Canadian Solar Int'l Ltd. v. United States (378 F. Supp. 3d 1292, 2019)

Kelly, Judge:

Before the court are several motions for judgment on the agency record challenging various aspects of the U.S. Department of Commerce's ("Commerce" or "the Department") determination in the third administrative review of the antidumping duty ("ADD") order on crystalline silicon photovoltaic products, whether or not assembled into modules, from the People's Republic of China ("the PRC"). See Crystalline Silicon Photovoltaic Cells, Whether or Not Assembled Into Modules, From the People's Republic of China, 82 Fed. Reg. 29,033 (Dep't Commerce June 27, 2017) (final results of [ADD] administrative review and final determination of no shipments; 2014-2015) ("Final Results") and accompanying Issues and Decision Mem. for the Final Results of the 2014-2015 [ADD] Administrative Review of Crystalline Silicon Photovoltaic Cells, Whether or Not Assembled into Modules, From [the PRC], A-570-979, (June 20, 2017), ECF No. 44-5 ("Final Decision Memo").

For the reasons that follow, the court sustains Commerce's selection of surrogate values for aluminum frames, nitrogen, polysilicon ingots and blocks, and financial ratios. The court also sustains Commerce's decision to include import data with reported zero quantities in its calculation of surrogate values and its decision to exclude Trina U.S.'s debt restructuring income as an offset to its indirect selling expenses. The court remands Commerce's selection of surrogate value for module glass, Commerce's application of an adverse inference in calculating Canadian Solar's dumping rate, and Commerce's rejection of Ningbo Qixin Solar Electrical Appliance Co., Ltd.'s ("Qixin") separate rate application.

BACKGROUND

On February 9, 2016, Commerce initiated the third administrative review of the ADD order on crystalline silicon photovoltaic cells, whether or not assembled into modules, from the PRC, for which the period of review would be December 1, 2014 through November 30, 2015. See Initiation of Antidumping and Countervailing Duty Administrative Reviews, 81 Fed. Reg. 6,832, 6,835 (Dep't Commerce, February 9, 2016). On March 28, 2016, after determining that it would not be practicable to examine individually each company for which a review was initiated, Commerce selected Canadian Solar International Limited and the collapsed entity of Trina Solar, comprised of Changzhou Trina Energy Co., Ltd. and Trina Solar (Changzhou) Science and Technology Co., Ltd. as mandatory respondents. See Respondent Selection Mem. at 6, PD 155, CD 104, bar code 3452853-01 (Mar. 28, 2016).

On December 22, 2016, Commerce published the preliminary results of the third administrative review. See Crystalline Silicon Photovoltaic Cells, Whether or Not Assembled Into Modules, From [the PRC], 81 Fed. Reg. 93,888 (Dep't Commerce Dec. 22, 2016) (preliminary results of [ADD] administrative review and preliminary determination of no shipments; 2014-2015) ("Preliminary Results") and accompanying Decision Mem. for Prelim. Results of the 2014-2015 [ADD] Administrative Review of Crystalline Silicon Photovoltaic Cells, Whether or not Assembled into Modules, From [the PRC], A-570-979, PD 499, bar code 3530538-01 (Dec. 16, 2016) ("Prelim. Decision Memo").

On June 27, 2017, Commerce published the final determination. See Final Results, 82 Fed. Reg. at 29,033. Commerce selected Thailand as the primary surrogate country for valuing the mandatory respondents' factors of production ("FOP"), see generally Final Decision Memo, and adopted surrogate values for, inter alia, semi-finished polysilicon ingots and blocks, aluminum frames, module glass, nitrogen, and overhead and financial expenses. Final Decision Memo at 21-22, 35-38, 45-50, 52-55, 66-71. Commerce applied partial AFA in calculating Canadian Solar International Limited's antidumping margin due to the failure of unaffiliated solar cell and solar module suppliers to provide FOP information. Final Decision Memo at 15-18. Commerce excluded Trina U.S.'s debt restructuring income from its calculation of Trina's U.S. indirect selling expense ratio. Id. at 84-85. Commerce included in the average unit surrogate value calculations import data with reported quantities of zero, finding "no basis to conclude that the zero quantity import data ... are errors or that these zero quantity imports result in unreliable and distortive [surrogate values]." Final Decision Memo at 86-87. Finally, Commerce rejected Qixin's separate rate application and assigned it the China-wide rate. Final Decision Memo at 90-92.

On July 7, 2017, Plaintiffs Canadian Solar International Limited; Canadian Solar (USA), Inc.; Canadian Solar Manufacturing (Changshu), Inc.; Canadian Solar Manufacturing (Luoyang), Inc.; CSI Cells Co., Ltd.; CSI-GCL Solar Manufacturing (YanCheng) Co., Ltd.; and CSI Solar Power (China) Inc. (collectively, "Canadian Solar") commenced this action pursuant to section 516A of the Tariff Act of 1930, as amended, 19 U.S.C. § 1516a(a)(2)(B)(iii) (2012). Summons, July 7, 2017, ECF No. 1 ; Compl., July 7, 2017, ECF No. 8. Canadian Solar moves for judgment on the agency record, challenging three aspects of the Final Results. Specifically, Canadian Solar challenges: 1) Commerce's application of partial AFA with respect to missing supplier information; 2) Commerce's use of import data under Thai Harmonized Tariff Schedule ("HTS") 7007.19.90000 to value Canadian Solar's module glass consumption; and 3) Commerce's use of import data under Thai HTS 2804.30.00000 to value its nitrogen consumption. See Mem. Points & Authorities Supp. Mot. J. Agency R. at 10-41, Mar. 7, 2018, ECF No. 54-1 ("Canadian Solar's Br.").

