Ellwood City Forge Co. v. United States

Slip Op. No. 25-68 UNITED STATES COURT OF INTERNATIONAL TRADE ELLWOOD CITY FORGE CO., ELLWOOD NATIONAL STEEL CO., ELLWOOD QUALITY STEELS CO., and A. FINKL & SONS, Plaintiffs, v. Before: Stephen Alexander Vaden, Judge UNITED STATES, Court No. 1:21-cv-00007 (SAV) Defendant, and BHARAT FORGE LTD., Defendant-Intervenor. OPINION [Sustaining Commerce’s Second Remand Determination] Dated: June 2, 2025 Paul K. Keith and Noah A. Meyer, Rock Creek Trade LLP of Washington, DC, for Plaintiffs Ellwood City Forge Co., Ellwood National Steel Co., Ellwood Quality Steels Co., and A. Finkl & Sons. With them on the brief was Jack A. Levy. Kara M. Westercamp, Senior Trial Attorney, Commercial Litigation Branch, Civil Division, U.S. Department of Justice, of Washington, DC, for Defendant United States. With her on the brief were Brian M. Boynton, Principal Deputy Assistant Attorney General, Patricia M. McCarthy, Director, and W. Mitch Purdy, Of Counsel, Office of the Chief Counsel for Trade Enforcement and Compliance, U.S. Department of Commerce. Court No. 1:21-cv-00007 (SAV) Page 2 Alexander Keyser, Fox Rothschild LLP, of Washington, DC, for Defendant-Intervenor Bharat Forge Limited. Vaden, Judge: Before the Court is the third installment in a case about Indian steel fluid end blocks. This saga began in early 2020 when Commerce initiated an antidumping investigation and selected an Indian producer, DefendantIntervenor Bharat Forge Limited (Bharat), as a mandatory respondent. Domestic petitioners Ellwood City Forge Company, Ellwood National Steel Company, Ellwood Quality Steels Company, and A. Finkl & Sons (collectively, Plaintiffs) challenged Commerce’s decision not to conduct an in-person verification of Bharat’s information because of the COVID-19 pandemic. This Court granted Commerce’s Motion for a Voluntary Remand to reconsider that decision. Order Granting Def.’s Mot. for Vol. Remand (Ellwood City I) at 5, ECF. No. 28. On remand, Commerce took new agency action but failed to follow the necessary procedural steps. This Court remanded for a second time. Ellwood City Forge Co. v. United States (Ellwood City II), 47 CIT __ , 654 F. Supp. 3d 1268, 1276-77 (2023). Commerce then completed an in-person verification at Bharat’s factory in India. The agency found a gap in the record and applied an adverse inference to two categories of information. Plaintiffs now argue that Commerce did not sufficiently address the broader reliability concerns raised by these gaps. The Court disagrees. Commerce’s Second Remand Determination will be SUSTAINED. Court No. 1:21-cv-00007 (SAV) Page 3 BACKGROUND The Court presumes familiarity with this case’s facts as described in its two previous opinions. See Ellwood City I, at 1–3, ECF No. 28; Ellwood City II, 47 CIT __ , 654 F. Supp. 3d at 1270–76. This opinion recounts only those facts relevant to the Court’s review of the Second Remand Determination. I. Procedural History This case concerns Commerce’s Final Determination in its antidumping investigation of Indian steel fluid end blocks. Forged Steel Fluid End Blocks from the Federal Republic of Germany, India, and Italy: Initiation of Less-Than-Fair-Value Investigations, 85 Fed. Reg. 2,394 (Dep’t of Com. Jan. 15, 2020). Fluid end blocks are blocks of forged steel that have been cut-to-length and modified for use in hydraulic oil and gas pumps. Fluid End Blocks from China, Germany, India, and Italy at 8– 10, Inv. Nos. 701-TA-632-635, 731-TA-1466, 731-TA-1468 (Final), USITC Pub. 5152 (Int’l Trade Comm’n Jan. 2021). The exact chemical and physical composition of a fluid end block depends on the amount of pressure to be placed on the fluid in the pump. Id. at 10. Fluid end blocks undergo a variety of processes to be suitable for use, including heat treating, milling, shaping, drilling, threading, and coating. Id. at 9. In some cases, manufacturers may add minor attachments to fluid end blocks at the request of customers, referred to by Commerce as “parts.” See Final Results of Redetermination Pursuant to Second Ct. Remand (Second Remand Determination) at 11, ECF No. 58. Court No. 1:21-cv-00007 (SAV) Page 4 Commerce selected Bharat as a respondent. See Forged Steel Fluid End Blocks from India: Preliminary Negative Determination of Sales at Less Than Fair Value and Postponement of Final Determination (Prelim. Determination), 85 Fed. Reg. 44,517 , 44,518 (Dep’t of Com. July 23, 2020). From January through July 2020, Commerce and Bharat exchanged questionnaires and responses. Prelim. Decision Mem. at 3 (PDM), J.A. at 83,289, ECF No. 71. Commerce’s questionnaires sought to determine if Bharat sold fluid end blocks in the United States at less-than-fair value. Commerce normally determines the fair value of a company’s products by looking at its sales price in its home market. 