OPINION and ORDER

POGUE, Senior Judge:

This consolidated action arises from the U.S. Department of Commerce's ("Commerce") countervailing duty ("CVD") investigation of aluminum extrusions from the People's Republic of China ("China").2 Before the court are the results of Commerce's redetermination on remand of the "all-others" CVD rate, pursuant to the Court of Appeals' decision in MacLeanFogg Co. v. United States, 753 F.3d 1237, 1246 (Fed.Cir.2014) ("MacLean-Fogg V").3

The court has jurisdiction pursuant to Section 516A(a)(2)(B)(i) of the Tariff Act of 1930, as amended, 19 U.S.C. § 1516a(a)(2)(B)(i) (2012)4 and 28 U.S.C. § 1581(c) (2012).

As explained below, because Commerce's decision to rely on simple averaging when calculating the "all-others" rate in this case was an unreasonable judgment in the application of 19 U.S.C. § 1671d(c)(5)(A)(i), this determination is remanded for reconsideration.

BACKGROUND

Where, as here, a countervailing duty investigation involves a large number of exporters and/or producers as potential respondents, Commerce is authorized to select a sample of these exporters and producers for individual examination (the "mandatory" respondents).5 In addition, the remaining exporters and producers may request an individualized examination as "voluntary" respondents.6 Companies not selected as mandatory or voluntary respondents receive a CVD rate that is calculated for "all-other" companies (the "all-others" rate),7 which must equal the weighted average of all "individually investigated" companies' rates,8 unless all such rates are zero/ de minimis or entirely based on "facts otherwise available," rather than the respondents' own submissions.9 Consequently, Commerce generally constructs the all-others rates by using the weighted average of the mandatory respondents' rates.10

Following this statutory scheme, in the CVD investigation at issue here, Commerce selected the three companies exporting the largest volume of subject imports during the period of investigation as the mandatory respondents.11 However, none of these three companies responded to Commerce's requests for information.12 Commerce therefore found that the mandatory respondents "withheld requested information and significantly impeded [the] proceeding,"13 and failed to act to the best of their abilities to cooperate in the investigation.14 Accordingly Commerce established CVD rates for the mandatory respondents based entirely on adverse facts available ("AFA").15 Meanwhile, two companies had requested and were granted individualized examinations as voluntary respondents, each ultimately receiving a non-zero, non-de minimis, non-AFA CVD rate.16

With regard to the all-others rate, agency regulations in force at the time of the investigation prohibited Commerce from including the voluntary respondents' CVD rates in the all-others rate calculation.17 As this Court explained when upholding this regulation in Maclean-Fogg I, Commerce's basis for excluding the voluntary respondents' rates from the all-others rate calculation was the concern that voluntary respondents are unrepresentative of the remaining companies (particularly where, as here, the three largest exporters/producers did not respond to Commerce's inquiries at all).18 The agency considered the voluntary respondents to be unrepresentative because, unlike the mandatory or the all-other respondents, the voluntary respondents are those that willingly submit their sales data of their own accord, presumably because their commercial practices are such that they have good reason to believe that their CVD rates will be lower than those set for the mandatory respondents (regardless of whether those mandatory respondents are cooperative or not), such that including the rates established for this self-selected group threatens distortion of the weighted-average of the more representative rates.19 But the Court of Appeals reversed this decision,20 invalidated 19 C.F.R. § 351.204(d)(3), and ordered this court to remand Commerce's all-others rate calculation, requiring the agency to include the two voluntary respondents' rates when determining the all-others rate in this case "under the general rule, [19 U.S.C.] § 1671d(c)(5)(A)(i)."21

