pieces of legislation.” It does not establish that agencies would comply, and it certainly does not establish that pieces of legislation would “come to [OMB]” before going to Congress. Indeed, elsewhere counsel noted that OMB frequently does not even know about agencies’ legislative proposals until they are introduced in Congress. Oral Arg. at 36:21-36:33. In short, the OMB’s own description of the documents is fair: the documents “serve as the OMB’s ‘intelligence’ on other agencies’ views regarding the nature of their obligations in the legislative clearance process” and they are used to help OMB personnel “determine the nature of then-interactions” with those agencies. Appellee’s Br. at 13.
We must evaluate these documents under the predominant internality test established in Crooker v. Bureau of Alcohol, Tobacco & Firearms, 670 F.2d 1051, 1073 (D.C.Cir.1981), read in the light of the Crooker en banc court’s preservation of the holding of an earlier case, Jordan v. DOJ, 591 F.2d 753, 763 (D.C.Cir.1978). In Jordan we had held that the Justice Department’s prosecutorial guidelines did not fall within the scope of Exemption 2. Crooker rejected much of Jordan’s legal analysis, 670 F.2d at 1073, and held that an agent’s training manual of the Bureau of Alcohol, Tobacco & Firearms (BATF), governing law enforcement investigation techniques, was covered by Exemption 2 because it was predominantly internal and its disclosure significantly risked circumvention of the law, id. at 1053. Crooker said, however, that its new approach would not have changed the result in Jordan because nothing in the Jordan opinion suggested that disclosure would risk circumvention of the law and, in any event, the documents at issue were not predominantly internal. Id. at 1075.
Crooker conceded that the investigatory technique policies described in the withheld portions of the BATF manual had an effect on the public at large, as would almost any agency policy. Id. at 1073. The court found “critical,” however, that “the manual is used for predominantly internal purposes; it is designed to establish rules and practices for agency personnel ...; it involves no ‘secret law' of the agency....” Id. The prosecutorial guidelines in Jordan, on the other hand, were “a source of ‘secret law,’ as important to the regulation of public behavior as if they had been codified.” Id. at 1075.
As applied to documents by which an agency guides its personnel in conduct affecting others, the distinction our cases draw seems a bit metaphysical, i.e., difficult to operationalize. As to any such document, it is possible to assert, with equal plausibility, that its “primary” purpose is to guide the agency’s employees or, by guiding the employees’ conduct, to affect the outside world. The puzzle is highlighted by Schiller v. NLRB, 964 F.2d at 1208 (1992), where we confronted documents concerning the NLRB’s litigation strategies with reference to the Equal Access to Justice Act. Of course the “prosecutorial strategies” at issue in Jordan might be viewed as simply a subset of “litigation strategies.” Yet, while the NLRB strategies clearly affected outsiders — -and were presumably intended to influence the other parties’ behavior — we said that they merely “established] rules and practices for agency personnel, and Mr. Schiller has given us no reason to think that the documents contain any ‘secret law.’ ” Id. at 1207. Accordingly we found them predominantly internal. Similarly, in Nat’l Treasury Employees Union v. U.S. Customs Serv. (“NTEU”), 802 F.2d 525, 531 (D.C.Cir.1986), we classified statements of criteria for agency employment as predominantly internal (a conclusion in*881directly bolstered, of course, by the documents’ clearly relating to “personnel”).
Must we then throw up our hands and arbitrarily choose one of two contradictory assertions? I think not. Two features may usefully distinguish Jordan from the three later cases. As we described Jordan in Crooker, the strategies we characterized as secret law were “as important to the regulation of public behavior as if they had been codified.” 670 F.2d at 1075. To the extent that the prosecutorial guidelines were the equivalent of flat-out no-proseeution rules, they switched the conduct in question from unlawful to de facto lawful, as would, for example, a clear determination not to prosecute marijuana offenses. Thus they (1) impacted primary conduct and, as we understood them, (2) they did so unequivocally (“as if they had been codified”). By contrast, the investigative techniques in Crooker, the litigation strategies in Schiller, and the employment criteria in NTEU appear to have been aimed at peripheral activity: in Crooker at parties’ concealment strategies; in Schiller at their behavior in agency adjudications; in NTEU at their role as job applicants. (Of course if persons dedicated to a career with the U.S. Customs were fully informed about the documents withheld in NTEU, they might mold their career paths to meet its interests; but such an effect seems remote enough to justify our having viewed the documents as predominantly internal.)