This action was consolidated with actions brought by Qixin, Shanghai BYD Co., Ltd. ("BYD"), Changzhou Trina Solar Energy Co., Ltd. et al. ("Trina"), SolarWorld Americas, Inc. ("SolarWorld"), and Sunpreme Inc. See Order, Sept. 26, 2017, ECF No. 41. Consolidated Plaintiffs and Plaintiff-Intervenors filed motions for judgment on the agency record, Mot. J. Agency R., Mar. 7, 2018, ECF No. 52 ; Pls.' R. 56.2 Mot. J. Agency R., Mar. 7, 2018, ECF No. 55 ; Mot. J. Agency R., Mar. 7, 2018, ECF No. 56 ; [SolarWorld's] Mot. J. Agency R., Mar. 7, 2018, ECF No. 57 ; Mot. J. Agency R., Mar. 7, 2018, ECF No. 60, each challenging various aspects of Commerce's Final Results. See Mem. Supp. Mot. J. Agency R. Submitted by Pl. Pursuant to R. 56.2 R. U.S. Ct. Int'l Trade, Mar. 7, 2018, ECF No. 52-1 ("Qixin's Br.") ; Mem. Supp. Mot. [Trina] J. Agency R., Mar. 7, 2018, ECF 55-1 ("Trina's Br."); [SolarWorld's] Mem. Supp. R. 56.2 Mot. J. Agency R., Mar. 8, 2018, ECF No. 63 ("SolarWorld's Br."). Specifically, Qixin challenges Commerce's denial of its separate rate application in the Final Results after omitting any reference to Qixin in the Preliminary Results. Qixin's Br. at 6-15. Trina challenges: 1) Commerce's use of Thai import data to value nitrogen; 2) Commerce's use of Thai import data to value module glass; 3) Commerce's decision to include in its calculation of surrogate values import data with no corresponding quantities; and 4) Commerce's exclusion of Trina U.S.'s debt restructuring income in its calculation of Trina's U.S. indirect selling expense ratio. Trina's Br. at 4-19. SolarWorld challenges: 1) Commerce's selection of Thai HTS 7604.29.90001 data to value the respondents' aluminum frames; 2) Commerce's surrogate value selection for module glass; 3) Commerce's selection of Styromatic's 2015 financial statements as best available information to calculate the respondents' overheard, selling, general and administrative expenses, and profit; and 4) Commerce's surrogate value selection for respondents' semi-finished polysilicon ingots and blocks. SolarWorld's Br. at 10-32.

JURISDICTION AND STANDARD OF REVIEW

The court has jurisdiction pursuant to 19 U.S.C. § 1516a(a)(2)(B)(iii) and 28 U.S.C. § 1581(c) (2012), which grant the court authority to review actions contesting the final determination in an administrative review of an [ADD] order. "The court shall hold unlawful any determination, finding, or conclusion found ... to be unsupported by substantial evidence on the record, or otherwise not in accordance with law." 19 U.S.C. § 1516a(b)(1)(B)(i).

DISCUSSION

Plaintiffs and Consolidated Plaintiffs challenge a total of five of Commerce's surrogate value determinations in the Final Results and raise four additional challenges. The court first addresses the arguments that Commerce's surrogate value selections for module glass, aluminum frames, nitrogen, semi-finished polysilicon ingots and blocks, and overhead, selling, general and administrative expenses, and profit are contrary to law and/or unsupported by substantial evidence. The court then addresses the arguments regarding the application of partial AFA, the decision to include import data with reported zero quantities in the calculation of surrogate values, the exclusion of Trina's debt restructuring income in its calculation of Trina's U.S. indirect selling expense ratio, and the rejection of Qixin's separate rate application in the Final Results.

I. Surrogate Value Selection

Plaintiff and Consolidated Plaintiffs challenge Commerce's surrogate value selection for module glass, aluminum frames, nitrogen, semi-finished polysilicon ingots and blocks, and overhead, selling, general and administrative expenses, and profit as unsupported by substantial evidence. The court addresses each of these challenges in turn.

A. Module Glass

Trina argues that Commerce's selection of tempered glass under Thai HTS 7007.19.90000 over float glass under Thai HTS 7005.29.90001 to value Trina's module glass was unsupported by substantial evidence because the float glass classification specified the thickness of module glass consumed by Trina. See Trina's Br. at 2, 12-14. SolarWorld argues that Commerce's selection of Thai import data for tempered glass over laminated glass was unsupported by substantial evidence because tempered glass fails to account for the additional surface treatments of module glass. See SolarWorld's Br. at 3, 19-21. Canadian Solar argues that Commerce's selection of the Thai data for tempered glass is unsupported by substantial evidence because the Thai data used by Commerce was aberrational. See Canadian Solar's Br. at 3, 19-31. Defendant responds that Commerce's decision is reasonable, as the Thai import data for tempered glass satisfied all of Commerce's surrogate value criteria and was not aberrational. See Def.'s Resp. Opp'n Pl.'s Mot. J. Agency R. at 6, 12-21, July 30, 2018, ECF No. 71 ("Def.'s Br."). For the reasons that follow, Commerce's selection of tempered glass over float glass and laminated glass is reasonable. Commerce's decision to use the Thai import data for tempered glass, however, is unsupported by substantial evidence and is remanded to Commerce for further consideration.

Where the subject merchandise is exported from a nonmarket economy country, Commerce calculates normal value based on FOPs. 19 U.S.C. § 1677b(c)(1). Commerce uses "the best available information" to value the FOPs, id., and has discretion to determine what constitutes the best available information, as this term is not defined by statute. QVD Food Co. v. United States, 658 F.3d 1318, 1323 (Fed. Cir. 2011). However, Commerce must ground its determination in the objective of the statute: to calculate accurate dumping margins. See Rhone Poulenc, Inc. v. United States, 899 F.2d 1185, 1191 (Fed. Cir. 1990) ; see also Parkdale Int'l v. United States, 475 F.3d 1375, 1380 (Fed. Cir. 2007). Commerce generally selects surrogate values that are publicly available, are product specific, reflect a broad market average, and are contemporaneous with the period of review. Qingdao Sea-Line Trading Co. v. United States, 766 F.3d 1378, 1386 (Fed. Cir. 2014) ; see also Import Admin., U.S. Dep't Commerce, Non-Market Economy Surrogate Country Selection Process, Policy Bulletin 04.1 (2004), available at http://enforcement.trade.gov/policy/bull04-1.html (last visited Apr. 11, 2019) ("Policy Bulletin 04.1"). Commerce's practice is to avoid using aberrational values as surrogate values. See generally Antidumping Duties; Countervailing Duties, 62 Fed. Reg. 27,296, 27,366 (Dep't Commerce May 19, 1997).

In this review, Canadian Solar reported using solar module glass, and Trina reported using coated glass and tempered glass (collectively, "module glass"), as FOPs. See Final Decision Memo at 45. The record contained Thai import data for tempered glass (HTS 7007.19.90000), float glass (HTS 7005.29.90001), and laminated glass (HTS 7007.29.90). Id. Here, Commerce reasonably selected tempered glass as the best available information to value the respondents' module glass inputs because record information and the respondents' descriptions indicated the module glass consumed by the respondents was tempered. See Final Decision Memo at 45-46. Commerce reasonably determined the laminated glass subheading was not the best available information because laminated glass is composed of "multiple layers of glass and plastic" and there was no evidence that such properties are present in the respondents' module glass inputs. See Final Decision Memo at 46-47.

SolarWorld maintains that Commerce's choice of tempered glass as a surrogate value substantially undervalued the module glass input because only laminated glass captures the additional costs associated with the surface treatments used on the respondents' module glass to increase its strength, safety, and durability. See SolarWorld's Br. at 19-21. SolarWorld's argument fails, however, as SolarWorld does not provide any evidence establishing that this surface treatment is comparable to the layering of laminated glass. See Final Decision Memo at 47.

