Opinion for the Court filed by Circuit Judge HARRY T. EDWARDS.
Concurring opinion filed by Circuit Judge RANDOLPH, in which Circuit Judge STEPHEN F. WILLIAMS joins.
HARRY T. EDWARDS, Circuit Judge:This appeal involves a claim under the Freedom of Information Act (“FOIA”) regarding certain reports dealing with nuclear power plant safety. The disputed reports are prepared by the Institute for Nuclear Power Operations (“INPO”), a utility consortium; since 1982, INPO has shared these reports with the Nuclear Regulatory Commission (“NRC” or “Commission”) pursuant to an agreement providing for the free exchange of nuclear information. The appellant, Critical Mass Energy Project (“Critical Mass”), seeks to secure copies of the reports from the Commission under FOIA.
In 1984, the NRC refused to release the INPO reports to Critical Mass, citing FOIA’s exemption for confidential commercial information; Critical Mass then brought suit challenging the agency’s action. The District Court issued a summary judgment in favor of the NRC; however, that judgment was vacated on appeal and the case was remanded to the District Court for further proceedings. See Critical Mass Energy Project v. NRC, 830 F.2d 278 (D.C.Cir.1987) (“Critical Mass I”). On remand, the District Court once again denied the appellant’s FOIA request, holding that the INPO reports were protected from *941disclosure under FOIA Exemption 4. Summary judgment was awarded to the Commission, and Critical Mass has appealed from that judgment. 731 F.Supp. 554.
On the record before us, we can find no basis to conclude that disclosure of the relevant reports would be likely to result in any significant impairment of either the effectiveness or efficiency of the NRC by virtue of anticipated antagonism in the relationship between INPO and the Commission. We therefore reverse the judgment of the District Court on this point. In addition, because we find that genuine issues of material fact remain concerning the likely effects of disclosure on the Commission’s information-gathering ability, we remand the case for further proceedings consistent with this opinion.
I. Background
A. Facts
In the wake of the Three Mile Island accident, the nuclear electric utility industry formed INPO, a voluntary membership organization designed to promote safety and reliability in nuclear power plant operations through peer review. INPO, which is comprised of and funded by utility companies that operate or construct nuclear plants in the United States, employs approximately 400 engineers and other staff. One of INPO’s principal programs is the Significant Event Evaluation and Information Network (“SEE-IN”), a system for collecting, analyzing and sharing information concerning construction and operations experiences within the nuclear power industry.
As part of SEE-IN, INPO produces three types of reports on nuclear power plant operations: Significant Event Reports (“SERs”); Significant Operating Event Reports (“SOERs”); and Operation & Maintenance Reminders (“0 & MRs”). As their titles suggest, SERs and SOERs involve operating events at nuclear facilities deemed “significant” by INPO engineers. SERs include descriptions of the events, discussions of their possible causes and effects and INPO’s analyses of the problems identified; SOERs are more detailed follow-up analyses, containing more wide-ranging, generic recommendations to utilities regarding plant construction, design and operation. 0 & MRs describe less significant operating and maintenance problems encountered in specific plants and discuss the means used to correct them. In preparing the SERs, SOERs and 0 & MRs, INPO analysts solicit comments and evaluations from working-level employees who have some familiarity with the events and problems raised in the reports.
INPO currently distributes copies of its SEE-IN reports to all INPO members and “participants,”1 to the Nuclear Safety Analysis Center (a membership organization of utilities engaged in the commercial production of electricity) and to Nuclear Electric Insurance Limited (an insurer of INPO utilities). In addition, INPO provides copies of particular reports to vendors whose products are mentioned in those reports, and to outside consultants and contractors where necessary to take corrective action suggested by the reports. Pursuant to a 1982 Memorandum of Agreement providing for the free exchange of nuclear safety information between INPO and the Government, INPO also transmits copies of all SEE-IN reports to the Commission. INPO sends its reports to each of the foregoing recipients with an understanding that the reports will not be disclosed to additional third parties without INPO’s consent.2
*942B. The Decision in Critical Mass I
In 1984, Critical Mass filed a FOIA request seeking access to INPO reports in the Commission’s possession. The NRC denied the request on the grounds that the reports were protected from disclosure under FOIA Exemption 4, which shields “commercial or financial information obtained from a person and privileged or confidential.” 5 U.S.C. § 552(b)(4) (1988). Critical Mass thereupon brought suit in the District Court challenging the Commission’s decision. Finding that the records properly came within the ambit of Exemption 4, the District Court granted the Commission’s motion for summary judgment.