On remand, Commerce applied 19 U.S.C. § 1671d(c)(5)(A)(i), as interpreted by MacLean-Fogg V, excluding the three mandatory respondents' AFA-based rates from the all-others calculation, but including the two non-zero, non-de minimis, non-AFA based voluntary respondents' rates.22 Considering the two voluntary respondents' rates, however, Commerce found that it could not weight-average these rates without impermissibly revealing the two companies' business proprietary information ("BPI") to each other.23 Normally, in such situations Commerce "would calculate a weighted-average countervailing duty rate using the publicly available, ranged values of the [individually examined] respondents' exports of subject merchandise to the United States, compare both this weighted-average rate and a simple average of [these] respondents' countervailing duty rates to the actual weighted-average rate (calculated using the proprietary export values) and assign to All Others the amount closer to the actual weighted-average countervailable subsidy rate."24 But in this case, although agency regulations require that all BPI submissions, including numerical data, be accompanied by publicly available summaries,25 the two voluntary respondents did not submit public, "ranged" versions of their BPI.26 During its investigation, Commerce chose not to enforce this requirement because the agency's regulations then expressly prohibited using the voluntary respondents' countervailable subsidy rates to calculate the all-others rate.27 Commerce therefore "did not find that it was necessary during the underlying investigation to request the publicly-ranged or indexed numerical data from the voluntary respondents."28

Thereafter, however, as noted above, the Court of Appeals for the Federal Circuit held that countervailable subsidy rates calculated for voluntary respondents in CVD investigations unambiguously fall within the meaning of "countervailable subsidy rates established for exporters and producers individually investigated," as used in 19 U.S.C. § 1671d(c)(5)(A)(i), and therefore that such rates must be used in the calculation of an all-others rate, so long as they are not zero/de minimis or based entirely on facts otherwise available or AFA.29 On remand, finding that "the publicly ranged sales data that could be used to calculate a weighted average all others rate based on publicly available data [were] not on the administrative record," Commerce therefore "based the revised all others rate on a simple average of the two voluntary respondents' calculated net subsidy rates."30

In commenting on the remand results below, the Aluminum Extrusions Fair Trade Committee ("AEFTC") — a petitioner in the underlying countervailing duty investigation and an intervenor in this action31 — argued, inter alia, that Commerce should have calculated the all-others rate using a weighted average of the two voluntary respondents' rates, contending that 19 C.F.R. § 351.304(c)(1) requires parties to submit publicly ranged versions of their BPI data, and that Commerce should therefore "reopen the record to obtain the publicly ranged data that is necessary for [Commerce] to calculate a weighted average all others rate."32 Commerce, however, declined to reopen the record.33 Rather, Commerce found that the voluntary respondents' publicly ranged sales data was neither "necessary [nor] warranted" because "the use of a simple average of the two voluntary respondents' net subsidy rates to calculate the all others rate is consistent with [Commerce's] practice in cases in which the publicly available ranged sales data are not on the record."34 AEFTC now challenges this determination.35 Alternatively, AEFTC also argues that Commerce should have established a single subsidy rate for both voluntary respondents, because the companies are affiliated.36

STANDARD OF REVIEW

The court will sustain Commerce's countervailing duty determinations on remand if they are in accordance with the remand order, are supported by substantial evidence, and are otherwise in accordance with law.37 Where the statute and regulations leave the agency with a measure of discretion, the court reviews such decisions for abuse of discretion.38 "An abuse of discretion occurs where the decision is based on an erroneous interpretation of the law, on factual findings that are not supported by substantial evidence, or represent an unreasonable judgment in weighing relevant factors."39 Moreover, Commerce's discretion "is bounded at the outer limits by the obligation to carry out its statutory duty of `determining dumping margins "as accurately as possible."'"40

DISCUSSION

AEFTC argues that Commerce's calculation of the all-others rate on remand — based on a simple average of the two voluntary respondents' subsidy rates — was contrary to law because (I) Commerce had found these two companies to be affiliated, and so should have established a single rate for both and then used that rate as the all-others rate41; or, in the alternative, because (II) Commerce used a simple (rather than weighted) average of the two voluntary respondents' subsidy rates, contrary to 19 U.S.C. § 1671d(c)(5)(A)(i).42 Each argument is addressed in turn.