Commerce also reasonably selected tempered glass over float glass to value Trina's module glass input because, although the float glass subheading more closely aligned with the thickness of Trina's input, the tempered glass subheading's lack of specification as to thickness indicated that it covered all thicknesses, including that consumed by Trina. Final Decision Memo at 46. Trina maintains Commerce failed to weigh the comparative importance of thickness against other factors in favor of selecting tempered glass. See Trina's Br. at 13. However, it is clear from Commerce's explanation that Commerce gave significant weight to the description by the parties that their module glass was tempered. See Final Decision Memo at 45-46. It is also clear that Commerce gave weight to the fact that the tempered glass subheading did not exclude the products with a thickness matching Trina's input. See Final Decision Memo at 45-46. The court will not reweigh the evidence, and Trina's argument thus fails.

However, while Commerce's choice of the tempered glass subheading to value Trina's and Canadian Solar's module glass is supported by substantial evidence, Commerce failed to provide a reasonable explanation as to why the Thai import data for tempered glass was not distorted by a small quantity of unusually costly imports from Hong Kong. Hong Kong imports accounted for only 0.4% of the total volume of Thai import data but constituted 60.2% of the total value of Thai imports during the period of review. See Final Decision Memo at 49; see also Thai Tempered Glass [attached as Ex. FR-1 to Canadian Solar's Oct. 31, 2016 Submission], PD 463-64, bar code 3518110-01 (Oct. 31, 2016). The average unit value ("AUV") for tempered glass imported into Thailand was $ 2.79 with the Hong Kong data, whereas excluding the Hong Kong data results in an AUV of only $ 1.11. Id.

The court addressed this same issue in SolarWorld II in the proceedings concerning the second administrative review of the ADD order covering crystalline silicon photovoltaic cells. See Canadian Solar's Br. at 22-24; see also Canadian Solar's Reply Br. at 5; see also SolarWorld Americas, Inc. v. United States, 42 CIT ----, 320 F.Supp.3d 1341 (2018) (" SolarWorld II"). In SolarWorld II the Court ordered Commerce to further explain on remand why Commerce's selection of Thai import data for tempered glass was reasonable in light of record evidence that imports from Hong Kong made up only 1.6% of the total volume but accounted for 75% of the total value. See SolarWorld II, 42 CIT at ----, 320 F.Supp.3d at 1354-55. The court held that it was not clear that Commerce had established a consistent practice of only assessing whether a surrogate value was aberrant in the aggregate, and that regardless, even if it was a consistent practice, Commerce had failed to explain how that practice was reasonable in light of the concerns raised by the Hong Kong imports as to the accuracy of the Thai data as a whole. See id., 320 F.Supp.3d at 1352-56 (noting Commerce's reliance on Issues and Decision Mem. For the Final Results of the 2012-2013 Admin. Review of Multilayered Wood Flooring from [the PRC] at 41-43, A-570-970 (July 8, 2015), available at https://enforcement.trade.gov/frn/summary/prc/2015-17368-1.pdf (last visited Apr. 11, 2019) ("Wood Flooring") ). Commerce's remand redetermination abandoning the use of the Thai data for tempered glass was later sustained. See SolarWorld Americas, Inc. v. United States, 42 CIT ----, 355 F.Supp.3d 1306 (2018). The reasoning in SolarWorld II applies with equal force to the present proceedings.

Commerce argues that the "relevant test is to determine whether the AUV in aggregate is aberrational," as otherwise "parties would advocate the manipulation of data by removing one or more lines they find objectionable." Final Decision Memo at 47. As in SolarWorld II, although this approach may be reasonable in other cases, it is not reasonable on this record without further explanation. See SolarWorld II, 42 CIT at ----, 320 F.Supp.3d at 1355.

Commerce unreasonably found that the Hong Kong imports were not low in quantity because Hong Kong ranged fourth in terms of quantity out of 22 countries from which Thailand imported tempered glass. Final Decision Memo at 49. Hong Kong imports account for only 0.4% of the total volume of imports of tempered glass into Thailand. The mere fact that there are other countries which represent an even smaller fraction of the imports into Thailand does not impact whether the imports from Hong Kong are of low volume.

Commerce also unreasonably concluded that the Hong Kong imports were not aberrational through reliance on benchmarks based on imports of negligible quantity from Denmark, France, Switzerland, and Mexico. Final Decision Memo at 49. These four other countries exported tempered glass to Thailand at prices equivalent to, or greater than, the Hong Kong imports. Id. The total import quantities from Denmark and France in the period of review were only 12 kilograms or less, while an import quantity of zero was recorded for Mexico and Switzerland. Id. These amounts are a negligible component of the total of approximately 2.26 million kilograms of tempered glass imported into Thailand in the period of review. See Thailand Import Data [attached as Ex. 2 to SolarWorld's Initial Surrogate Value Comments] at 2, PD 365-382, CD 424-436, bar code 3489132-01 (July 19, 2019). Substantial evidence requires Commerce to address evidence that supports its finding as well as that which "fairly detracts from its weight." Universal Camera Corp. v. NLRB, 340 U.S. 474, 477-78, 71 S.Ct. 456, 95 L.Ed. 456 (1951) (quoting Consol. Edison Co. v. NLRB, 305 U.S. 197, 229, 59 S.Ct. 206, 83 L.Ed. 126 (1938) ). Commerce notes the negligible volume of imports from these countries but fails to address the potentially distortive effect of the benchmarks' minuscule import quantities. Final Decision Memo at 49. It is unreasonable to rely on benchmarks that are exclusively based on negligible import quantities without addressing the impact this negligible volume has on the reliability of the benchmarks. See Blue Field (Sichuan) Food Industrial Co., Ltd. v. United States, 37 CIT ----, ----, 949 F.Supp.2d 1311, 1327-28 (2013) (holding as unsupported by substantial evidence a finding by Commerce that a surrogate value was not aberrational because, inter alia, that finding was made in reliance on benchmarks based on "miniscule import volumes"). Commerce's finding that tempered glass imports from Hong Kong are not aberrational through comparison with improperly vetted benchmarks based on negligible quantities is thus unsupported by substantial evidence.

