In this action challenging a final determination issued by the United States Department of Commerce ("Commerce") in Commerce's Third Administrative Review of the countervailing duty order on crystalline silicon photovoltaic cells, whether or not assembled into modules ("solar cells") from the People's Republic of China ("PRC"), covering the period from January 1, 2014, through December 31, 2014. See Crystalline Silicon Photovoltaic Cells, Whether or Not Assembled Into Modules, From the People's Republic of China: Final Results of Countervailing Duty Administrative Review, and Partial Rescission of Countervailing Duty Administrative Review; 2014, 82 Fed. Reg. 32,678 (Dep't Commerce July 17, 2017) ("Final Results"), amended by Crystalline Silicon Photovoltaic Cells, Whether or Not Assembled Into Modules, From the People's Republic of China: Final Results of Countervailing Duty Administrative Review, and Partial Rescission of Countervailing Duty Administrative Review; 2014, 82 Fed. Reg. 46,760 (Dep't Commerce Oct. 6, 2017) ("Amended Final Results"), Changzhou Trina Solar Energy Co., Ltd., Trina Solar (Changzhou) Science & Technology Co., Ltd. (collectively, "Trina"), Consolidated Plaintiffs BYD (Shangluo) Industrial Co., Ltd. ("Shangluo BYD") and Shanghai BYD Co., Ltd. ("Shanghai BYD") (collectively, "BYD"); and Canadian Solar Inc., Canadian Solar International, Ltd., Canadian Solar Manufacturing (Changshu), Inc., Canadian Solar Manufacturing (Luoyang), Inc., Canadian Solar (USA) Inc., CSI Cells Co., Ltd., CSI Solar Power (China) Inc., CSI Solartronics (Changshu) Co., Ltd., CSI Solar Technologies Inc., and CSI Solar Manufacture Inc. (collectively, "Canadian Solar") request the court hold aspects of Commerce's final determination to be unsupported *1323by substantial evidence or otherwise not in accordance with law.
The United States ("Defendant") asks that the court sustain Commerce's Final Results of its third administrative review. SolarWorld Americas, Inc. ("SolarWorld") requests the court to uphold other portions of Commerce's Final Results as supported by substantial evidence and otherwise consistent with law and asserts that other portions are not.
BACKGROUND
Commerce first published a countervailing duty order on solar cells from the People's Republic of China ("PRC") on December 7, 2012. See Crystalline Silicon Photovoltaic Cells, Whether or Not Assembled Into Modules, From the People's Republic of China: Countervailing Duty Order, 77 Fed. Reg. 73,017 (Dep't Commerce Dec. 7, 2012). In 2016, Commerce initiated its third administrative review of this countervailing duty, covering the period from January 1, 2014 to December 31, 2014. Canadian Solar and Trina were selected as mandatory respondents ("Respondents") and issued questionnaires along with the Government of the PRC ("GOC"). See Decision Memorandum for Final Results and Partial Rescission of Countervailing Duty Administrative Review: Crystalline Silicon Photovoltaic Cells, Whether or Not Assembled Into Modules, from the People's Republic of China; 2014, 1 (Dep't Commerce July 10, 2017) ("I & D Memo"). On January 9, 2017, Commerce published its preliminary results of the administrative review. Crystalline Silicon Photovoltaic Cells, Whether or Not Assembled Into Modules, From the People's Republic of China, 82 Fed. Reg. 2,317 (Dep't Commerce Jan. 9, 2017) (Prelim. Results) and accompanying issues and decision memorandum (Prelim. I & D Memo ). After receiving submissions from interested parties, Commerce issued its Final Results on July 17, 2017, later amended on October 6, 2017. 82 Fed. Reg. 32,6781 ; Amended Final Results, 82 Fed. Reg. 46,760. The Amended Final Results calculated a subsidy rate of 18.16 ad valorem for Canadian Solar, 17.14 percent ad valorem for Trina, and 17.49 ad valorem for non-selected companies under review. 82 Fed. Reg. at 46,761. Plaintiff and Plaintiff-Intervenors challenge several aspects of the Final Results, as amended.2
JURISDICTION AND STANDARD OF REVIEW
The court has jurisdiction pursuant to 28 U.S.C. § 1581(c) and 19 U.S.C. § 1516a(a)(2) (2012). Commerce's results in a countervailing duty proceeding are upheld unless "unsupported by substantial evidence on the record, or otherwise not in accordance with law[.]" 19 U.S.C. § 1516a(b)(1)(B)(i).
*1324DISCUSSION
Because the parties present a variety of fact-specific claims, the following opinion addresses the factual background for each in turn. Each section notes which parties are bringing a given claim.
I. Export Buyer's Credit Program
a. Commerce's decision to apply adverse facts available to cooperating parties
i. Background
In the course of the third administrative review, Commerce requested information about the Exports Buyer's Credit Program ("EBCP') from the GOC. Canadian Solar, and Trina. See Prelim I & D at 2-3.3 The latter two submitted affiliate and customer certifications of non-use applicable to the period of review stating that U.S.-based customers had not benefitted from the EBCP. See Trina Section III Questionnaire Response at 75, P.D. 81-98 (May 3, 2016); Trina Benchmark Submission (BPI Version) at Ex. 10., C.D. 105-107 (Nov. 30, 2016); Canadian Solar Section III Questionnaire Response at Vol. II, Ex. 20, P.D. 116-22, C.D. 31-37 (June 10, 2016). The GOC, however, refused to provide information on potential third-party bank involvement in the EBCP,4 arguing that such information was irrelevant to Commerce's determination regarding whether the program had been used by the relevant parties. I & D Memo at 13.
Unlike in the second administrative review,5 in which Commerce declined to apply an adverse inference with regard to facts otherwise available ("AFA") against otherwise cooperating respondents based on the GOC's refusal to provide requested EBCP information, Commerce here concluded that the intervening 2013 revisions6 to the EBCP made Respondents' certifications of non-use insufficient to establish non-use. See I & D Memo at 13; Prelim. I & D Memo at 30-31. The 2013 change to the program allowed for the involvement *1325of third party banks in the EBCP and Commerce reasoned that, given the GOC's refusal to answer questions regarding whether and how these banks extended credit, it was impossible to verify Respondents' certifications of non-use. See I & D Memo at 13. Commerce found that "[a]bsent the requested information, the GOC's claims that the respondent companies did not use this program [were] not reliable" and therefore applied AFA to all parties in calculating the amount of subsidization based on the EBCP. Id.