Following argument before another panel of this court, the District Court’s judgment was vacated on appeal. The decision in Critical Mass I first notes that, under FOIA, the agency carries the burden of justifying the decision not to disclose the disputed reports. Critical Mass I, 830 F.2d at 281 & n. 14. The decision then observes that our precedent has established “two prime requirements” for confidentiality. Id. at 281.3 First, the agency must show “that the information it seeks to shield would customarily not be released to the public by the person from whom it was obtained.” Id. (internal quotation omitted). The decision finds this requirement to be satisfied by INPO’s limited distribution of its reports to a well-defined (albeit large) class of recipients and its articulated policy forbidding unauthorized disclosure of the reports to third parties.
Second, the Critical Mass I decision declares that “the agency must demonstrate that disclosure [would] ... harm a specific interest that Congress sought to protect by enacting the exemption.” Id. (internal quotation omitted). On this point, the court notes that, under the holding of National Parks & Conservation Ass’n v. Morton, 498 F.2d 765 (D.C.Cir.1974), we have recognized two such protected interests: (1) the Government's need to have access to eom-mercial and financial data and (2) the need to safeguard persons submitting such data to the Government from the competitive harms that might result from general publication. Focusing on these interests, National Parks established the following test for the application of Exemption 4:
[Commercial or financial matter is “confidential” ... if disclosure of the information is likely ... either ... (1) to impair the Government’s ability to obtain necessary information in the future; or (2) to cause substantial harm to the competitive position of the person from whom the information is obtained.
National Parks, 498 F.2d at 770 (footnote omitted). In the decision under review in Critical Mass I, the District Court had relied on the first prong of this test, finding that disclosure of the INPO reports would be likely to impair the Commission’s future ability to acquire the information they contained.4
This court rejected the judgment of the District Court in Critical Mass I for lack of evidentiary support in the record. INPO had made no contention that it would discontinue either preparing the reports or disseminating them to its members if disclosure were required under FOIA; INPO had merely represented that it would no longer voluntarily submit the reports to the Commission. Accordingly, the decision in Critical Mass I holds that, in order to show a significant impairment to its information-gathering ability, the Commission had to demonstrate “either (1) that cessation of INPO’s voluntary submission of these reports would in fact deprive the agency of the information contained therein, or (2) that alternative means for obtaining the ... reports would create a significant risk that ... [their] value ... would decrease.” 830 F.2d at 283. There was no dispute on the first point because the Commission had conceded that it possessed “ample statutory authority” to obtain the INPO reports *943by compulsory means. Id. On the second point, however, the parties disagreed over the possible effect of public disclosure on the quality of the information and analyses in the reports. Because the District Court had made no findings on this second point, the case was remanded to allow the Commission to document its “necessarily uncertain prediction,” id. at 286, that disclosure of the INPO reports would result in “impairment ... significant enough to justify withholding the information.” Id. (quoting Washington Post Co. v. HHS, 690 F.2d 252, 269 (D.C.Cir.1982)).
In remanding, the decision in Critical Mass I amplifies somewhat on the National Parks test, indicating that, with respect to the question of confidentiality under Exemption 4, nondisclosure also may be appropriate if the agency can show that public disclosure will significantly harm the agency’s interests in its efficient operations or in the effective execution of its statutory responsibilities. Critical Mass I, 830 F.2d at 286-87. Thus, the District Court was instructed to afford the Commission the opportunity to demonstrate that disclosure of the INPO reports would damage “some identifiable agency interest relating to program effectiveness or efficiency.” Id. at 287.