Commerce's selection of the Thai data is remanded for further explanation of why its selection is reasonable in light of the fact that imports from Hong Kong only accounted for 0.4 % of the total volume of Thai imports, but constituted 60.2 % of the average import value during the period of review. The AUV of imports of tempered glass from Hong Kong into Thailand ($ 413.10 per kilogram) is nearly 150 times greater than the AUV of all tempered glass imports into Thailand (including imports from Hong Kong) ($ 2.79 per kilogram). See Canadian Solar's Reply at 3. Commerce's current explanation for not disaggregating data is that it simply does not do so as a matter of policy. It justifies its policy on the grounds of administrative burden and to avoid potentially distortive manipulation of import data. Final Decision Memo at 47, 54 (citing Polyethylene Terephthalate Film, Sheet, and Strip from [the PRC]: Issues and Decision Mem. for the Final Results of the 2011-2012 Admin. Review at 12, A-570-924, (June 5, 2013), available at https://enforcement.trade.gov/frn/summary/prc/2013-13985-1.pdf (last visited Apr. 11, 2019); see also SolarWorld II 42 CIT at ----, 320 F.Supp.3d at 1354-1355. Commerce's justification may be sufficient for most cases. At some point, however, input data may diverge so significantly from other input data that it renders the data set, as a whole, unreliable. Ideally, Commerce, not the court, should identify where that point lies. For module glass, the distortive input data has reached the point where the court cannot say that Commerce selection is reasonable.

B. Aluminum Frames

SolarWorld challenges Commerce's valuation of the respondents' aluminum frames using Thai HTS 7604.29.90001, which covers non-hollow aluminum profiles. See SolarWorld's Br. at 11-19; see also Final Decision Memo at 35-38. SolarWorld contends that Commerce's selection is contrary to law and not supported by substantial evidence, arguing that the selection is not in accordance with the HTS, fails to account for the additional manufacturing processes that the frames in question undergo, and runs contrary to CBP rulings on the merchandise in question. SolarWorld's Br. at 11-19. Defendant responds that Commerce's decision is reasonable because Thai HTS 7604.29.90001 represents the most specific proposed surrogate value on the record. Def.'s Br. at 23. For the reasons that follow, Commerce's selection of import data under HTS 7604.29.90001 is supported by substantial evidence.

As discussed above, where the subject merchandise is exported from a nonmarket economy country, Commerce calculates normal value based on FOPs. 19 U.S.C. § 1677b(c)(1). Commerce selects a surrogate value by which it values the FOPs and makes that selection "based on the best available information regarding the values of such factors in a market economy country or countries." Id. Although Commerce has broad discretion in deciding what constitutes the best available information, see QVD Food Co. v. United States, 658 F.3d 1318, 1323 (Fed. Cir. 2011) (noting the absence of a definition for "best available information" in the ADD statute), it must ground its selection of the best available information in the overall purpose of the statute, which is to calculate accurate dumping margins. See Rhone Poulenc, Inc., 899 F.2d at 1191 ; see also Parkdale Int'l. v. United States, 475 F.3d 1375, 1380 (Fed. Cir. 2007). Commerce considers the best available information to be (1) specific to the input; (2) tax and import duty exclusive; (3) contemporaneous with the period of review; (4)

representative of a broad market average; and (5) publicly available. See Policy Bulletin 04.1.

Here, Commerce reasonably concluded that import data under HTS 7604.29.90001, which covers non-hollow aluminum profiles, constitutes the best available information to value Trina's aluminum frames. See Final Decision Memo at 35-38. Commerce noted that Trina sufficiently demonstrated that the aluminum frames in question are non-hollow, aluminum profiles, and that Commerce found no evidence on the record to contradict this description. Final Decision Memo at 35. Moreover, as in the underlying investigation, Commerce emphasized that HTS 7604 presents the best available information to value the inputs in question because it covers alloyed aluminum profiles, whereas HTS 7616.99.90-suggested by SolarWorld-comprises an "other" subheading that includes dissimilar products. Id. Commerce reasoned that HTS 7616 applied to completely different products than the aluminum frames at issue because this heading includes products like "nails, tacks, staples, screws, bolts, nuts, screw hooks, rivets, cotters, cotter pins, washers, knitting needles, bodkins, crochet hooks, embroidery stilettos, safety pins, other pins and chains, and cloth, grill and netting of aluminum wire." Final Decision Memo at 37-38. Based on the inclusion of such unrelated items, Commerce reasonably concluded that HTS 7616.99.9909 does not constitute the best available information. Additionally, Commerce reasonably determined that HTS 7604's descriptions, such as "[a]luminum bars, rods and profiles," indicate that the subheading used-HTS 7604.29-includes non-hollow aluminum profiles, such as those listed by Trina. Id. at 35. Given these explanations, Commerce's determination that HTS 7604 is more specific than the available alternatives on the record is supported by substantial evidence.

SolarWorld's argument that evidence specific to this review renders the selection of HTS 7604.29 unreasonable is unpersuasive. See SolarWorld's Br. at 11, n.3. SolarWorld first argues that the aluminum frames should not be valued using HTS 7604 because the HTS defines aluminum bars, rods, or profiles as having "a uniform cross section along their whole length ..., [sic] provided that they have not thereby assumed the character of articles or products of other headings." SolarWorld's Br. at 11-12 (citing Ex. SC-17 [attached to Trina's Sec. C Suppl. Questionnaire Resp.] at SuppC-24, PD 286-287, CD 314-319, bar code 3480083-02 (June 21, 2016) ). SolarWorld cites to an exhibit showing a drawing of the frame as evidence that the frames cannot fit the definition of HTS 7604, but Commerce reasonably concluded that the drawing in question shows "a single, uniform cross section," thus countering SolarWorld's argument. See Final Decision Memo at 36. SolarWorld offers no evidence to counter this determination. Moreover, Commerce's task is not to classify the solar frame inputs for customs purposes, but to select the best available information to value the FOPs in question.

See 19 U.S.C. § 1677b(c)(1). SolarWorld fails to proffer evidence detracting from Commerce's conclusion that the frames are more similar to the goods under HTS 7604 than any other data on the record.

SolarWorld also argues that the aluminum frames in question cannot be categorized as aluminum profiles due to the extent to which the product undergoes further processing. SolarWorld's Br. at 12-14. According to SolarWorld, the fact that "Trina's aluminum frames are fabricated products that have been further manufactured into a finished and final form" makes them not properly classifiable as aluminum profiles under HTS 7604.29.90001. SolarWorld's Br. at 14. The argument fails, however, given that HTS 7604 does not specify whether it covers finished or unfinished aluminum profiles, as Commerce explained. Final Decision Memo at 37. Moreover, the International Trade Commission's ("ITC") definition of aluminum profiles applies to goods "that have been subsequently worked after production ... provided that they have not thereby assumed the character of articles or products of the other headings." See Final Decision Memo at 36 (emphasis in original, quoting Jiangsu Jiasheng Photovoltaic Technology Co., Ltd. v. United States, 38 CIT ----, ----, 28 F.Supp.3d 1317, 1337 (2014) ). It is also reasonably discernible from Commerce's reference to the ITC's definition of profiles as covering goods "which have been subsequently worked after production," Final Decision Memo at 35, that the work performed on the profiles underscored by SolarWorld is not sufficient to render Trina's aluminum profiles more similar to the finished products covered by a different subheading. SolarWorld proffers no evidence detracting from Commerce's determination.