I. Affiliation

AEFTC first argues that because Commerce found that the two voluntary respondents were affiliated, Commerce should have established a single rate for both and then used that rate as the all-others rate.43 Commerce responds that its finding that the two voluntary respondent companies were "affiliated persons" under 19 U.S.C. § 1677(33) is insufficient to compel the conclusion that these companies form a single entity for which a single subsidy rate is appropriate.44 Specifically, Commerce's regulations provide that the agency will generally attribute subsidies to (and hence establish CVD rates for) "the products produced by the corporation that received the subsidy,"45 except "[i]f two (or more) corporations with cross-ownership produce the subject merchandise."46 Where two or more corporations "with cross-ownership" produce the subject merchandise, Commerce "will attribute [and countervail for] the subsidies received by either or both corporations to the products produced by both corporations," using a single CVD rate.47

The regulations further provide that "[c]ross-ownership exists between two or more corporations where one corporation can use or direct the individual assets of the other corporation(s) in essentially the same ways it can use its own assets,"48 adding that this standard will normally be met "where there is a majority voting ownership interest between two corporations or through common ownership of two (or more) corporations."49 Moreover, in explaining these regulatory provisions, Commerce has expressly stated that mere affiliation between two companies, within the meaning of 19 U.S.C. § 1677(33), is insufficient for a finding of cross-ownership which would support the establishment of a single subsidy rate for two or more companies.50

Here, Commerce found that the record of this investigation contained "no evidence" that the two voluntary respondents "have the ability to direct the individual assets of one another as if they were their own."51 Commerce accordingly concluded that cross-ownership among these companies was not established, and hence determined that a single subsidy rate for both would not be appropriate.52 As AEFTC has not provided any "new information or argument that would warrant reconsideration ... on this point,"53 Commerce's finding that the two voluntary respondents do not meet the regulatory criteria for cross-ownership is not contested.54 Furthermore, AEFTC presents no specific argument to support its contention that, contrary to the agency's regulations, a finding of affiliation is sufficient to require Commerce to assign a single subsidy rate to the affiliated companies.55

Accordingly, because Commerce's regulations provide that only companies that are cross-owned may have their subsidies attributed to one another56 (the reasoned basis for which57 is not explicitly challenged here58), and because AEFTC provides no evidentiary support or argument to impugn Commerce's finding that the two voluntary respondents here are not cross-owned,59 Commerce's determination to calculate separate subsidy rates for these companies is therefore sustained.

II. Simple Averaging

Next, in the alternative, AEFTC argues that Commerce acted contrary to law by using a simple average of the two voluntary respondents' subsidy rates — rather than a weighted average, as required by 19 U.S.C. § 1671d(c)(5)(A)(i) — to establish the all-others rate.60 Specifically, AEFTC argues that Commerce unreasonably determined that weight-averaging the rates would impermissibly reveal BPI, because the agency's regulations require that all BPI submissions have correlating public versions, and provide a methodology for converting BPI into public information.61 AEFTC argues that Commerce abused its discretion by neither requesting that the respondents apply this methodology to their BPI and submit public versions of the necessary data, nor itself applying this methodology to convert the BPI to usable data.62

Commerce responds by pointing to 19 U.S.C. § 1677f(b) — which provides that information designated as BPI "shall not be disclosed to any person without the consent of the person submitting the information," other than to certain U.S. Government officials and authorized applicants under an administrative protective order63 — explaining that "[a]t no point during the investigation did counsel for either voluntary respondent authorize Commerce to reveal proprietary information to the other,"64 and arguing that weight-averaging these subsidy rates to arrive at the all-others rate was therefore foreclosed by the fact that doing so would reveal the two companies' BPI to each other.65