Following the decision in Critical Mass I, INPO intervened as a co-defendant. After additional proceedings before the trial court, summary judgment was once again awarded in favor of the NRC. Critical Mass Energy Project v. NRC, 731 F.Supp. 554 (D.D.C.1990). In issuing this judgment, however, the trial court declined to rely on any likely impairment to the Government’s information-gathering capability, deeming the effects of disclosure on the quality of the information contained in the INPO reports “an issue neither more declarations nor a parade of witnesses could definitively resolve in advance of the event.” Id. at 557. Instead, the District Court found that disclosure would be likely to diminish the Commission’s regulatory efficiency and effectiveness.
The District Court assumed that if disclosure were required under FOIA, INPO would resist future Commission attempts to obtain access to its reports, and predicted that Commission efforts to acquire the reports by compulsory means would engender “both acrimony and some form of litigation with attendant expense and delay.” Id. Thus, the trial court concluded that disclosure would change the relationship between the Commission and INPO from one of “collaborators in a quest for optimum industry safety” into one involving, “[i]f not outright antagonists, ... at best ... wary allies, working independently of one another’s efforts, and mistrustful of one another’s initiatives or overtures.” Id. On the strength of this conclusion, the District Court held that the Commission’s “efficiency and effectiveness would be impaired were it not permitted to honor its commitment to INPO to keep the INPO reports in confidence.” Id. Accordingly, the trial court again found the reports to be protected from disclosure under FOIA Exemption 4, and entered summary judgment for the Commission and INPO. The present appeal ensued.
II. Analysis
A. The Impact of Disclosure Upon the Commission’s Efficient Operations and Effective Execution of Its Statutory Responsibilities
Critical Mass contends that reversal is required because the District Court had no basis upon which to hold that the Commission’s “efficiency and effectiveness would be impaired were it not permitted to honor its commitment to INPO to keep the INPO reports in confidence.” 731 F.Supp. at 557. We agree.
For one thing, the trial court recognized that, other than INPO’s assertion that it would no longer voluntarily submit its reports to the Commission, the record is “largely unrevealing as to how the defendants would react to a decision rendering the INPO reports disclosable under FOIA.” Id. at 556. And, as noted above, the Commission concedes that it has “ample statutory authority ... to compel the production of the INPO reports.” Critical Mass I, 830 F.2d at 283-84. Furthermore, on the *944record before us, there is no evidence whatsoever to support the District Court’s holding that disclosure of the INPO reports will significantly impair the effectiveness or efficiency of the agency by virtue of anticipated antagonism in the relationship between NRC and INPO.
Whatever the ultimate disposition of this case, the Commission and INPO will still share a common goal of promoting nuclear power plant safety. Moreover, INPO, as the representative of the nuclear power industry, will continue to possess a powerful incentive to cooperate with the Commission in the pursuit of this goal. Through such concerted efforts, INPO has a firsthand opportunity to demonstrate the efficacy of industry self-regulation to the Government agency responsible for regulating nuclear power.5
Even were we to accept the District Court’s finding that the relationship between INPO and the Commission would somehow suffer if the reports lost their confidential status, however, we could not uphold the trial court’s application of Exemption 4 on this basis. In premising its summary judgment award solely on the possibility that disclosure of confidential records might create friction between a Government agency and the preparer of those records, the District Court misconstrued the type of “identifiable agency interest relating to program efficiency” upon which an Exemption 4 claim legitimately can be based.