Finally, SolarWorld cites two CBP rulings, N139353 and N238208, which classify aluminum frames for solar panels under HTS 7616.99.5090 and HTS 8541.90.0000, respectively, to support its argument that HTS 7604 is not the best available information. See SolarWorld's Br. at 15-18. Commerce correctly noted, however, that it is not bound by CBP rulings for U.S. imports when selecting import values from surrogate countries, but rather should select the best available information on the record. Final Decision Memo at 37. Moreover, Commerce correctly pointed out that HTS 7616.99 covers an "other" subheading, which would solely comprise aluminum articles not already identified elsewhere. Id. Additionally, Commerce noted that the CBP rulings did not provide explanations for why the selected headings were appropriate, thus precluding Commerce from weighing the ruling against evidence on the record. Id. In light of Commerce's response, as well as the evidence relied upon by Commerce, the decision to select HTS 7604.29.9001 is reasonable.

C. Nitrogen

Commerce valued Trina's and Canadian Solar's nitrogen inputs using Thai import data under HTS 2804.30.00 (i.e, Hydrogen, rare gases and other non-metals; Nitrogen). Final Decision Memo at 52. Trina and Canadian Solar argue that Commerce's selection of the Thai import data is unsupported by substantial evidence because this data is aberrational and unreliable. See Canadian Solar's Br. at 31-39; Canadian Solar's Reply Br. at 12-15; Trina's Br. at 4-12; Trina's Reply Br. at 2-11. In the alternative, Canadian Solar argues that if Commerce uses the Thai import data, then input data from the United States and Switzerland included in the Thai data should be excluded as distortive. See Canadian Solar's Br. at 39-41; Canadian Solar's Reply Br. at 16-17. Defendant responds that Commerce reasonably determined that the Thai import data was not aberrational or unreliable, and that Commerce's selection is supported by substantial evidence. See Def.'s Br. at 26-31.

When analyzing whether data is aberrational, Commerce generally compares that data to (1) the AUV of data on the record for other countries at a level of economic development comparable to the non-market economy in question, and (2) the AUV for that input in the country at issue in prior years. Final Decision Memo at 53. Commerce's practice is to view data based on small quantities as not inherently distorted. Id. at 54.

Commerce's comparison of the Thai data to data from other economically comparable countries and from the past review reveals that this data is within the range of the relevant benchmarks as defined by Commerce's practice. Final Decision Memo at 53. Commerce compared the Thai import data and the import data available on the record for five other countries economically comparable to the PRC and found the Thai data fell within the range of the other AUVs. Id. at 53. Specifically, the Thai AUV ($ 9.36) was greater than the AUVs of three other countries on the record (Bulgaria with $ 0.08, Romania with $ 0.08, and Mexico with $ 0.25), but lower than the remaining two countries (South Africa with $ 26.27, Ecuador with $ 17.16). Id. at 53. Commerce also concluded that the mere fact that the volume of Thai imports was lower than the volume of imports from each of Bulgaria, Romania and Mexico did not, in and of itself, demonstrate distortion. Id. at 54.

Commerce also concluded that a comparison with the nitrogen AUV in the prior administrative review did not indicate that the Thai data was aberrational. Final Decision Memo at 54. The AUV for nitrogen in the Thai import data in the prior administrative review was $ 11.68, around 19% higher than in the present review. Id. at 54. Accordingly, Commerce's determination that the Thai import data is not aberrational is reasonable.

Canadian Solar argues that high value imports from the United States and Switzerland distort the value of Thai imports and should be excluded. See Canadian Solar's Br. at 39-41; Canadian Solar's Reply Br. at 16-17. Imports from these two countries account for 1.57% of imports into Thailand during the POR but comprise 22.6% of the total value of Thai imports. Canadian Solar's Br. at 39. Commerce reasonably declined to disaggregate the Global Trade Atlas ("GTA") data for nitrogen and exclude the imports from the United States and Switzerland on the basis of its preference for using a full dataset to avoid "cherry-picked import data in a[ ] [surrogate value] calculation." Final Decision Memo at 54. While Commerce's policy of not disaggregating data may not be reasonable in all circumstances, the court cannot say that Commerce's refusal to disaggregate the GTA data for nitrogen in this case is unreasonable. The AUV of nitrogen imports from the United States and Switzerland is around fifteen times greater than the AUV of all nitrogen imports into Thailand (including imports from the United States and Switzerland). See Final Decision Memo at 53; see also Summary of Import Data for Nitrogen & Oxygen from Economically Comparable Countries [attached as Ex. 3 to SolarWorld's Rebuttal Surrogate Value Comments] at 2, PD 397-98, CD 482-84, bar code 3490795-02 (July 26, 2016). It is reasonably discernable that Commerce did not consider this degree of price variation within the Thai data as sufficient to justify disaggregation where it had already concluded that credible benchmarks exhibit significant price variation. See Final Decision Memo at 53-54 (identifying import values for other potential surrogate countries in this period of review as ranging between $ 0.08-$ 27.27 per kilogram, and identifying price volatility between the present period of review and the prior period of review in South Africa ($ 26.27 vs. $ 5.46), Romania, ($ 0.08 vs. $ 0.13) and Ecuador ($ 17.16 vs. $ 4.84) ). Trina argues that Commerce unreasonably declined to use certain price quotes and invoices as benchmarks against which to assess whether the Thai import data was aberrational. See Trina's Br. at 6-7; Trina's Reply at 10. Commerce stated that it generally does not use price quotes because it cannot verify the conditions under which the quotes were solicited and whether they have been selected from within a broader range of quotes, and because price quotes do not represent actual prices or broad ranges of data. Final Decision Memo at 52-53. Commerce also stated that it considered individual prices to be unrepresentative of broad market averages. Id. at 53. Commerce's reasoning for declining to use the price quotes and prices as benchmarks is reasonable. Id. at 53.