Trina and Canadian Solar argue that Commerce disregarded record evidence showing that they did not receive support from the EBCP. Trina Compl. at ¶ 10; Trina Br. at 7-9; Canadian Solar Br. at 8-14. Canadian Solar further argued that it was improper to use AFA against a cooperating party and that Commerce's decision is arbitrary as it contradicts Commerce's previous rulings in similar situations. Canadian Solar Br. at 9, 12-13. Trina disputes Commerce's assertion that the non-use of third party banks was unverifiable and contends that Commerce could have requested further documentation on Trina's loans in order to verify non-use. Trina Br. at 14; Reply Brief of Plaintiffs Changzhou Trina Solar Energy Co., and Trina Solar (Changzhou) Science and Technology Co., Doc No. 79 ("Trina Reply Br.") at 11-13 (Sept. 21, 2018). Canadian Solar argues that the affidavits of non-use were sufficient and no other documentation was necessary. Canadian Solar Br. at 10-12. Defendant claims that Commerce was correct in finding that respondents used the EBCP based on an adverse inference given the GOC's failure to cooperate fully, and that Commerce was under no obligation to attempt to verify Canadian Solar or Trina's submissions. Def. Br. at 10-20.
ii. Discussion
When an interested party "withholds information that has been requested" by Commerce, Commerce may need to "use the facts otherwise available" to reach a decision. 19 U.S.C. § 1677e(a)(2)(A). Under 19 U.S.C. § 1677e(b) an adverse inference as to what such facts show may be used only if a party has failed to cooperate to the best of its ability, meaning that a party has failed to "put forth its maximum effort to provide Commerce with full and complete answers to all inquiries in an investigation." Nippon Steel Corp. v. United States, 337 F.3d 1373, 1382 (Fed. Cir. 2003). As the court has previously held in a similar case, if a foreign government fails to cooperate in a countervailing duty case, Commerce may apply AFA even if the collateral effect is to "adversely impact a cooperating party." Archer Daniels Midland Co. v. United States, 917 F.Supp.2d 1331, 1342 (CIT 2013). Commerce, however, should "seek to avoid such impact if relevant information exists elsewhere on the record." Id.
The court recently issued two opinions regarding the Export Buyer's Credit Program Changzhou Trina Solar Energy Co. v. United States, 195 F.Supp.3d 1334 (CIT 2016) ("Changzhou I") and Guizhou Tyre Co. v. United States, Slip Op. 18-140, 348 F.Supp.3d 1261, 2018 WL 5307676 (CIT Oct. 17, 2018) that at least facially seem to conflict. In Changzhou I, the court upheld Commerce's use of AFA in determining that respondents used the program. 195 F.Supp.3d at 1355. The court based its decision on Commerce's reasonable explanation that an understanding of how an exporter would be involved in the program was necessary to determine usage and that the GOC failed to cooperate in providing this information. Id. In contrast, in Guizhou the court found Commerce's explanation regarding the use of the EBCP by *1326respondents unavailing. Slip 352 F.Supp.3d at 1270-72, 2018 WL 5307676, at *4. In that case, Commerce found non-use submissions by respondents unverifiable because the GOC refused to provide documents regarding the functioning of the EBCP under the GOC's new regulations.7 Id. The court found that Commerce failed to show there was a gap in the record warranting the use of AFA and that Commerce was "conflat[ing] the program's operation with its use," and thus remanded for reconsideration. Id. at 1270-72, 1280-81, at *4, *12. Although both opinions share some factual similarities to this case, neither is fully dispositive given the record before the court.
In Changzhou I, unlike here, the respondents did not submit customer certifications of non-use, so that issue was not before the court. See Countervailing Duty Investigation of Certain Crystalline Silicon Photovoltaic Products From the People's Republic of China: Final Affirmative Countervailing Duty Determination, 79 Fed. Reg. 76,962 (Dep't Commerce Dec. 23, 2014), accompanying Issues and Decision Memorandum at 93 ("Changzhou I I & D Memo"). In addition to the lack of complete documentation of non-use, in that review Commerce supplied detailed reasoning for why documentation from the GOC was necessary. Changzhou I I & D Memo at 91-94. Here, Commerce provided reasoning as to why the GOC's failure to respond adequately made it impossible for it to understand fully the operation of the EBCP, but it failed to show why a full understanding of the EBCP's operation was necessary to verify non-use certifications.
In Guizhou, as here, Commerce found that the GOC's refusal to explain if and how third-party banks were involved in the EBCP made respondent's claims of non-use unverifiable. Certain New Pneumatic Off-the-Road Tires From the People's Republic of China: Final Results of Countervailing Duty Administrative Review; 2014, 82 Fed. Reg. 18,285 (Dep't Commerce Apr. 18, 2017) accompanying Issues and Decision Memorandum at 23-24 ("Guizhou I & D Memo"). In the Guizhou court's review of that determination, it found that the GOC's lack of response was "rendered immaterial by responses from [respondents]." 352 F.Supp.3d at 1269-71, 2018 WL 5307676, at *3. The court here does not resolve the materiality issue at this juncture, rather it finds that Commerce simply failed to provide reasoning sufficient for this court to find that its determination was supported by substantial evidence. In this case, Commerce claims its ability to verify the certifications was stymied by a lack of understanding of if and how third party banks were involved in the distribution of loans. I & D Memo at 13. Commerce, however, did not explain why the GOC's failure to explain this program was necessary to assess claims of non-use and why other information accessible to respondents was insufficient to fill whatever gap was left by the GOC's refusal to provide internal bank records. Further, Commerce did not explain how an adverse inference regarding the operation of the EBCP logically leads to a finding that respondents used the program. The use of facts available allows Commerce to render determinations when information is missing and it may use an adverse inference if respondents fail to cooperate, but is *1327not permitted to skip either of these steps on the way to use of AFA.
Under 19 U.S.C. § 1677e(b) Commerce may use AFA to choose among facts of record, but the choice must fill in the information that is actually missing. Further, although it is true that Commerce need not consider information submitted by respondents that cannot be verified, Commerce must first reasonably show that such information is, in fact, unverifiable. See 19 U.S.C. § 1677m(e) ; see also Papierfabrik August Koehler SE v. United States, 843 F.3d 1373, 1382-83 (Fed. Cir. 2016) (holding that if the requirements of § 1677m(e) are not met, Commerce need not consider information submitted by an interested party). What type of information requested from the GOC would have made these claims verifiable? And what information, such as loan agreements and other relevant documents ostensibly held by Respondents, might have sufficed to provide Commerce the assurance it needed? Commerce does not explain.
Accordingly, the court remands this matter and instructs Commerce to explain what information specifically the GOC failed to provide that led it to resort to facts available and the facts as to which it drew an adverse inference in arriving at the conclusion that Respondents benefited from the EBCP. Commerce should further explain if and how certifications of non-use are unverifiable in the absence of the GOC's cooperation. If Commerce determines that it is able to verify non-use by not unreasonably onerous means, the court instructs it to do so.
b. Adverse rate selected for Export Buyer's Credit Program
i. Background
If Commerce continues to find that respondents used the EBCP, and the court sustains that determination, then the court must assess whether the rate selected for the EBCP is supported by substantial evidence. As with several other calculation issues addressed here, in the interest of judicial and attorney economy the court addresses this issue, which has been briefed fully. Canadian Solar contests Commerce's calculation of an adverse rate for the program. Canadian Solar Br. at 14-17. After finding no program identical to the EBCP in the same administrative review, Commerce identified a similar program in the same proceeding to use as a basis for calculating the rate for the EBCP. I & D Memo at 18-19. Commerce calculated a rate of 5.46 percent ad valorem, for the EBCP by utilizing the rate "calculated for company respondent Lightway Green New Energy Co., Ltd.'s usage of the Preferential Policy Lending to the Renewable Energy Industry program in the 2012 administrative review of this proceeding." Prelim I & D Memo at 32. Commerce explained that the Lightway Green New Energy Co. Policing Lending Program ("Lightway Program") was similar because both it and the Export Buyer's Credit Program provided access to loans. I & D Memo at 19.