The decision in Critical Mass I recognizes that the NRC might have “substantial, indeed compelling, reasons why it would choose not to exercise the coercive powers at its disposal” to require INPO to share its reports if they are not voluntarily furnished. 830 F.2d at 284 n. 26. But no such “substantial” or “compelling” reasons have been shown in this case. In Critical Mass I, the NRC had argued that disclosa-bility would require it to resort to its subpoena power in order to obtain INPO reports in the future. Stressing the “unavoidable delays” and “practical problems that dot the subpoena route,” the Commission had claimed that these compulsory procedures were “not a reasonable alternative means for getting ... [the] information.” Id. at 286 n. 36. The NRC now concedes that it can obtain the INPO reports by requiring nuclear plant licensees to file INPO reports with the Commission pursuant to its licensing regulations. And, the Commission has pointed to nothing in the record to indicate that use of its regulatory authority, a far less burdensome method for obtaining information than the subpoena process, is not “a reasonable alternative means” for it to acquire INPO reports (apart from possible effects on the quality of the reports). There is no evidence suggesting either that licensees would disobey regulations requiring them to submit copies of INPO reports to the Commission promptly after receiving them from INPO, or that the processing of these submissions would impose a substantial burden upon the Commission’s resources.
Nor did the District Court base its decision upon any such significant impairment to the agency’s efficiency interests. The only alleged “damage to some identifiable agency interest” cited by the District Court was the possibility that disclosure might spawn litigation between the Commission and INPO, along with “attendant cost and delay.” Were these possible effects sufficient to justify application of Exemption 4, however, a person submitting valuable information to the Government voluntarily could dictate whether that information was confidential for purposes of FOIA; and, this third-party control could be achieved even with respect to information that is otherwise attainable by the agency through regulation. So long as the third-party threatened to resist production of the information if it were found subject to public dissemination, the agency could base an exemption claim on alleged adverse *945effects to its operations that exercise of its coercive power would likely entail. By thus allowing the confidential status of information to depend entirely upon the will of the party submitting that information to the Government, we would ignore the law of this circuit holding “that the test for confidentiality is an objective one.” National Parks, 498 F.2d at 766. We always assume that a request for confidentiality is based in part on the subjective preferences of the parties, but those preferences must be supported by objective justifications.6
Although our opinion in Critical Mass I recognizes agency interests in efficiency and effectiveness, the decision did not purport to change the law of the circuit regarding the objective test of confidentiality. Indeed, Critical Mass I left but a small window of opportunity for the NRC “to show, if it so [chose], that exercising its full authority [to compel disclosure of the INPO reports through] regulation would damage some identifiable agency interest relating to program effectiveness or efficiency.” 830 F.2d at 287. The court surely did not mean to suggest that the agency could satisfy this test with a mere suggestion that those against whom the regulation will be enforced will be unhappy.
On the record before us, we can find no evidence to support a holding that the NRC’s resort to the exercise of regulatory authority to secure the INPO reports will significantly impair “some identifiable agency interest relating to program effectiveness or efficiency.” Critical Mass I, 830 F.2d at 287. We therefore reverse the judgment of the District Court on this point.
B. The Impact of Disclosure Upon the Commission’s Information-Gathering Ability
Information will also be deemed confidential for purposes of Exemption 4 if “disclosure of the information is likely ... to impair the Government’s ability to obtain necessary information in the future.” National Parks, 498 F.2d at 770. As noted earlier, the District Court did not premise its grant of summary judgment on this ground. Both parties, however, have addressed this possible basis for nondisclosure in their briefs and at oral argument, and so we may properly consider whether the District Court’s summary judgment award could rest upon the likely impairment to the Commission’s information-gathering ability. See Proctor v. State Farm Mut. Auto. Ins. Co., 676 F.2d 308, 326 (D.C.Cir.) (appellate court may affirm grant of summary judgment on ground not relied upon by lower court, provided opposing party has had fair opportunity to dispute facts material to new theory), cert. denied, 459 U.S. 839, 103 S.Ct. 86, 74 L.Ed.2d 81 (1982).