Although Commerce acknowledges that weight-averaging would be possible (without improperly divulging the two respondents' BPI to each other) if the BPI were accompanied with public versions of the data, as required by 19 C.F.R. § 351.304(c)(1),66 such information is not on the record here because the voluntary respondents did not submit public versions of their BPI, and Commerce "did not find that it was necessary during the underlying investigation to [enforce 19 C.F.R. § 351.304(c)(1) and] request the publicly-ranged or indexed numerical data from the voluntary respondents" because, at that time, 19 C.F.R. § 351.204(d)(3) expressly prohibited using the voluntary respondents' rates in the all-others rate calculation.67 On remand, responding to AEFTC's suggestion that Commerce re-open the record to obtain these missing publicly ranged data, so the two voluntary respondents' rates may be weight-averaged without divulging BPI to each other, the agency continued to find that such data were neither "necessary [n]or warranted"68 because using a simple average of the two rates, rather than "expend[ing] additional administrative resources and further delay[ing] the ultimate resolution of this proceeding," was a "reasonable choice."69

But the statute unequivocally and without exception requires that Commerce base the all-others rate on the weighted average of individually-investigated non-zero, non-de minimis, non-AFA rates.70 Defendant argues that "the statute and regulation are silent as to Commerce's methodology for calculating an all-others rate when using data derived from two respondents in a weighted-average calculation would divulge the firms' business proprietary data to each other, in violation of the administrative protective order."71 But in fact the statute (requiring Commerce to use a weighted-average when calculating the all-others rate72) and the regulations (requiring publicly ranged data for all BPI submissions73) preclude the situation Commerce describes. If "calculating an all-others rate when using data derived from two respondents in a weighted-average calculation would divulge the firms' business proprietary data to each other,"74 19 U.S.C. § 1671d(c)(5)(A)(i)'s requirement that Commerce use a weighted average may be satisfied by employing the publicly ranged data.75

Commerce argues that the missing public data here are nevertheless unnecessary because using a simple average is also reasonable.76 But this argument ignores the accuracy-enhancing value placed by the statute on accounting for the individually-investigated respondents' relative sales volumes by weight-averaging to arrive at the all-others rate.77 Moreover, as AEFTC points out, here there was a significant disparity in the volume of sales between the two respondents whose rates were averaged to arrive at the all-others rate.78 Accordingly, taking a simple average of the two gives significantly more weight to the respondent with the lower sales volume, resulting in an all-others rate that is materially different from what it would otherwise be if it were properly weighted based on the relative size of each respondent's total sales.79 Thus Commerce's argument — that completing the record with publicly ranged values of the two voluntary respondents' BPI was neither necessary nor warranted because a simple average would suffice — is unreasonable in light of the statute's clear preference for the accuracy-enhancing value of weight-averaging and the particular facts of this case.80

Additionally, Commerce incompletely characterizes its practice in cases where weight-averaging two respondents' rates would impermissibly reveal their BPI to each other as either taking a simple average of the two or taking an average weighted by the respondents' publicly ranged sales values.81 In fact Commerce's reasonable practice in such situations is to take both averages and compare each to the actual weighted-average (using BPI available to the agency), in order to arrive at the nearest approximation of the all-others rate contemplated by 19 U.S.C. § 1671d(c)(5)(A)(i).82 This practice reasonably reconciles 19 U.S.C. § 1671d(c)(5)(A)(i) (the all-others provision) with 19 U.S.C. § 1677f(b) (the BPI provision) because it most closely approximates the result contemplated by the all-others provision without violating the BPI provision. And while Commerce relies on a handful of recent contrary determinations, in which the agency has resorted to simple averaging in the absence of publicly available data,83 none of these decisions provide any significant reasoning or explanation to indicate why the necessary public information was absent from the record or why it could not be obtained, or how such simple averaging comports with the statutory directive to weight-average individual rates when calculating the all-others rate,84 and accordingly none of these determinations supports the agency's argument here.