Canadian Solar challenges the use of the Lightway Program rate, arguing that it is not an appropriately similar program. Canadian Solar Br. at 15. China's Ex-Im Bank is the administrator of the EBCP, a program that provides credit to foreign importers of Chinese products and loans. See I & D Memo at 12-13. Canadian Solar argues the record contains evidence of a more similar program, the Export Seller's Credit Program, which provides a more directly comparable rate than the Lightway Program. Canadian Solar Br. at 15-17. Canadian Solar argues that the Lightway Program is not a similar program in that, although it calls for financial institutions *1328to offer loans to renewable energy, these loans are not specifically related to exports. Canadian Br. at 16. They argue that Commerce should have further explained why it chose the Lightway Program over the Export Seller's Credit Program. Canadian Solar Br. at 15, 17. The Defendant argues that Commerce acted in accordance with its established practice in selecting the Lightway Program. Def. Br. at 21-24.
ii. Discussion
Commerce has discretion when calculating the appropriate AFA rate, as neither the relevant statute nor regulations limit how Commerce must select programs that are "similar," or if no similar program exists, one "from a proceeding that the administrating authority considers reasonable to use." 19 U.S.C. § 1677e(d)(1) ; see SolarWorld Americas, Inc. v. United States, 229 F.Supp.3d 1362, 1366 (CIT 2017).8 Commerce has developed a methodology by which to determine the appropriate AFA rate in accordance with the governing statute. See SolarWorld Americas, 229 F.Supp.3d at 1366 (describing Commerce's five-step hierarchy in selecting an AFA rate). First, Commerce assesses whether another cooperating company in the same proceeding used the identical program, and if so, Commerce applies the "highest non-de minimis rate calculated for a cooperating company for the identical program in the same proceeding." Id. If no such figure exists, Commerce applies "the highest non-de minimis rate calculated for a cooperating company for a similar program in the same proceeding." Id. If no such figure can be calculated, then Commerce continues to step three, and if necessary, step four and five until it arrives at the appropriate metric. Id. 9
Here, Commerce was unable to satisfy step one of the methodology as there was no alternative rate for the EBCP in the Third Administrative Review. See Prelim. I & D Memo at 30-32; I & D Memo at 18-21. Thus, Commerce turned to step two and found a sufficiently similar program from an earlier administrative review, the Lightway Program. Prelim. I & D Memo at 31-32; I & D Memo at 19. Commerce predicated this finding of similarity on both the EBCP's and the Lightway Program's distribution of loans. I & D Memo at 19. With this finding, Commerce applied the rate from the Lightway Program to calculate an AFA rate for the EBCP. Id. Canadian Solar's argument that *1329Commerce needed to compare the Lightway Program to Canadian Solar's proffered Export Seller's Credit Program fails. See Canadian Solar Br. at 15. Under Commerce's established methodology and consistent with the plain text of the statute, Commerce selects a similar program, not necessarily the most similar program. See 19 U.S.C. § 1677e(d)(1)(A)(i). Additionally, Canadian Solar argues that Commerce failed to explain why the Lightway Program was similar. Canadian Solar Br. at 17; but see I & D Memo at 19 (stating that the Lightway Program was chosen because it, like Export Buyer's Credit Program, was a loan program). Although a more detailed description might be helpful, it is not required.
Commerce has broad discretion in determining and applying an AFA rate, so long as it "reasonably balance[s] the objectives of inducing compliance and determining an accurate rate." SolarWorld Americas, 229 F.Supp.3d at 1366. Commerce used its developed methodology to arrive at the AFA rate of 5.46 percent. The court finds no error in this regard.
II. Provision of Aluminum for LTAR
a. Specificity determination of aluminum program
i. Background
Canadian Solar claims that the provision of aluminum extrusions for less than adequate remuneration (LTAR) is not properly classified as a specific subsidy. Canadian Solar Br. at 26. Commerce found, as it did in the second administrative review, that the provision of aluminum extrusions for less than adequate remuneration was a de facto specific subsidy because the industries that used aluminum extrusions were "limited in number" and no new information disturbed that finding. I & D Memo at 22.10 Both Commerce and Canadian Solar agree that the six categories of industries that use aluminum extrusions are: "(1) building and construction, (2) transportation, (3) electrical, (4) machinery and equipment, (5) consumer durables, and (6) other industries." I & D Memo at 21; Canadian Solar Br. at 27.
Canadian Solar argues that the named industries listed as using aluminum extrusions are themselves diverse and also that Commerce failed to inquire as to whether the number of industries in the "other industries" category would render the subsidy non-specific. Canadian Solar Br. at 26-27. According to Canadian Solar, Commerce was obligated to conduct a more searching analysis regarding the diversity of these industries as the industries identified are broad categories that contain numerous sub-industries. Id. at 27-28. SolarWorld argues that the catchall "other industries" by itself does not mean the subsidy was widely used by numerous other industries and that Canadian Solar's relies on outdated case law. SolarWorld's Response to Plaintiff's Rule 56.2 Motion for Judgment upon the Agency Record, Doc. No. 69 at 27-28 (July 20, 2018) ("SolarWorld Resp."). The Government responds that the information received from GOC shows that the "recipients of aluminum extrusions (on an industry basis) are limited in number." Def. Br. at 25.
ii. Discussion
Commerce is empowered to assess countervailing duties if, after investigating *1330a subsidy, it finds that the subsidy "1) provides a financial contribution to a person, 2) a benefit is thereby conferred, and 3) the subsidy is specific." Bethlehem Steel Corp. v. United States, 223 F.Supp.2d 1372, 1378 (CIT 2002) ; see also 19 U.S.C. § 1677(5). No party challenges Commerce's determinations regarding financial contribution or benefit conferred. Domestic subsidies are specific, and thus countervailable, when "[t]he actual recipients of the subsidy, whether considered on an enterprise or industry basis, are limited in number." 19 U.S.C. § 1677(5A)(D)(iii)(I).11 In determining whether a subsidy is provided to a group of enterprises or industries, Commerce is not required to "determine whether there are shared characteristics among the enterprises or industries" that receive or are eligible for a subsidy: variety amongst the industries receiving a given subsidy is not the test for specificity. 19 C.F.R. § 351.502(b). That being said, Commerce is under an obligation to compare the industries receiving the subsidy to the industry makeup of the country at issue as a whole. This is a necessary step in the analysis in order to "avoid the imposition of countervailing duties in situations where, because of the widespread availability and use of a subsidy, the benefit of the subsidy is spread throughout an economy." Uruguay Round Agreements Act, Statement of Administrative Action, H.R.Rep. No. 103-316, vol. 1, at 930-31 (1994), reprinted in 1994 U.S.C.C.A.N. 4040, 4242 ("SAA").12 By way of analogy, if the GOC had instead said that only two sectors used aluminum extrusions - manufacturing and non-manufacturing - on a shallow level that would seem to satisfy the requirement that the subsidy be available to only a limited number of industries, but in reality such a distribution is not specific as those two categories encompass all possible industries.