We initially note that the information contained in the INPO reports is sufficiently important to the Commission’s exercise of its statutory responsibilities to be “necessary information” under National Parks. See Washington Post Co., 690 F.2d at 268. It is true that INPO reports contain essentially the same objective data concerning safety-related incidents that the Commission already obtains directly from reports prepared by licensees.7 However, INPO reports offer candid analyses from a diverse group of industry professionals; and many of these analyses come from persons other than those who are assigned to prepare the “official” incident reports for the NRC. There can be no doubt that access to the insights of the knowledgeable INPO analysts is an invaluable benefit to the Commission in its attempts to fulfill its broad statutory mandate to promote safety in the nuclear power field.
*946The remainder of our inquiry is narrowly circumscribed by the concessions of the parties and the prior proceedings in this case. First, it is undisputed that, whether or not INPO reports are disclosed to the public, INPO will continue to create those reports, will prepare them with the same degree of diligence as it has in the past and will distribute them to its members. Second, the Commission has expressed confidence that it will be able to gain access to those reports from licensees, and INPO has not suggested that it can prevent its members from submitting the reports to the agency pursuant to Commission regulations. Third, the INPO reports traditionally have been widely distributed to all INPO members, to all domestic and some international nuclear utilities, to various contractors and suppliers, to persons engaged in research associated with nuclear safety, and to NRC agents and officials. The limited question before us, then, is whether the qualitative value of the reports would decrease if they became available to the public-at-large.
Throughout these proceedings, the Commission and INPO have argued that the confidentiality of INPO reports is critical to their accuracy and value. They note that INPO engineers base their reports largely on subjective, often self-critical, input solicited from working-level employees at the sites of the incidents under examination. The Commission and INPO argue that these employees might not speak as freely about their involvement in such incidents or offer their theories of possible causes of safety incidents if they knew such information would be more generally available to the public-at-large. As a result, these parties claim, disclosure would seriously diminish the value of the INPO reports.
We fully agree that the Commission would be justified in withholding the INPO reports if limiting their circulation were truly necessary to maintain their quality. However, we have searched the present record in vain for any evidence that would justify such a finding.
From the evidence before us, it is unclear that the employees whose input is solicited for INPO reports either know or care about the confidential nature of their statements. As noted above, INPO reports are distributed not only to the Commission, but also to all domestic nuclear utilities, some international utilities, plant suppliers and an occasional consultant or contractor. Thus, while these records are not generally available to the public-at-large, they do receive wide distribution throughout the nuclear energy community. Moreover, employees seem perfectly willing to talk with INPO investigators even though it is understood that their comments will be reviewed by their employers and by federal regulators. Critical Mass claims that, based on the existing broad dissemination of INPO reports, whatever candor currently finds its way into the INPO reports will not be significantly reduced by the prospect that the reports might also be made available to the general public.
Responding to Critical Mass’ claims, the Commission and INPO argue that nuclear plant employees are comfortable speaking candidly within the nuclear power community, where their speculations and ruminations about plant events are unlikely to be misunderstood or taken out of context. The Commission and INPO claim, however, that employees would be reluctant to talk freely if they knew that their subjective comments might publicly embarrass colleagues or cast an unfavorable light on the industry generally. We cannot determine whether the views of Critical Mass or the Commission and INPO are correct; on the present record, the effects of disclosure upon employee candor simply cannot be gauged with any degree of confidence. The record contains absolutely no testimony from working-level employees regarding the importance of confidential treatment of INPO reports to their willingness to speak to INPO analysts. In fact, the record fails to demonstrate whether the employees interviewed by INPO analysts are even informed or aware of INPO’s current limited distribution policy.8
*947The only evidence presented before the District Court on this critical point were affidavits from officials of the Commission and INPO theorizing that disclosure would be inimical to employee candor. Such self-interested speculations by officials of the entities resisting disclosure, however, are hardly the type of evidence upon which a finding of likely impairment can be based.