Trina further argues that the Thai data should be treated as unreliable due to a "significant and inexplicable discrepancy" between the quantity and value of nitrogen imports reported from the United States into Thailand, and nitrogen exports recorded from the United States to Thailand by the ITC Dataweb. See Trina's Br. at 7-10; Trina's Reply Br. at 9-10. The ITC data shows that the United States exported 788,319 kilograms of nitrogen to Thailand during the period of review. See Analysis of U.S. Origin Nitrogen Included in Thai Imports of Nitrogen [attached as Enclosure 2 to Trina's Administrative Case Br.], PD 533, CD 588, bar code 3538946-01 (Jan. 25, 2017). By contrast, the Thai data shows that only 2,070 kilograms of nitrogen were imported from the United States during the period of review. Id. The U.S. data contains an AUV of approximately $ 0.15 per kilogram, while the AUV for the U.S. component of the Thai import data contains an AUV of approximately $ 140 per kilogram. Id. Commerce reasonably concluded, however, that country-specific export data are an inappropriate benchmark by which to evaluate corresponding import values. Final Decision Memo at 55. As Commerce explained, this is because different reporting and inspection requirements mean that each shipment of merchandise is likely to be treated differently. Id. (quoting Certain Activated Carbon from [the PRC]: Issues and Decision Mem. for the Final Results of the First [ADD] Admin. Review at 32, A-570-904, (Nov. 3, 2009), available at https://enforcement.trade.gov/frn/summary/prc/E9-27083-1.pdf (last visited Apr. 11, 2019) ). Trina's argument fails as Commerce has provided a reasonable explanation for the divergence between the U.S. export data and Thai import data, and a reasonable justification for not using the U.S. export data as a benchmark by which to evaluate the Thai import data.

D. Semi-Finished Polysilicon Ingots and Blocks

In the Final Results, Commerce determined that the best available information by which to value the respondents' semi-finished polysilicon ingots was the international price for solar-grade polysilicon. Final Decision Memo at 21-22. SolarWorld challenges Commerce's decision, arguing that the surrogate value does not reflect the substantial additional processing and value added by turning raw polysilicon into an ingot or block. SolarWorld's Br. at 29-32. SolarWorld proposes that Commerce instead construct a cost, starting with the world-market price of raw polysilicon and adding the costs required to produce a unit of ingot or block. Id. at 31-32. Defendant responds that, considering the available data on the record, Commerce reasonably chose to value respondents' ingots and blocks using a surrogate for the primary raw input. Def.'s Br. at 21-23. For the reasons that follow, Commerce's determination to value respondents' semi-finished polysilicon ingots and blocks with the world market price for raw polysilicon is reasonable.

Here, Commerce determined that the world market price for raw polysilicon constituted the best available information for valuing respondents' semi-finished polysilicon ingots and blocks. Final Decision Memo at 21-22. In the absence of data values for ingots and blocks, Commerce used a value for raw polysilicon because respondents' ingots and blocks are primarily composed of polysilicon. Id. With respect to SolarWorld's contention that the world market price for polysilicon is missing certain processing costs, Commerce explained that most of the processing required to produce ingots and blocks from raw polysilicon is performed by large, expensive machinery and is thus accounted for in the manufacturing costs. Id. Indeed, Commerce stated that the record did not present sufficient evidence to show that the additional processing stages add a significant amount of value beyond the original cost of polysilicon. Id. Given Commerce's decision to use a value for the main component of the good, it is reasonably discernible that Commerce viewed the option as imperfect but recognized that it would result in a more accurate surrogate value than would a value for a component that is not the main component of the good. See Rhone Poulenc, Inc., 899 F.2d at 1191. It is also reasonably discernible that Commerce determined, consistent with its practice, that the world market price for raw polysilicon is as specific as possible to the input (lacking data for the input itself), is contemporaneous, publicly available, and represents a broad market average. See Policy Bulletin 04.1. Commerce's determination that this value constitutes the best available information is reasonable.

E. Respondents' Surrogate Financial Ratios

In the Final Results, Commerce determined that Styromatic's 2015 financial statements constituted the best available information to calculate respondents' overhead, selling, general and administrative expenses, and profit. Final Decision Memo at 66-71. SolarWorld challenges Commerce's decision, arguing that Commerce should have selected the available financial statements of Thai company Ekarat Engineering Public Company Limited ("Ekarat") because Ekarat is a producer of identical merchandise, thus making it the best available information. SolarWorld's Br. at 21. Defendant responds that Commerce reasonably selected Styromatic's 2015 financial statements, as they constituted the best available information considering they were complete and contemporaneous with the period of review, and Styromatic's primary business is the production of comparable merchandise. Def.'s Br. at 34. For the reasons that follow, Commerce's determination is supported by substantial evidence.

As discussed, Commerce determines the normal value of the subject merchandise based on the FOPs utilized. 19 U.S.C. § 1677b(c)(1). Commerce values the FOPs using the best available information, to which it adds "an amount for general expenses and profit." Id. Commerce selects a surrogate value for each input from a source in a market economy country that is economically comparable to the NME country and a significant producer of the merchandise in question. Id. §§ 1677b(c)(4)(A)-(B) ; 19 C.F.R. § 351.408(b). Commerce determines the amount for manufacturing overhead, general expenses, and profit using publicly available financial data from a producer of identical or comparable merchandise. 19 C.F.R. § 351.408(c)(4). In choosing which surrogate data to use, Commerce considers "the quality and specificity of the statements, as well as whether the statements are contemporaneous with the data used to calculate production factors." Final Decision Memo at 67. Additionally, where Commerce "has reason to believe or suspect that a company may have received countervailable subsidies, financial ratios derived from that company's financial statements may not constitute the best available information." Id.

Commerce selected Styromatic's 2015 financial statements as the best available information because they were contemporaneous, audited, and came from a company that produced merchandise comparable to the subject merchandise during the period of review. Final Decision Memo at 71. SolarWorld asserts that Ekarat's financial statements are the best available information because the company is the only Thai manufacturer that produces identical merchandise. SolarWorld's Br. at 21. Defendant responds that Commerce reasonably explained that the majority of Ekarat's revenue comes from distribution transformers and services, which are not comparable to the subject merchandise. Def.'s Br. at 32-33.

Commerce reasonably concluded that Styromatic's 2015 financial statements constituted the best available information. Although Commerce acknowledged that Ekarat's 2015 consolidated financial statements indicate that Ekarat is a manufacturer of solar modules, Commerce reasonably concluded that Ekarat is primarily a manufacturer of distribution transformers. Final Decision Memo at 68. First, Ekarat's 2015 financial statements indicate that "[m]ost of the company's revenue came from the sale of Distribution Transformer," thus establishing that Ekarat was primarily engaged in the manufacturing of non-comparable merchandise during the POR. Ekarat 2015 Financial Statements [attached as Ex. 11 to SolarWorld's Initial Surrogate Value Comments] at 18, PD 365-382, CD 424-436, bar code 3489062-01 (July 18, 2016) ("Ekarat Financial Statements"). Second, the same statements indicate that in 2015, "the company had the sale revenue from sales of Distribution Transformer to the Metropolitan Electric Authority and the Provincial Electric Authority ... in the amount of ... Baht 267.60 million or ... 12.92%, ... of the Transformer revenue and Services revenue of the company." Ekarat Financial Statements at 18. From this figure, it is discernible that Ekarat made a total of 2,071.21 million baht in transformer revenue and services revenue (267.60 divided by 0.1292). The financial statements also indicate that Ekarat's total revenue for every customer sector was 2,092.12 million baht in 2015. Id. So, 2,071.21 million-Ekarat's 2015 transformer revenue and services revenue-is 99% of 2,092.12 million-the company's total 2015 revenue for customer sectors. Although SolarWorld argues that Ekarat can produce solar cells, see SolarWorld's Br. at 21-22 (citing Ekarat Financial Statements at 11), Commerce reasonably concluded that the vast majority of Ekarat's revenue comes from the sale of non-comparable merchandise and thus Ekarat should not be considered a producer of identical merchandise. Final Decision Memo at 68. As Ekarat's financial statements indicate, "[m]ost of the company's revenue came from the sale of" distribution transformers. Ekarat Financial Statements at 18.