Thus, although Commerce was under no obligation pursuant to its regulations to compare the characteristics of the six industries listed by the GOC against one another, Commerce should have determined whether these six industries made up a significant enough portion of all Chinese industries to render the subsidy nonspecific despite the use of only six categories to describe these industries.13 The categories mentioned here - building and *1331construction; transportation; electrical; machinery and equipment; consumer durables; and other industries14 - appear to represent a large swath of industries that could be further broken down into numerous sub-industries.15 Commerce needed to explain how subsidizing these broad industries amounts to a specific rather than a generally available subsidy. It is nonsensical to simply count the number of proffered industries, regardless of their composition, in order to determine specificity. Such a cursory test would allow gamesmanship in specificity determinations by allowing a respondent to simply recharacterize what is in fact a limited number of industries as numerous industries in order to avoid such a finding.
The court renders no decision on whether the provision of aluminum in the GOC is, in fact, a countervailable subsidy, but remands for Commerce to reconsider its methodology in arriving at this conclusion.
b. Challenge to Commerce's use of Comtrade and IHS datasets
i. Background
If, on remand, Commerce continues to find that the provision of aluminum amounts to a countervailable subsidy, it must additionally reconsider the data used to arrive at the appropriate benchmark. Canadian Solar and Trina challenge Commerce's decision to average data from IHS technology ("IHS") and UN Comtrade ("Comtrade") to determine the appropriate aluminum extrusion benchmark.
They contend that whereas IHS data specifically pertains to aluminum frames for solar modules, the type used by Trina and Canadian Solar, Comtrade data is broader and encompasses multiple broad Harmonized Tariff Schedule ("HTS") subheadings at the six-digit level (7604.21, 7604.29, and 7610.10)16 which covers many products not used by Respondents. Trina Br. at 15; Canadian Solar Br. at 29-31. Trina and Canadian Solar argue that the annual average figure provided by IHS provides a more accurate benchmark for the POR because it pertains specifically to aluminum used in the production of solar panels, and thus accords with Commerce's preference for the "narrowest category of products encompassing the input product." Trina Br. at 17; see also Canadian Solar at 29-31. Trina further argues that even if Commerce's normal preference for monthly average has merit, there is no way for Commerce to know whether Comtrade's monthly data more accurately reflects price fluctuations over time, because those *1332fluctuations may very well be caused by the irrelevant merchandise within the HTS subheadings. Trina Br. at 19.
SolarWorld counters that the inclusion of Comtrade data is necessary, given Commerce's preference for monthly over annual data in setting benchmark prices. SolarWorld Resp. at 31. SolarWorld argues that monthly data allows Commerce to match subsidy program pricing with world benchmark prices more accurately. Id. The Government argues that given the noncritical flaws in both datasets, averaging them was the appropriate course of action. Def. Br. at 26-31.
ii. Discussion
When goods are provided for LTAR, Commerce determines the amount of the subsidy by comparing remuneration actually paid with adequate remuneration. See 19 U.S.C. § 1677 (5)(E)(iv). Commerce employs a three-tiered hierarchy to determine the appropriate remuneration benchmark. In the absence of an undistorted tier one benchmark, e.g., an "actual transaction [price] in the country in question," Commerce turns to a tier two benchmark, e.g., a "world market price" for goods in question. C.F.R. § 351.511(a)(2); see also Essar Steel Ltd. v. United States, 678 F.3d 1268, 1273 (Fed. Cir. 2012). When there is more than one dataset representing the world market price, then "the Secretary will average such prices to the extent practicable, making due allowance for factors affecting comparability." 19 C.F.R. § 351.511(a)(2)(ii).17
In this instance, Commerce combined two datasets, IHS Technology and UN Comtrade, the latter of which uses broad HTS categories. Balancing Commerce's preference for monthly data and its desire for data specific to the relevant input, Commerce averaged these two datasets. But, in doing so, Commerce failed to properly make allowance for "factors affecting comparability." See 19 C.F.R. § 351.511(a)(2)(ii).
Commerce prefers monthly data points ostensibly to track potential market fluctuations over the period of review or investigation,18 but there is no statutory or regulatory basis for allowing this preference to overcome vital comparability concerns. This preference does not allow Commerce to include largely irrelevant data in its average of world market data sets. Put simply, not all flaws in data are equally problematic. Although some degree of nonspecificity is tolerable, when a dataset is vastly overinclusive of products not covered by the relevant CVD order, that flaw is not equivalent to the flaw in otherwise product-specific arising from its annual average. Here, Commerce made little effort to counter claims that Comtrade data was based on too broad a product category to provide an accurate world market price, stating only that Commerce was familiar with the data and that the HTS descriptions were suitable for constructing a world price for aluminum extrusions. I & D Memo at 23-24. Although Commerce often has to use less than ideal data, here it has the option to use only seemingly product-specific data.
Thus, the court concludes that Commerce's decision to average the Comtrade *1333and IHS datasets without properly considering whether the Comtrade data was too flawed to be probative of the world market price for the input at issue renders the decision unsupported by substantial evidence. Accordingly, the court remands this case to Commerce with instructions either to use solely the IHS dataset in its calculation of the appropriate benchmark or else explain why the inclusion of the Comtrade data does not produce a fatally inaccurate result.
III. Provision of Solar Glass for LTAR
a. Background
Canadian Solar and Trina similarly challenge Commerce's benchmark determination with regard to solar glass. Commerce found the GOC's solar glass market distorted, and so used world market indicators to calculate a benchmark for adequate remuneration. I & D Memo at 29. In its final calculation, Commerce averaged monthly data points from Comtrade19 with annually-reported data from IHS. I & D Memo at 29-30. Commerce defends this amalgam arguing that, although neither set was perfect, neither was so deficient as to merit rejection. Def. Br. at 31-36. Like the data used in setting the benchmark for aluminum extrusions, the Comtrade data in question contains monthly data points, but is less specific to solar glass, whereas the IHS data is an average annual figure, but one specific to solar glass. Id. at 32.
Canadian Solar and Trina contend that Commerce should not include the Comtrade data due to a critical lack of product specificity. Canadian Solar Br. at 21-23; Trina Br. at 22-23. They argue that Commerce should use only the IHS data given that its specificity to solar glass, whereas the Comtrade data includes HTS headings that include, but are not limited to, solar glass. Canadian Solar Br. at 17-19, see also Trina Br. at 20-23. They argue the glass included in these headings often does not possess the specific characteristics necessary for use in solar cells.20
Further, Canadian Solar argues that Comtrade's monthly benchmarks project a distorted picture of the solar market, showing price variability unsupported by the record and disputed by party submissions.21 See Canadian Solar Br. at 19-20. Canadian Solar acknowledges that including only Comtrade data from solar glass-producing nations makes the data less-flawed, but they further note that the Comtrade data does not contain data from certain major solar glass producing countries, including the United States, which is one of the top three solar glass producing countries globally. Canadian Solar Br. at 22-23. Finally, Trina contends that although the record indicates that there is a difference between broadly-defined tempered glass and the more specific solar glass, Commerce depended on the description of the headings referencing tempered glass as justification for its position that the headings included solar glass. Trina Br. at 23.