We therefore conclude that a genuine dispute of material fact remains concerning the effects of disclosure on the quality of INPO reports and remand to the District Court for additional proceedings directed to this question. As in Critical Mass I, we recognize that any finding by the District Court on this issue may admit of some uncertainty. This is not fatal to a disposition of this case, however, because the adverse effects of disclosure for purposes of Exemption 4 need not be proven with mathematical certainty. See Public Citizen Health Research Group v. FDA, 704 F.2d 1280, 1291 (D.C.Cir.1983). The District Court should review the evidence submitted by the parties and, as a trial court must do in countless other contexts, resolve material disputes of fact and offer a legal judgment based upon its factual findings, bearing in mind that the NRC carries the “burden of demonstrating that these documents are within FOIA exemption 4,” Critical Mass I, 830 F.2d at 281, and that the issue is one of “necessarily uncertain prediction,” id. at 286.
Evidence relevant to the court’s inquiry would include details concerning the full extent to which INPO reports are currently circulated, to determine whether their limited confidentiality is truly maintained; testimony or affidavits from employees interviewed by INPO analysts concerning the importance of limited circulation to their willingness to speak freely; and any evidence indicating the degree to which employees are currently aware of and reasonably rely on the limited circulation of INPO reports as a prerequisite to speaking to interviewers. See Formaldehyde Inst. v. HHS, 889 F.2d 1118, 1124-25 (D.C.Cir.1989) (reviewing evidence concerning effects of public disclosure upon peer reviews).
III. Conclusion
For the reasons stated, we reverse the judgment of the District Court and remand for further proceedings.
So ordered.
. INPO’s participants are domestic companies that serve as suppliers to nuclear plants and international utility organizations that operate or construct nuclear facilities in foreign countries.
. SEE-IN reports contain the following statement:
Reproduction of not more than ten copies by each recipient for its internal use or use by its contractors in the normal course of business is permitted. This report should not be otherwise transferred or delivered to any third person, and its contents should not be made public, without the prior agreement of INPO.
Exhibits A, B & C to Supplemental Affidavit of Terence J. Sullivan, reprinted in Joint Appendix (“J.A.”) 553, 574, 589.
. The court found that, apart from the issue of their confidentiality, the INPO reports came within the scope of Exemption 4 because they were "commercial” and had been "obtained from a person." Id.
. There had been no contention that disclosure of the reports would have had adverse competitive effects upon INPO.
. Indeed, the interests of the Commission and INPO have been entirely in accord throughout this litigation; both parties have opposed disclosure of the INPO reports on the grounds that public access would impair their quality. If the need for limited circulation of these records is so tenuous, however, that it cannot be demonstrated, we would hardly expect INPO to jeopardize its valuable working relationship with the Commission over their disclosure.
. Thus, we made it clear in Critical Mass I that, in order to prove "confidentiality" under Exemption 4, "the agency must demonstrate that disclosure will harm a specific interest that Congress sought to protect by enacting the exemption.” 830 F.2d at 281 (internal quotation omitted). This is an objective test, and it is not limited by the parties’ preferred means of doing business in connection with their exchange of information.
. Under 10 C.F.R. § 50.73 (1990), licensees must prepare and submit Licensee Event Reports ("LERs”) within 30 days of discovery of any safety-related "reportable event” containing objective data concerning that event.
. Nor is it entirely clear to what extent employees in fact risk exposure by speaking to INPO *947representatives. The reports do not identify the sources of the information on which they are based, and subjective data from interviewed employees enters the final reports only after being sifted and filtered by the analytical process through which the reports are prepared. Thus, insofar as the Commission or INPO may claim that employee fear of retaliation would inhibit frank discussion, it is unclear whether any such fear would be realistic. But see Transcript of Deposition of Terence J. Sullivan, reprinted in J.A. at 448-49 (containing the statement of INPO’s Executive Vice President that, in at least some cases, an informed reader might be able to deduce the names of employees involved in reported incidents).