That the necessary public data are absent from the record here is due to Commerce's own decision not to enforce its regulatory requirement and request the necessary data from the submitting parties.85 Commerce initially "did not find that it was necessary" to complete the public record, including public versions of the voluntary respondents' BPI, because at the time 19 C.F.R. § 351.204(d)(3) precluded using the voluntary respondents' data in the all-others rate calculation.86 But even then Commerce faced a challenge to the legality of this regulation,87 and it was unreasonable to presume that the ultimate outcome of this litigation would favor the agency, as indeed it did not.88

Moreover, even if the regulation excluding the voluntary respondents' rates from the all-others rate calculation had not been challenged and invalidated, and even if Commerce had been correct that the voluntary respondents' information would be used solely to calculate their own individual rates, Commerce also failed to consider another important concern. By leaving the record of the voluntary respondents' rates calculations sealed from public scrutiny (because the voluntary respondents' rates are based on their non-public information), Commerce failed to recognize the value of ensuring that all aspects of the administrative record — including the evidentiary bases for the voluntary respondents' rates themselves (regardless of whether or not they are included in the all-others rate) — are as publicly available as they can be. There is "a fundamental public interest in transparency in government,"89 and "[t]he parties to a lawsuit are not the only people who have a legitimate interest in the [public] record compiled in a legal proceeding."90 "[L]itigants in other similar cases have a legitimate need (and a right) to review the facts underlying a[n agency]'s decision" in order to "discern the relevance and significance of [that] decision vis-à-vis their own cases."91 Here, Commerce unreasonably failed to weigh the importance of completing the public record with regard to these voluntary respondents, even if the agency were correct that their subsidy rates would not be included in the all-others rate, for by doing so Commerce automatically precluded potential future interested parties from ascertaining the reasoning underlying the voluntary respondents' rates.

Finally, Commerce's decision on remand not to expend the minimal effort required to correct the error and obtain the missing public versions of the necessary BPI was, based on the record here, clearly an unreasonable exercise of judgment. Although Commerce initially saw no apparent need for the public data, this was no longer true at the time of the remand proceeding. 19 C.F.R. § 351.304(c)(1), which requires all BPI to be accompanied by public versions "in sufficient detail to permit a reasonable understanding of the substance of the information," provides a clear and unambiguous formula for converting proprietary numerical data into publicly available summaries thereof: "numerical data will be considered adequately summarized if grouped or presented in terms of indices or figures within 10 percent of the actual figure," and "[i]f an individual portion of the numerical data is voluminous, at least one percent representative of that portion must be summarized."92 Accordingly, all that Commerce had to do to follow the statutory prescription that individual rates be weight-averaged to arrive at the all-others rate, without impermissibly revealing BPI, was to send a letter to each of the voluntary respondents, referencing the regulation and requesting that their BPI be publicly ranged in accordance with the provided formula for doing so, or even to simply itself apply the formula to the BPI. Instead the agency chose to forego the statutory requirement and distort the accuracy of its all-others rate calculation. Thus to the extent that Commerce had discretion to avoid enforcement of the public versions requirement, the agency failed to exercise its judgment reasonably, and therefore abused its discretion.

Accordingly, while Commerce is correct that it has no general duty to reopen the record during remanded proceedings,93 the particular circumstances presented here require that, where the necessary data is missing from the record due to Commerce's own failure to fully administer the legal framework, it is the agency's responsibility to go back and fix errors that are material to the remand proceeding. Commerce's determination to use a simple average of the two individually-investigated respondents' subsidy rates in this case is therefore remanded for reconsideration. On remand, Commerce may either reopen the record and enforce 19 C.F.R. § 351.304(c)(1)'s requirement that the voluntary respondents submit public versions of their BPI, or (as AEFTC suggests94) itself publicly range the BPI, if the agency finds that doing so is appropriate here.

CONCLUSION

For all of the foregoing reasons, Commerce's determination to rely solely on simple averaging when calculating the all-others rate in this case pursuant to 19 U.S.C. § 1671d(c)(5)(A)(i) is remanded for reconsideration, consistent with this opinion. Commerce shall have until September 24, 2015, to complete and file its remand results, the parties shall have until October 9, 2015, to file any comments, and the agency shall have until October 19, 2015, to respond.

It is SO ORDERED.