SolarWorld disagrees. It argues that the inclusion of the Comtrade data was correct *1334given Commerce's established preference for using monthly over annual data. SolarWorld Resp. at 33-34. It asserts that monthly-reported data better accounts for market fluctuations. Id. Canadian Solar responds that although monthly data points are preferable when available, there is no statutory requirement mandating that Commerce use monthly rather than annual data and contend that the use of IHS data is supported by the record. Canadian Solar Reply Brief in Support of Canadian Solar's Rule 56.2 Motion for Judgment upon the Agency Record, Doc. No. 80 at 10-11 (Sept. 21, 2018) ("Canadian Solar Reply"); Reply Brief of Trina, Doc. No. 79 at 8-10 (Sept. 21, 2018) ("Trina Reply").
b. Discussion
As with the above discussion of the proper datasets for calculating a benchmark for aluminum extrusions, Commerce similarly calculated its solar glass benchmark by averaging two tier-two datasets. Also as with the aluminum extrusions data, Commerce did not sufficiently determine the adequacy of these datasets or explicate their comparability.
Solar glass is a type of flat glass, which in turn is one of the [ [ ] ] main types of glass on the global market. Letter on Behalf of Canadian Solar to the Dep't Commerce re: Benchmark Submission at Ex. 5 C.R. 101-102 (No. 30, 2016). Flat glass represents about [ [ ] ]% of the global glass market, but of that [ [ ] ]% less than [ [ ] ]% is solar photovoltaics. Id. Thus the vast majority of flat glass is not suitable for solar panels, but is used in the production of other merchandise such as windows, glass doors, windshields, etc. See id. Comtrade data includes glass under the six-digit HTS headings 7007.19 and 7007.29.22 Neither of these HTS basket headings are specific to solar glass, but rather includes many types of safety glass (often a type of flat glass) unrelated to solar glass. Id. In contrast, the IHS data is specific to solar glass. Prelim. I & D Memo at 19.
In its brief, Defendant cited the court's earlier decision in Changzhou Trina Solar Energy Co., Ltd., v. United States, regarding the averaging of IHS data and Global Trade Atlas (GTA) data in a prior administrative review to arrive at the proper benchmark for solar glass. Slip. Op. 18-31, 2018 WL 1649629 (CIT Mar. 27, 2018) (" Changzhou II"); Def. Resp. Br. at 28. In that case, the court upheld Commerce's decision on remand to average IHS and GTA to arrive at a benchmark price for solar glass for reasons similar to those given in this case. See Changzhou II, 2018 WL 1649629 at *7. The court notes that plaintiffs challenge the use of the Comtrade data in this case, but there was no challenge to the use of GTA data in the previous case, and thus the court had neither the appropriate record nor any reason to inspect the adequacy of the GTA data. Id. at *6. Thus, the court finds the previous decision unhelpful.
Finally, the court finds Commerce's reliance on potential price fluctuations in the solar glass industry as a reason for including the Comtrade data unpersuasive. The only indication on the record before the court showing price fluctuations in solar glass is from the Comtrade data itself. In contrast, submissions by respondents during the administrative review show minimal solar glass price fluctuations. Canadian Solar Br. at 19-20. Commerce did not inquire into whether the fluctuations in the *1335Comtrade data were due to solar glass rather than other merchandise contained in the HTS headings. Without answering that threshold question, the court cannot be certain that the addition of the Comtrade data does not create the appearance of fluctuations in the solar glass market were none actually exist. Accordingly, the court cannot assess whether the inclusion of the Comtrade data makes the resulting benchmark more or less reliable.
Although averaging datasets is appropriate in certain circumstances, when one dataset is far more specific to the product at issue, that data may be more probative even if it is based on a yearly average rather than a monthly one. See 19 C.F.R. § 351.511(a)(2)(ii) (requiring the department to take into account "factors affecting comparability"). As with the datasets used to calculate the aluminum extrusions rate, Commerce failed to adequately evaluate and explain the relative adequacy and comparability of both datasets.23 Although Commerce may prefer to employ monthly data in its analysis, this cannot supersede regulatory considerations requiring the use of adequate data Here, Commerce failed to meaningfully assess the reliability of the Comtrade data and included it despite indications that it is overinclusive in regards to the types of glass included in the data, underinclusive in failing to include solar glass-producing countries, and by failing to assess whether the monthly fluctuations were due to price variability of solar glass or merely related to other merchandise contained in the HTS headings at issue.
In finding the Comtrade data to be a sufficient world market price metric without adequate evaluation, Commerce made a decision unsupported by substantial evidence or otherwise not in accordance with law. On remand, Commerce is instructed to use the IHS data alone in constructing a benchmark for the world market price for solar glass and otherwise address the court's concerns as to the Comtrade data and explain why its inclusion is appropriate.
IV. Provision of Polysilicon for LTAR
a. Background
In its original investigation, Commerce determined that the provision of polysilicon at LTAR in the PRC was a countervailable subsidy based on AFA. Prelim. I & D Memo at 33-35. Commerce determined that a benefit was being conferred based on the provision of polysilicon for LTAR, and thus sought to determine adequate remuneration in order to assess the appropriate countervailing duty.24 Id. at 34, I & D Memo at 31.
*1336It is Commerce's practice to determine remuneration by comparing the government price to a "market-determined price for the good or service resulting from actual transactions in the country in question," which it refers to as a tier-one metric. See 19 C.F.R. § 351.511(a)(2)(i).25 When no such price is available, or when there is reason to conclude that actual transactions are distorted, then Commerce resorts to its tier-two metric and determines adequate remuneration by comparing the government price to a world market price if it is reasonable to assume purchasers in the relevant country would be able to access that price. See 19 C.F.R. § 351.511(a)(2)(ii) ; Countervailing Duties: Final Rule 63 Fed. Reg. 65,348, 65,377 (Dep't Commerce Nov. 25, 2018) ("Preamble"). Because Commerce found that the PRC's involvement in the polysilicon market distorted the domestic prices, Commerce resorted to tier-two world market price data in order to determine the appropriate polysilicon subsidy benchmark.26 I & D Memo at 31; See 19 C.F.R. § 351.511(a)(2). Commerce did not explain how the GOC's market interference would result in import distortion.
Canadian Solar challenges this determination, arguing that because all of its polysilicon purchases were imported from market-economy suppliers outside the PRC, rather than domestically-purchased, they would not be distorted by the GOC's market interference. Canadian Solar Br. at 41-42. Canadian Solar thus contends that Commerce should have used Canadian Solar's purchases as a first-tier metric. Id. at 42. SolarWorld objects, arguing that although these purchases were imports, the "third-country suppliers would be forced to lower prices in order to compete with the artificially low prices in China." SolarWorld Resp. at 44. The Defendant responds that after finding that the Chinese market was distorted, it acted within its discretion in rejecting the imports as tier-one evidence and resorting to tier-two. Def. Br. at 37-38.
b. Discussion
As mentioned above, Commerce's determination that the polysilicon program was countervailable was based on the use of AFA. Prelim. I & D Memo at 18. As indicated previously, when a party refuses to cooperate and withholds requested information necessary to a determination, Commerce may need to rely on "facts otherwise available" in making a determination. 19 U.S.C. § 1677e(a)(2). Although under 19 U.S.C. § 1677e(b) Commerce may, under certain circumstances, draw an adverse inference as to facts affecting a cooperating party, it should attempt to avoid doing so when adequate information exists elsewhere on the record that would avoid this collateral effect. Archer Daniels, 917 F.Supp.2d 1331 at 1342 ; see also Fine Furniture (Shanghai) Ltd. v. United States, 865 F.Supp.2d 1254, 1262 (CIT 2012) (holding that "[w]here the respondents have placed evidence on the record consistent with the Department's regulations for calculating benchmarks ... Commerce would be expected to consider such evidence."). The court in Fine Furniture also found that an "alternative benchmark meeting such criteria" that did not "adversely affect a cooperative party ... would be superior *1337to one which does adversely affect a cooperative party." 865 F.Supp.2d at 1262.27 This policy helps ensure that non-government respondents continue to cooperate in administrative reviews and gives Commerce more opportunities to collect data that may best reflect a rate set in accordance with market principles.