Moreover, Commerce's practice is to decline to use financial statements of companies that are not profitable, and Ekarat Solar Co., Ltd., Ekarat's subsidiary described as a solar cell producer, operated at a loss in 2015. See Final Decision Memo at 69 (citing Canadian Solar Rebuttal Surrogate Value Comments at Ex. SVR-1, PD 391, CD 474, bar code 3490637-01 (July 25, 2016) ). Accordingly, Commerce's conclusion that Styromatic's 2015 financial statements-rather than those of Ekarat-represented the best available information is reasonable.

II. Commerce's Application of Partial AFA to Value Canadian Solar's Unreported FOPs

Canadian Solar challenges Commerce's application of partial AFA as unlawful and unsupported by substantial evidence. Canadian Solar's Br. at 10-19. Canadian Solar argues that because it cooperated with Commerce in the administrative proceedings, application of partial AFA was unlawful. Id. at 12-13. Canadian Solar further contends that Commerce's determination that Canadian Solar could have induced cooperation from its unaffiliated suppliers to provide FOP data was not supported by substantial evidence. Canadian Solar's Br. at 11, 13-17; see also Final Decision Memo at 15-18. Commerce reasoned that because the suppliers failed to cooperate by not providing the FOP information, 19 U.S.C. § 1677e(a) and (b) permit Commerce to apply AFA to account for the missing information. Final Decision Memo at 15-18. Commerce explained its use of AFA as necessary because Canadian Solar was "in a position to exercise leverage to induce cooperation from its uncooperative solar cell and solar module suppliers." Final Decision Memo at 16. For the reasons that follow, Commerce's application of partial AFA in calculating Canadian Solar's dumping margin is neither in accordance with law nor supported by substantial evidence.

To calculate accurate dumping margins, Commerce requests information from respondents. Where information necessary to calculate a respondent's dumping margin is not available on the record, Commerce applies "facts otherwise available" in place of the missing information. See 19 U.S.C. § 1677e(a). Where Commerce "finds that an interested party has failed to cooperate by not acting to the best of its ability to comply with a request for information," Commerce may apply "an inference that is adverse to the interests of that party in selecting among the facts otherwise available." Id. § 1677e(b). A respondent complies "to the best of its ability" where it "put[s] 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). The Court of Appeals has held that Commerce may, under certain circumstances, apply AFA in calculating a cooperative respondent's antidumping margin where it finds that the respondent could have induced an uncooperative supplier's cooperation. Mueller Comercial de Mexico S. De R.L. de C.V. v. United States, 753 F.3d 1227, 1233-34 (Fed. Cir. 2014).

In Mueller, Mueller was a cooperative respondent but did not possess all the production cost information Commerce needed to calculate its antidumping margin. 753 F.3d at 1230. Commerce requested data directly from Mueller's two main suppliers, but only one supplier provided the information. Commerce used facts otherwise available pursuant to 19 U.S.C. § 1677e(a), concluding that the unavailable production cost data was related to acquisition cost data on the record. Id. Commerce explained

Although premised on the adverse inference that [Mueller's uncooperative supplier's]

actual cost information would not be favorable - otherwise [the supplier] may not have elected to withhold it from the Department - the selected facts available are intended to produce an accurate, non-punitive, dumping margin for Mueller.

Issues and Decision Mem. for Final Results of [ADD] Administrative Review: Certain Circular Welded Non-Alloy Steel Pipe from Mexico at 16, A-201-805, June 13, 2011, ECF No. 59 (Mueller Comercial de Mexico, S. de R.L. de C.V. v. United States, Ct. No. 11-00319) ("Mueller IDM"). Commerce selected the most discounted transaction data from the responsive supplier, inferring that all merchandise sold to Mueller by the uncooperative supplier came at such a discount. Mueller, 753 F.3d at 1230. This selection enabled Commerce to calculate the uncooperative supplier's cost of production, and ultimately resulted in a higher dumping rate for Mueller. Id. Commerce justified its decision, in part, based on a finding that Mueller "could and should have induced" the cooperation of the uncooperative supplier, and that an adverse inference affecting Mueller was necessary to induce the uncooperative supplier's cooperation since it "could otherwise evade its antidumping rate by funneling its goods through Mueller." Id. at 1233. Mueller challenged the final results at the U.S. Court of International Trade ("CIT"), arguing that Commerce's application of AFA was improper, given Mueller's cooperation. Id. at 1230-31. After the CIT sustained Commerce's determination, Mueller appealed to the Court of Appeals for the Federal Circuit.

The Court of Appeals examined whether Commerce's determination could be sustained under 19 U.S.C. § 1677e(a). It held that Commerce may, under subsection (a), rely on inducement or evasion rationales where reasonable under the circumstances, and where "the predominant interest in accuracy is properly taken into account." Id. at 1233. The Court of Appeals explained that its holding was "justified and required, even if Commerce is viewed as acting entirely under subsection (a) in determining Mueller's rate." Id. The Court of Appeals' use of "even if" is properly read as "assuming, as we do here," given that Commerce indeed acted under the authority of subsection (a) in the administrative proceeding, and the Court undertook no prior analysis of subsection (b) in Mueller. Moreover, the Court's phraseology demonstrates the Court's subtle recognition of the importance of its holding-that policy rationales are permissible under subsection (a)-in light of the fact that subsection (b) has traditionally called for policy rationales, such as deterrence. See, e.g., Gallant Ocean (Thailand) Co., Ltd. v. United States, 602 F.3d 1319, 1323 (Fed. Cir. 2010) (quoting F.lli De Cecco Di Filippo Fara S. Martino S.p.A. v. United States, 216 F.3d 1027, 1032 (Fed. Cir. 2000) (an AFA rate should be "a reasonably accurate estimate of the respondent's actual rate, albeit with some built-in increase intended as a deterrent to non-compliance") ). The Court of Appeals noted that such considerations could be reasonable in Mueller's case, given that Mueller had an existing relationship with its supplier and thus "could potentially have refused to do business" as a tactic to induce cooperation. Mueller, 753 F.3d at 1234-35. As for the evasion rationale, the Court of Appeals explained that the uncooperative supplier-itself a mandatory respondent in the administrative proceeding-could potentially evade its own AFA rate by exporting its goods through Mueller if Mueller were assigned a favorable antidumping rate. Id. at 1235.