At bottom, the policy expressed in the documents here is no more than a set of instructions to agency staff on how to bargain with other agencies on an issue much less connected to their primary conduct than submission to OMB approval of then-legislative or budget proposals, to wit, the dissemination of information. See Jukes Memorandum (Feb. 20, 2001). This external effect seems about as remote from the public’s primary conduct as one can imagine. Nor do these documents regulate the primary conduct of other agencies (assuming for the moment that doing so would bring them within the scope of Jordan)', they deal only with the agencies’ horsetrading with OMB on issues of bureaucratic information exchange — the referral process. The panel characterizes that process as “a key part of the legislative clearance process.” Maj. Op. at 872. But OMB’s referral leverage strategies seem to stand in relation to the basic clearance process in much the way the criminal investigative procedures, administrative litigation strategies, and agency employment criteria at issue in Crooker, Schiller and NTEU relate to criminal law enforcement, administrative policy, and agency management of personnel, i.e., so peripherally to the affected subjects’ primary conduct that they are properly seen as “predominantly internal.”
Second, the documents are not at all comparable to any kind of codification; within their mandate, one can easily imagine temporary, partial accommodations. While the bargaining strategy may well force other agencies to make a choice, it is a far cry from the decriminalization of a whole class of conduct.
The majority summarizes its view with the declaration that where “agency documents have significant external effects on other government agencies, we cannot deem them ‘predominantly internal.’ ” Id. at 874. I note that this is the first case ever in which a document’s “external effects” operate in the first instance on other federal agencies. I do not regard that fact as dispositive: if the initial impact fell on another government agency in such a way as to have clearly defined effects on the public’s primary conduct, it would not make sense to view the documents as “predominantly internal.” And quite possibly an agency policy seriously impacting other *882agencies’ primary conduct would fail the internality test. Neither effect is present here. The case fits comfortably within Crooker, Schiller and NTEU.
Circumvention of the law. The second prong of the “high 2” exemption is met if “disclosure would significantly risk circumvention of agency regulations or statutes.” NTEU, 802 F.2d at 528. NTEU illustrates how we apply the criterion. There we observed that because the agency’s evaluation procedures were supposed to measure “actual experience and proven ability,” in theory “advance knowledge of their content should not affect the rating of the candidates.” Id. at 529. This conclusion would hold, however, only “if all applicants can be depended upon to be meticulously correct in describing then-past experience and their quantified or quantifiable abilities.” Id. In fact, affidavits from agency individuals suggested that applicants could embellish many aspects of them applications “in a manner that is not strictly fraudulent, or that cannot be proven to be fraudulent.” Id. In light of these affidavits, we found “that release of the plans creates a significant risk that the Service’s applicant evaluation program will be seriously compromised.” Id.
Here OMB expresses the concern that if other agencies “knew OMB’s beliefs concerning their views or the views of sister agencies, they could use this information to impede and frustrate legislative clearance requirements,” Appellee’s Br. at 26, thus circumventing the legislative clearance process set out in Circular A-19. Thus the claim does not appear to relate to the documents’ normative instructions on referral of documents to bypass agencies. As in NTEU, in theory the information in the documents should not affect whether or not an agency is subject to the clearance process. In practice, however, this may hold only if agencies approach the process with meticulous integrity. It is not fanciful to imagine that they might change their behavior in response to the information. Indeed, the majority believes that the explanation of OMB’s policies has “significant external effects on the behavior of other agencies.” Maj. Op. at 873. To the extent that agencies are willing to game the system, the information in these documents could help them do so.
OMB’s submissions on this issue, however, are on the vague side. The Assistant Director for Legislative Reference of the OMB said in his affidavit that disclosure of these materials “would reveal aspects of OMB’s evaluative process concerning submission of agencies’ documents to Congress without OMB’s clearance and the manner in which relevant opinions and recommendations were formed.” Jukes Deck at 11. But this statement, rather than being addressed directly to circumvention, seems simply to assert the raw truism that forced disclosure will reveal something about OMB’s thinking process. The issue, though, is how agencies might use those insights to undermine OMB’s efforts to assure compliance. The record on that problem being too opaque for a well-founded decision, I would remand to the district court for further proceedings. See Sussman v. United States Marshals Serv., 494 F.3d 1106, 1113 (D.C.Cir.2007).
My colleagues argue that the documents at issue in this case “lie at the core of what FOIA seeks to expose to public scrutiny.” Maj. Op. at 874. Disclosure is, of course, FOIA’s primary policy. See Crooker, 670 F.2d at 1074. But as Crooker reminds us, “it will not do for us to act on the primary purpose of the statute to the exclusion of all other express congressional concerns,” such as “preserving the effective operation of governmental agencies.” Id. Here, the *883effectiveness potentially at stake is the President’s ability to corral the government’s far flung agencies, many if not all of them beholden to interest groups whose agenda may not track the President’s, into support of a common, coherent program. A fair application of the test developed by Crooker demonstrates that the documents are predominantly internal. If further proceedings establish that their disclosure risks circumvention of the law, as seems quite plausible, OMB should be able to protect them. I would therefore remand for a better grounded decision on the circumvention issue.