Here, Canadian Solar did not take issue with Commerce's calculation of the world market price, but rather they argue that Commerce should not have resorted to this tier-two metric in the first place. Nothing in the record before the court indicates that Chinese manipulation of domestic transactions had any effect on arms-length import prices, and without such a determination, it is impossible to assess whether Commerce correctly resorted to a tier-two benchmark.
Rather than address Canadian Solar's submissions, which Canadian Solar claims show that all polysilicon purchases involved "arms-length import transactions with market economy suppliers," Commerce dismissed the evidence without consideration, stating simply that Commerce had already found actual transactions in China distorted. Canadian Solar Br. at 42; I & D Memo at 31. This is circular. Commerce's determination that the actual transactions were distorted presupposed the appropriateness of applying a subsidy rate to these cooperating parties derived from application of AFA in finding there was a subsidy program. Simply stating that the market was distorted fails to give cooperating parties a meaningful chance to rebut that they benefitted from any subsidy resulting in market distortion. Commerce should have considered Canadian Solar's proffered evidence and either accepted it as a tier-one metric or explained how these imports may have been distorted by GOC interference in the market. Only if the latter occurred would it be appropriate to resort to a tier-two metric. In sum, Commerce must first explain why Canadian Solar's submission, a potential tier-one benchmark, is not usable. Such an explanation should not be based entirely on the adverse inference used to determine that the GOC's influence in polysilicon distorted the market. Although SolarWorld's claim regarding import price depression in order for importers to compete in the PRC's national market may certainly be the case, without any such determination on the record, or even sufficient information about polysilicon's fungibility or the dynamics of the market, the court cannot accept it.
Accordingly, with respect to Commerce's decision to resort to a tier-two benchmark with regards to polysilicon, the court finds that Commerce's decision was unsupported by substantial evidence and remands this issue. On remand Commerce should either use Canadian Solar's import data as a tier-one benchmark or else explain how the GOC's purported interference with the polysilicon market would distort arms-length import transactions in a way that makes this data unreliable.
V. Inclusion of International Freight Charges
a. Use in calculating polysilicon, aluminum, and solar glass benchmarks
i. Background
In calculating the appropriate benchmark Commerce added international freight charges to the calculations for polysilicon, *1338aluminum, and solar glass. I & D Memo at 33. Canadian Solar argues that [ [ ] ] Commerce should not include international freight charges. Canadian Solar Br. at 32-34. Commerce rejected this argument in its final determination. See I & D Memo at 33. SolarWorld and the Government contend that the statute only requires that the benchmark be adjusted based on market conditions and does not concern the character of purchases by specific respondents during the period of review. SolarWorld Resp. at 34-36; Def. Br. at 38-39.
ii. Discussion
The statute at issue requires Commerce to determine a world market price based on "prevailing market conditions." 19 U.S.C. § 1677(5)(E).28 Commerce's regulation states that in assessing the adequacy of remuneration with a tier-one or tier-two metric,29 Commerce is required to "adjust the comparison price to reflect the price that a firm actually paid or would pay if it imported the product." 19 C.F.R. § 351.511(a)(2)(iv). The regulation explicitly includes transportation as an input in calculating prevailing market conditions and further assumes for its purposes that a firm imports the product at issue. Id.; see also Creswell Trading Co. v. Allegheny Foundry Co., 141 F.3d 1471, 1478 (Fed. Cir. 1998) (finding that a world market price must include the cost of shipping). This regulation does not require Commerce to base its decision on the purchasing decisions of the parties before it in a given proceeding, but rather requires that Commerce calculate the market conditions based on a party's hypothetically importing a given product. See Beijing Tianhai Industry Co., Ltd. v. United States, 52 F.Supp.3d 1351, 1374 (CIT 2015) (holding that a reference to a "firm" in 19 C.F.R. § 351.511(a)(2)(iv) does not refer to the respondent but to "a hypothetical firm located in the [country at issue].")
Canadian Solar mistakenly asserts that Commerce's calculations should be based on the specific circumstances of the Respondents.30 As the court has indicated, however, Commerce has determined that benchmark calculations are assessed based on a hypothetical importer making a market-price purchase, not the specific parties in a proceeding. See Beijing, 52 F.Supp.3d at 1374. Basing calculations on a hypothetical importer rather than the respondent in a given administrative review is not a plainly erroneous interpretation of 19 C.F.R. § 351.511(a)(2)(iv). See Auer v. Robbins, 519 U.S. 452, 461, 117 S.Ct. 905, 137 L.Ed.2d 79 (1997) (finding that an agency's interpretation of its own regulations controlling unless plainly erroneous or inconsistent with the regulation). Commerce found, freight charges would be paid by an importer of aluminum, solar glass, and polysilicon. See Prelim. I & D Memo at 34, 36, 37. Commerce was thus required to add international freight charges in order to calculate an accurate *1339benchmark for the products in question. See Essar Steel Ltd. v. United States, 678 F.3d 1268, 1274 (Fed. Cir. 2012) (finding that adding freight to the world market price was required by 19 U.S.C. § 1677(5)(E) as a prevailing market condition). Accordingly, Commerce's inclusion of international freight charges in its benchmark calculations is supported by substantial evidence and is in accordance with law.
b. Calculation of International Freight Charges for Benchmarking
i. Background
In determining the proper freight charge to be added for each benchmark calculation, Commerce averaged two data sets from Maersk and Xenata. I & D Memo at 33. The Xenata data was calculated based on actual prices while the Maersk data was based on price quotes. Id. at 34. Canadian Solar argues that because the Maersk Data is based on price quotes rather than finalized contracts, this data should be excluded from the calculation. Canadian Solar Br. at 35-36. Further, Canadian Solar contends that the use of such data is not in accordance with Commerce's own regulations. Id.
ii. Discussion
Under 19 C.F.R. § 351.511(a)(2), Commerce is required to calculate "the price that a firm actually paid or would pay if it imported" the product at issue. In making this calculation, this court has previously allowed Maersk's datasets based on price quotes for international freight charges rather than finalized contracts. See TMK IPSCO v. United States, 222 F.Supp.3d 1306, 1320-21 (CIT 2017) (finding that using Maersk's estimates was reasonable). Commerce's regulations require that "[w]here there is more than one commercially available world market price, [Commerce] will average such prices to the extent practicable." 19 C.F.R. 351.511(a)(2)(ii).
Commerce did not err in averaging the two datasets in determining the proper international freight benchmark. The regulation requires only that the price calculated reflect what a firm paid or would pay. This determination can properly be made on generally-available price quotes, so long as the source of those price quotes is a reputable source. Given Maersk's prominent position in the shipping market, Commerce properly considered the Maersk data to be a reliable world market price and averaged it with the Xenata data. As long as Commerce adequately justified why it chose to average the given datasets, as it did here, a set will not be excluded simply because it is based on a price estimate rather than completed contracts. See I & D Memo at 33-34. The benchmark calculation is sustained.