A close reading of section 1677e(b) and Mueller reveals that application of an inference adverse to the interests of a cooperating respondent under subsection (b) is not contemplated by the statute nor the Court of Appeals. First, the plain language of section 1677e(b) expressly limits Commerce's application of an adverse inference to the interests of the party that failed to cooperate. 19 U.S.C. § 1677e(b)(1) (where an interested party fails to cooperate, Commerce "may use an inference that is adverse to the interests of that party"). Second, in the determination giving rise to Mueller, Commerce expressly disavowed imposing an adverse inference against a cooperating party:

The Department has found that the necessary information is absent from the record because [Mueller's uncooperative supplier] failed to cooperate, and has not made a finding that Mueller failed to cooperate. Accordingly, the Department has not applied an adverse inference against the interest of Mueller. Instead, the Department has selected from the facts otherwise available, the best information to use in place of [the uncooperative supplier's] withheld cost data.

Mueller IDM at 16. The Court of Appeals did not reject Commerce's selection of facts available under subsection (a) despite its adverse effect on Mueller. The Court of Appeals' key observation-that subsection (a) "does not provide for the specific facts that should be used as a gap-filling mechanism"-set the stage for Mueller's lasting insight-that policy rationales are not reserved exclusively for the realm of subsection (b). See Mueller, 753 F.3d at 1234. "The statute on its face does not preclude Commerce from relying on the same considerations under subsection (a) for an AFA determination as used under subsection (b)." Id.

Here, several of Canadian Solar's unaffiliated suppliers of solar modules did not report information on their FOPs, information Commerce needed to calculate Canadian Solar's antidumping rate. Final Decision Memo at 15. Commerce applied "partial AFA" in place of the missing information. Final Decision Memo at 17. First, it noted that section 1677e(a) "provides that the Department shall apply 'facts otherwise available' if an interested party or any other person withholds information that has been requested." Id. at 15. Next, Commerce stated that subsection (b) "provides that the Department may use an adverse inference in applying the facts otherwise available when a party has failed to cooperate by not acting to the best of its ability to comply with a request for information." Id. Thereafter it selected Canadian Solar's highest consumption rates for FOPs for solar cells and modules sold in the United States "because the suppliers in question failed to cooperate by not acting to the best of their abilities to comply with a request for information." Id. at 18. In other words, Commerce used an adverse inference invoking 1677e(b) based on the suppliers' lack of cooperation to calculate Canadian Solar's rate.

Commerce's application of an adverse inference under subsection 1677e(b) is contrary to law. The plain meaning of the statute, as confirmed by Mueller, does not provide for an adverse inference against a cooperative respondent under subsection 1677e(b). It is undisputed that Canadian Solar cooperated in the administrative review. See Canadian Solar's Br. at 10; Def.'s Br. at 34-41; Oral Arg. at 00:14:45-00:15:00 (Defendant agreeing that Canadian Solar was a cooperative respondent); See also Unreported [FOPs] Mem. at 5, PD 517, CD 585, bar code 3533043-01 (Dec. 16, 2016) (stating that Commerce may apply AFA in determining a cooperative respondent's dumping margin to induce cooperation of the uncooperative party). As discussed, the statute allows Commerce to select facts available under subsection (a), not apply an adverse inference against a cooperating party under subsection (b). Commerce may consider inducement and evasion rationales in selecting amongst facts available under subsection (a), so long as accuracy remains the predominant concern. Mueller, 753 F.3d at 1233. The Court of Appeals in Mueller explained that Commerce is not prohibited "from drawing adverse inferences against a non-cooperating party that have collateral consequences for a cooperating party," but that was not the set of facts presented. Id. at 1236. Generally, Commerce's decision to rely on subsection (a) versus (b) (or vice versa) is critical, as the two subsections involve varying levels of review and trigger separate lines of jurisprudence. Id. at 1232 ("[t]hese two subsections have different purposes."), 1234 (discussing the jurisprudence under subsection (b) separately). In the case of a cooperating party, selection of facts available under subsection (a) where Commerce considers an inducement or evasion rationale necessarily involves a more searching review. Commerce must conduct "a case-specific analysis of the applicability of deterrence and similar policies," placing a "greater emphasis on accuracy" where the decision affects a cooperating party. Id. at 1234 (citing Changzhou Wujin Fine Chem. Factory Co., Ltd. v. United States, 701 F.3d 1367, 1379 (Fed. Cir. 2012) ). Here, Commerce purports to impose "partial AFA" pursuant to section 1677e(b), relying principally on Mueller. As explained, Mueller addressed a determination Commerce made under subsection (a), and its lessons thus apply to the selection of facts available under subsection (a). Commerce's application of an adverse inference-purportedly under subsection 1677e(b)-is therefore contrary to law.

To the extent Commerce purports to rely on 19 U.S.C. § 1677e(a) for its application of partial AFA, its determination that Canadian Solar could have potentially induced its uncooperative suppliers to cooperate is unsupported by substantial evidence. Commerce cites both subsections 1677e(a) and (b) in its determination, but found, "pursuant to [ section 1677e(b) ] of the Act, that the application of partial AFA is warranted." Final Decision Memo at 18. The statute makes clear that Commerce must invoke subsection (a) to reach subsection (b). Therefore, Commerce's invocation of subsection (a) would seem to be solely in service of its subsection (b) analysis. Nonetheless, in its determination Commerce discusses and purports to rely upon Mueller throughout, see Final Decision Memo at 16-17, a case that relies on a subsection (a) analysis. See Mueller, 753 F.3d at 1230, 1232. It is therefore possible that Commerce also attempts to rely upon subsection (a) or some combination of subsection (a) and (b). In any event, the record does not support Commerce's determination pursuant to an analysis under subsection (a), as Commerce fails to show that Canadian Solar had the type of long-standing relationships with its suppliers that would give it leverage in the marketplace.

Although the Court of Appeals in Mueller does not expound upon what constitutes an "existing relationship," surely more is required than what is present here. 753 F.3d at 1235 ; see also Changzhou, 701 F.3d at 1379 (holding that there was no justification for an AFA rate based in deterrence where the rate would affect only the cooperating party). Here, Commerce based its determination that Canadian Solar has long-term relationships with its suppliers on the fact that it maintains "supplier-specific accounts in the accounting system" and corporate divisions dedicated to the purchasing of solar cells. Final Dec