VI. Use of Value-Added Tax in calculating LTAR
i. Background
In its final determination, Commerce included value-added tax (VAT) in determining the appropriate benchmark prices. I & D Memo at 38-39. Canadian Solar claims that the addition of VAT is not an allowable adjustment under the regulation. Canadian Solar Br. at 36-38. They claim that VAT cannot be properly categorized as a delivery charge or import duty, but is rather an indirect tax, the inclusion of which inflates the apparent benefit received. Id. at 37 ; see 19 C.F.R. § 351.102(b)(28). In addition, Canadian Solar argues that because VAT is later recouped, it should not be included. Canadian Solar Br. at 38.
SolarWorld counters that failing to include the VAT would distort the benefit *1340calculated. SolarWorld Resp. at 39-40. The Government argues that the mandate of the statute is to calculate what an importer would pay at the time of import-which includes VAT-and that in calculating this figure, Commerce does not account for potential reimbursement after import. Def. Br. at 42-43.
ii. Discussion
Under the relevant regulation, Commerce adjusts benchmark prices to include delivery charges and import duties that an importer would pay in order to arrive at the "delivered price." See 19 C.F.R. § 351.511(a)(2)(iv). The court has previously found the inclusion of VAT in this calculation appropriate. See Beijing, 52 F.Supp.3d at 1372-1375 (holding that the inclusion of VAT in a tier-two benchmark was correct). The relevant statute says that when determining the adequacy of remuneration Commerce should take into account market conditions including "price, quality, availability, marketability, transportation, and other conditions of purchase or sale." 19 U.S.C. § 1677(5)(E).
Here, Canadian Solar's contention that the inclusion of VAT is not allowable under the regulation fails. The relevant regulation states that the benchmark should be adjusted to reflect the "delivered price" meaning the "price that a firm actually paid or would pay if it imported the product." 19 C.F.R. § 351.511(a)(2)(iv).31 Commerce determined that VAT would be paid in this situation and Canadian Solar has offered no evidence showing that to be incorrect. It is immaterial that the regulations may classify VAT as an indirect tax; the regulation does not say that the delivered price only includes delivery charges and import duties, just that it does include these figures. VAT is properly classified as a market condition under the statute, and is thus may be included in calculating the benchmark for adequate remuneration. See 19 U.S.C. § 1677(5)(E). Therefore, Commerce's addition of VAT to the benchmark was supported by substantial evidence and otherwise in accordance with law.
VII. Electricity Subsidy
a. Background
After finding that the GOC failed to fully cooperate in responding to questions regarding the alleged provision of electricity for LTAR, Commerce applied AFA to determine that there was a specific subsidy and subsequently to calculate the benefit conferred.32 I & D Memo at 40-41; Prelim. I & D Memo at 27-28. When a party fails to respond to "the best of its ability" with reasonable requests for information, Commerce may apply AFA in order to prevent that party from benefitting *1341from its non-cooperation and to encourage future participation. See Nippon Steel Corp., 337 F.3d 1373 at 1382 ; Fine Furniture (Shanghai) Ltd. v. United States, 748 F.3d 1365, 1373 (Fed. Cir. 2014). Commerce relied on AFA in finding that there was a subsidy, that the subsidy was specific, and in calculating the benchmark. I & D Memo 40-42. Commerce selected the highest rate schedule on record for each reported category and used those rates to set a benchmark. See I & D Memo at 41; Prelim. I & D Memo at 28. Commerce states that this calculation is reasonable because without information on how electrical rates are determined, it is plausible that a respondent would bear the highest rate regardless of a plant's location. I & D Memo at 41; see also Def. Br. at 44, 46.
Canadian Solar argues that Commerce failed to adequately determine that the electricity subsidy was specific. Canadian Solar Br. at 39-41. According to Canadian Solar, it could qualify only as a domestic subsidy under 19 U.S.C. § 1677(5A)(D) because electricity cannot be imported or exported.33 Id. at 40. In its reply brief, Canadian Solar further argues that Commerce failed to make a specificity determination at all and simply concluded that the provision of electricity was countervailable without an explanation. Canadian Solar Reply at 16-19. Finally, Canadian Solar and Trina disagree with Commerce's application of the highest rate from six different provinces in China given that a particular factory cannot exist in more than one location. Canadian Solar Br. at 41; Trina Br. at 26. Canadian Solar argues that Commerce failed to cite anything supporting the notion that the subsidy was geographically specific. Canadian Solar Br. at 40.34 Trina further argues that the GOC supplied enough information to prove that electricity pricing varies by province. Trina Br. at 24. Trina contends that the record proves that plants at issue would not have been subjected to the electricity rates of provinces in which they were not located. Id. Trina presents various ways in which Commerce could have calculated the rate while still applying AFA. Trina Br. at 27.
i. Discussion
The court has upheld the application of AFA in determining that a given subsidy was specific when a party has failed to cooperate. See RZBC Grp. Shareholding Co. Ltd. v. United States, 100 F.Supp.3d 1288, 1296-97 (CIT 2015) (sustaining an application of AFA to determine that a calcium carbonate subsidy was specific). In previous cases involving electricity subsides from China, the court has found an adverse inference appropriate when a party did not provide enough information to determine whether the rate was set according to market principles. See Fine Furniture, 865 F.Supp.2d at 1262-63.
In this instance, Commerce characterized the GOC's refusal "to answer questions related to regional electrical differences, including differences between industries" as preventing it from determining from direct evidence whether the subsidy was specific. I & D Memo at 41. It decided solely on this failure to answer some questions that the subsidy was specific. Commerce, however, failed to explain the particular facts as to which it *1342was it was drawing an adverse inference and how that analysis subsequently results in a finding of specificity under one of the criteria listed in 19 U.S.C. § 1677(5A).
AFA is not a magic phrase that permits Commerce to skip an analysis of the record. Here, the Government states that the GOC failed to adequately respond with information necessary for Commerce to understand whether the electricity prices were set in accordance with market principles. In response to Commerce's questions, the PRC supplied numerous documents detailing the provision of electricity in China. See Conf. Joint App'x 1 at 35-520. Rather than explain what information was missing, or how these submissions were deficient, Commerce makes the conclusory statement that the GOC failed to comply and thus Commerce can rightly determine that there is a subsidy, that it confers a benefit, and that it is specific such that the subsidy is countervailable. Prelim. I & D Memo at 27-28. Without combing through the submissions and guessing as to why Commerce found these submissions inadequate, the court is unable to ascertain how Commerce made its decision. The record is simply unclear.
Although Commerce does mention that the electricity program was found countervailable in an earlier administrative review, that earlier decision based on a different record does not clarify how Commerce found for this review period that the provision of electricity continued to provide a financial contribution, whether the subsidy conferred a benefit, and, most relevant here, whether this provision was specific within the meaning of the statute. See 19 U.S.C. § 1677(5)(B) ; see also Albemarle Corp. & Subsidiaries v. United States, 821 F.3d 1345, 1356 (2016) (stating that Commerce must "base its decisions on the record of the administrative proceeding before it in each review"). Although the Court of Appeals for the Federal Circuit's decision in Albemarle had to do with Commerce's duty to recalculate dumping margins in each administrative review, except in certain limited circumstances, the principle underlying that decision holds true here. The purpose of administrative reviews is to reassess earlier determinations in the light of data relevant to the period of review. See id. at 1356-57. If Commerce simply relies on findings from an earlier administrative review in reaching a decision in a current review, it abdicates its responsibility and undermines the central purpose of periodic reviews.
This is not to say that Commerce cannot reference the reasoning of a previous administrative review. But Commerce must give respondents a meaningful opportunity to dispute earlier findings and offer evidence of changes. If respondents fail to do so, then Commerce might state that no new evidence merits a reconsideration of a decision made in a previous administrative review, but Commerce must at the very least explain why a decision made in an earlier review should control.
Simply stating that the GOC did not fully comply is insufficient, Commerce must actually engage in an analysis of the information on the record and explain how adverse inferences lead to the conclusion that the provision of electricity in China is a countervailable subsidy. Otherwise stated, Commerce must connect the dots: how does the GOC's partial response-or failure to respond fully-reasonably lead to a finding of a specific subsidy even with use of AFA?
If, on remand, Commerce properly concludes that the provision of electricity in the PRC amounts to a countervailable subsidy, Commerce's benchmark determination based on record evidence with appropriate *1343adverse inferences is consistent with its regulations for calculating benchmarks. See 19 C.F.R. § 351.511(a)(2). The GOC refused to provide certain details regarding variation of provincial electricity rates and whether these rates were calculated based on market principles. See I & D Memo at 41. Accordingly, Commerce can apply an adverse inference to the GOC's electricity rate submissions and select the highest rates for each electrical category and use those to set a benchmark. See 19 U.S.C. § 1677e(b) ; 19 C.F.R. § 351.511(a)(2)(iii) ; Fine Furniture, 865 F.Supp.2d at 1260-1262 (upholding Commerce's decision to set the benchmark rate for electricity equal to the highest rate reported in the provincial price schedules submitted by the GOC when Commerce was unable to determine whether the prices were set in accordance with market principles). Commerce's goal in setting a benchmark rate is to best approximate the market rate of electricity, not to choose the rate respondents were most likely to pay in an electricity market Commerce argues is tainted by the GOC's interference.
Finally, although Trina provided potential alternative modes of calculating the benchmark, it has not shown that these calculation methods result in a better estimate of the market rate for electricity. It is not this court's place to substitute its judgment for that of Commerce by selecting a different method of calculation where Commerce has acted within its lawful discretion and made a reasonable decision. See Inland Steel Indus., Inc. v. United States, 188 F.3d 1349, 1360-61 (Fed. Cir. 1999). In sum, assuming a countervailable subsidy exists, Commerce acted in accordance with the law in using the highest of all provincial rates on the record to calculate the benchmark.
Accordingly, the issue regarding the provision of electricity is remanded for Commerce to explain how it arrived at its conclusion that such provision was a countervailable subsidy. Commerce should cite specific information on the record, noting any allowable adverse inferences, in making its decision.
VIII. Golden Sun Demonstration Program
i. Background
During the original investigation, Commerce found the Golden Sun Demonstration Program ("GSDP") to be countervailable. See Crystalline Silicon Photovoltaic Cells, Whether or Not Assembled into Modules, From the People's Republic of China: Final Affirmative Countervailing Duty Determination and Final Affirmative Critical Circumstances Determination, 77 Fed. Reg. 63,788 (October 17, 2012) and accompanying issues and decision memorandum at 10-11 ("I & D Memo 2012"). The GSDP was created in 2009 to assist in the construction of photovoltaic electricity-generation projects. See Prelim. I & D at 43. In its final determination, Commerce found no new information warranting a reexamination of the program and concluded that the subsidy was untied and attributable to Canadian Solar's total sales. I & D Memo at 47. The Government argues that it must only look at the grant when it was bestowed and need not inquire as to how the grant was used specifically. Def. Br. at 48-50.
Canadian Solar contends that GSDP was meant to subsidize the generation of electricity and not the production of solar cells. Canadian Solar Br. at 44. Although Canadian Solar admits that it received GSDP funds, it asserts that Commerce has mischaracterized the program's purpose as providing assistance in the production of solar cells. Id. at 43-45. Canadian Solar *1344contends that at the time of bestowal, the subsidy could only be properly attributed to power generation operations and thus Commerce improperly assessed countervailing duties on the production of solar cells. Canadian Solar Br. at 43-45. SolarWorld contends that it is unclear from the record whether the grant was intended for electricity production or solar cell production. SolarWorld Resp. at 45-46.
ii. Discussion
In the 2012 Final Determination from the initial investigation, Commerce found that the Golden Sun program subsidized "solar-powered projects." I & D Memo 2012 at 12 (October 17, 2012). Based on submissions by the GOC, in its preliminary determination Commerce found that the program supported:
(1) The use of large-scale mining, commercial enterprises, and public welfare institutions to construct the user's side of the electrical grid for photovoltaic power generation demonstration projects; (2) increasing the power supply capacity in remote locations; and (3) construction of large-scale grid-connected photovoltaic power generation demonstration projects in solar energy rich regions." Crystalline Silicon Photovoltaic Cells, Whether or Not Assembled Into Modules, From the People's Republic of China: Preliminary Affirmative Countervailing Duty Determination C-570-980 POI 01/01/2010-12/31/2010 at 15-16 (Mar. 26, 2012) ("Prelim. I & D Memo 2012")
Based on this information, Commerce determined that grants under this program constitute a subsidy of enterprises "involved in the construction of solar-powered projects." Prelim. I & D Memo 2012 at 16.
Commerce's regulations mandate that "[i]f a subsidy is tied to the production or sale of a particular product; the Secretary will attribute the subsidy only to that product." 19 C.F.R. § 351.525(b)(5)(i). When a subsidy is not tied to a specific product, however, it is Commerce's practice to attribute the benefits of a subsidy based on the stated purpose at the time of bestowal,35 as untied subsidies are attributed to all products sold because they benefit all production. Preamble, 63 Fed. Reg. at 65,400 ("the current benefit of an untied subsidy will be attributed to the firm's total sales.")
Canadian Solar argues that the language of the program description is not specific to the production of solar cells, but is for energy production broadly. While this may be true, Commerce reasonably understands this program to include the subsidization of the production of solar cells, despite the inference needed to reach this conclusion. It is reasonable to assume that creating photovoltaic power generation necessitates the production of solar cells as a component of this endeavor. Although Canadian Solar may not use the funds received through this program specifically in the production of solar cells, Commerce need only look at the purpose of the subsidy at the time it is bestowed and not exactly how it is used by companies. Therefore, Commerce's decision is supported by substantial evidence and in accordance with law.
CONCLUSION
For the foregoing reasons, the court remands Commerce's challenged determinations as regards to its determination on the Export Buyer's Credit Program, the inclusion of Comtrade data in calculating *1345the world market rate for aluminum extrusions and solar glass, Commerce's decision to revert to a tier-two benchmark in determining the price for polysilicon without considering Respondent's proffered evidence, and the finding that provision of electricity constitutes a specific and thus countervailable subsidy. All other determinations are sustained. The court remands for proceedings consistent with this opinion. Remand results should be filed by January 29, 2019. Objections are due February 28, 2019 and Responses to Objections are due March 15, 2019.