19 U.S.C. § 1677b(a)(1)(B)(i). Because Bharat had no sales in India, however, Commerce used the alternate “constructed value” methodology authorized by the Tariff Act. Id. § 1677b(a)(4). “Constructed value equals the cost of materials and fabrication or other processing, plus an amount for selling, general, and administrative expenses, as well as an amount for profit.” Vincentin S.A.I.C. v. United States, 43 CIT __ , 404 F. Supp. 3d 1323 , 1335 (2019) (citing 19 U.S.C. § 1677b(e)). Commerce began its constructed value calculation by assigning a control number to each type of fluid end block Bharat produces. A “control number,” often called a “CONNUM,” denotes a unique product1 based on relevant physical characteristics. See Xi’an Metals & Min. Imp. & Exp. Co. v. United States, 45 CIT __ , 520 F. Supp. 3d 1314 , 1321 n.4 (2021). Control numbers are made of digits that are “merely numerical stand-ins for a product’s physical characteristics[.]” Navneet 1 A reader may easily substitute “product” any time he reads “control number.” Court No. 1:21-cv-00007 (SAV) Page 5 Educ. Ltd. v. United States, 47 CIT __ , Court No. 1:22-cv-00132, 2023 Ct. Int'l Trade LEXIS 194, at *27 (Dec. 29, 2023). These digits are ordered according to a descending hierarchy where each successive digit represents a characteristic less important than the prior digit. See Matra Ams., LLC v. United States, 48 CIT __ , 681 F. Supp. 3d 1339, 1351 (2024). Thus, the first digit in a control number represents the most important characteristic and the last digit the least important. Id. Commerce calculates the constructed value of each control number by aggregating the costs associated with each step in the production process. Pls.’ Corrected Comments on Second Remand Results (Pls.’ Comments) at 5, ECF No. 69; see also PDM at 11–12, J.A. at 83,297–98, ECF No. 71. On July 23, 2020, Commerce issued its Preliminary Negative Determination finding that fluid end blocks from Bharat were not sold in the United States at less than fair value. Prelim. Determination, 85 Fed. Reg. at 44,517–18. It also committed to verify the information Bharat submitted. Id. COVID-19 changed Commerce’s plans, and it chose not to conduct an in-person verification. Ellwood City I, at 2, 4, ECF No. 28. Commerce instead issued a “questionnaire in lieu of performing an onsite verification” to Bharat on September 2, 2020. See Letter from Commerce to Bharat (Sept. 2, 2020), J.A. at 83,332–38, ECF No. 71. Despite Plaintiffs’ objections, Commerce proceeded with this approach. Ellwood City I, at 2, ECF No. 28. Commerce changed tact yet again and determined that — although it could not verify Bharat’s information — it would use the information Bharat provided as “facts available” in making its determination. See Forged Steel Fluid End Blocks from Court No. 1:21-cv-00007 (SAV) Page 6 India: Final Negative Determination of Sales at Less Than Fair Value (Final Determination), 85 Fed. Reg. 80,003 , 80,004 (Dep’t of Com. Dec. 11, 2020); 19 U.S.C. § 1677e(a) (flush language) (allowing Commerce to use “facts otherwise available” when making a determination as long as there is a gap in the record). II. The Remands Plaintiffs timely filed suit. They alleged that Commerce improperly forwent on-site verification. Compl. ¶¶ 22–23, ECF No. 8; Pls.’ Rule 56.2 Mot. for J. on the Agency R. Br. at 2–3, ECF No. 21. This Court granted Commerce’s Motion for a Voluntary Remand, permitting it to “perform on-site verification” and noting that travel restrictions to India were now significantly relaxed so that on-site verification was once again possible. Ellwood City I, at 2–3, 5, ECF No. 28; Def.’s Mot. for Vol. Remand at 5-6, ECF No. 24. On remand, Commerce did not perform an on-site verification. Instead, it claimed that it “reconsidered [its] use of a questionnaire in lieu of on-site verification” and decided that it considered Bharat’s information verified. Ellwood City II, 47 CIT __ , 654 F. Supp. 3d at 1277–78 (internal quotation marks omitted) (citing First Remand Determination at 4, ECF No. 29). Commerce argued that, because it could not conduct an in-person verification during the pandemic, it need not do so now — notwithstanding COVID’s abatement. First Remand Determination at 8–9, ECF No. 29. Commerce also claimed that it had not taken new agency action despite reversing its prior position on whether Bharat’s questionnaire responses constituted a lawful verification. Ellwood City II, 47 CIT __ , 654 F. Supp. 3d at 1276–77. Court No. 1:21-cv-00007 (SAV) Page 7 The Court remanded to Commerce for a second time. It explained that under Department of Homeland Security v. Regents of the University of California, 591 U.S. 1 (2020), an agency has two possible paths on remand: “[It] can offer a fuller explanation of its reasoning at the time it made the decision,” or it “can take new agency action….” Ellwood City II, 47 CIT __ , 654 F. Supp. 3d at 1277 (citing Regents, 591 U.S. at 20–21). An agency “is not limited to its prior reasons but must comply with ... procedural requirements” when it takes new action. Id. (quoting Regents, 591 U.S. at 21). For example, “when an agency rescinds a prior policy its reasoned analysis must consider the ‘alternative[s]’ that are ‘within the ambit of the existing [policy].’” Regents, 591 U.S. at 30 (quoting Motor Vehicle Mfrs. Ass’n. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 51 (1983) (alterations in original)). And when deviating from a past practice or policy, an agency “must ‘be cognizant that longstanding policies may have engendered serious reliance interests that must be taken into account.’” Id. (quoting Encino Motorcars, LLC v. Navarro, 579 U.S. 211, 222 (2016)). The Court ordered Commerce to comply with the Regents requirements for new agency action by explaining (1) why it chose not to do on-site verification, (2) the range of other alternatives the agency considered and why it rejected them, and (3) why its decision to use only questionnaires did not violate Plaintiffs’ reliance interests. Ellwood City II, 47 CIT __ , 654 F. Supp. 3d at 1278–79. III. Present Dispute In response to the Court’s Second Remand Order, Commerce completed an inperson verification at Bharat’s factory in Pune, India, from October 30 through Court No. 1:21-cv-00007 (SAV) Page 8 November 7, 2023. Second Remand Determination at 3, ECF No. 58. Commerce then issued its Second Remand Determination on February 7, 2023. Id. at 1. Commerce continued to calculate a de minimis dumping margin for Bharat. Id. at 2. Commerce could not verify the molybdenum content for one steel grade Bharat sold. See Second Remand Determination at 12–13, ECF No. 58. Molybdenum is a chemical element often used in steel production. Uses of Molybdenum Alloys in the Aerospace and Steel Industry, STANFORD ADVANCED MATERIALS, https://perma.cc/9YXC-9WUA (last visited June 2, 2025). Each control number had one digit representing the molybdenum content of the fluid end blocks. Pls.’ Comments at 5–6, ECF No. 69. The antidumping order only covers fluid end blocks that have between .15 and 3 percent molybdenum content by weight. Final Determination, 85 Fed. Reg. at 80,004. For two products of the same steel grade, Bharat presented documentation showing both as having out-of-scope molybdenum content. See Letter from Commerce to Bharat at 3 (Sep. 2, 2020), J.A. at 83,336, ECF No. 71; Pls.’ Pre-Verification Comments at 4–6, J.A. 99,397–99, ECF No. 71. Bharat confessed at verification that this was incorrect and presented updated documents from a third party showing the steel grade to have in-scope molybdenum content. Second Remand Determination at 12, ECF No. 58. Commerce could not verify some of these new certificates, creating a gap in the record. Def.’s Resp. to Pls.’ Comments on Second Remand Results (Def.’s Comments) at 7–8, ECF No. 65. Commerce applied an adverse inference to fill that gap, finding that Bharat did not comply to the best of its ability. Second Remand Determination at 22–23, ECF No. 58. Based on this Court No. 1:21-cv-00007 (SAV) Page 9 adverse inference, Commerce concluded that all sales of this steel grade have in-scope levels of molybdenum. Id. at 12–13. It therefore assigned the highest individual dumping margin calculated for Bharat’s other U.S. sales to all sales of this steel grade. Id. Commerce also found that Bharat failed to report “parts” costs for two control numbers. The last and least important digit in each control number represented minor attachments purchased by Bharat and added to the fluid end blocks at customers’ request. Second Remand Determination at 11, ECF No. 58. For two products, Bharat “reported [each] with a physical characteristic for parts” but failed to report the corresponding costs. Pls.’ Comments at 8, ECF No. 69. Commerce found that Bharat’s failure to report these costs created a gap in the record. Def.’s Comments at 7, ECF No. 65. It also found that Bharat failed to comply to the best of its ability to provide these parts costs. Second Remand Determination at 23, ECF No. 58. As a result, Commerce chose to fill the gap in the record with an adverse inference. Id. Commerce increased “the direct material costs” for each missing parts cost “by the highest amount of parts [costs] reported for any [control number].” Id. at 5. Plaintiffs now argue that the Second Remand Determination is not supported by substantial evidence. Pls.’ Comments at 5, ECF No. 69. They allege that Commerce misunderstood the molybdenum issue and chose a method of verification for other steel grades that would not uncover similar errors. Id. at 7–8. Plaintiffs also argue that Commerce failed to address the broader reliability concerns raised by Court No. 1:21-cv-00007 (SAV) Page 10 the errors in Bharat’s cost reporting. Id. at 10–11. In Plaintiffs’ view, “Commerce fails to support its assertions that ... [these] errors ... were limited in scope” with substantial evidence. Id. at 5. The Government responds that Commerce fully complied with the Court’s Second Remand Order. Def.’s Comments at 1, ECF No. 65. Commerce asserts it was not required to use a broader adverse inference. Id. at 5–7. It also notes that, during its week-long verification tailored to Plaintiffs’ concerns, it uncovered almost no errors beyond those already identified. Id. at 3, 8–9. Commerce maintains that Plaintiffs improperly ask it to “proactively justify” any decision not to draw an adverse inference against all of Bharat’s data. Id. at 9. JURISDICTION AND STANDARD OF REVIEW This Court has exclusive jurisdiction over Plaintiffs’ challenge to Commerce’s Final Determination under 19 U.S.C. § 1516a(a)(2)(B)(i) and 28 U.S.C. § 1581 (c), which grant the Court authority to review actions contesting final affirmative determinations, including any negative part of such determinations, in an antidumping order. The Court must sustain Commerce’s “determination, finding, or conclusion” unless it is “unsupported by substantial evidence on the record, or otherwise not in accordance with law[.]” 19 U.S.C. § 1516a(b)(1)(B)(i). Substantial evidence “means such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.” Consol. Edison Co. v. NLRB, 305 U.S. 197, 229 (1938) (citations omitted). Additionally, “The court reviews remand determinations Court No. 1:21-cv-00007 (SAV) Page 11 for compliance with the court’s remand order.” Nakornthai Strip Mill Pub. Co. Ltd. v. United States, 32 CIT 1272 , 1274 (2008) (citations omitted). DISCUSSION In Ellwood City II, the Court ordered Commerce to comply with the requirements for new agency action set out in Regents. Commerce was to explain (1) why it chose not to do on-site verification, (2) the range of other alternatives the agency considered and why it rejected them, and (3) why its decision to use only questionnaires did not violate Plaintiffs’ reliance interests. Ellwood City II, 47 CIT __ , 654 F. Supp. 3d at 1278–79. Now that the agency has conducted in-person verification, no party disputes that it has complied with Regents. Pls.’ Comments at 5, ECF No. 69. Plaintiffs’ only remaining argument is that Commerce’s determination was deficient because Commerce failed to account for how errors revealed at verification undermined the entirety of Bharat’s information. Id. This Court disagrees. Commerce has complied with both Regents’ requirements, and its Second Remand Determination is supported by substantial evidence as well as Federal Circuit precedent. The Second Remand Determination is SUSTAINED. I. Commerce satisfied its Regents obligations. First, Commerce addressed any issues arising from Ellwood City’s reliance interests when Commerce returned to its past practice. When deviating from a consistent past practice or policy, an agency “must ‘be cognizant that longstanding policies may have engendered serious reliance interests that must be taken into account.’” Regents, 591 U.S. at 30 (quoting Encino Court No. 1:21-cv-00007 (SAV) Page 12 Motorcars, 579 U.S. at 222 ). Because Commerce’s longstanding policy was to perform in-person verifications, Ellwood City had a reliance interest rooted in this past practice that Commerce needed to consider. See id. at 30. By conducting an in-person verification, Commerce gave the parties the same treatment they would have received but-for the pandemic. The harm to Ellwood City stemming from its reliance interests is gone because Commerce has returned to its past practice. No party disputes this. Pls.’ Comments at 5, ECF No. 69; Def.’s Comments at 5, ECF No. 65. The Court therefore finds that Commerce satisfied the first Regents requirement by returning to its past practice and conducting an in-person verification. Second, Commerce reconsidered its decision not to conduct on-site verification and examined if it had any alternative options for verification. Commerce answered by conducting an in-person verification. Second Remand Determination at 3, ECF No. 58. Because Commerce returned to its practice of in-person verification, there is no need for it to consider alternatives closer to that policy. No party objects to Commerce’s decision. See Pls.’ Comments at 5, ECF No. 69. The Court therefore finds that Commerce appropriately “deal[t] with the problem afresh.” Regents, 591 U.S. at 21 (quoting SEC v. Chenery Corp., 332 U.S. 194, 201 (1947)). II. Procedure no longer serving as an obstacle, the Court turns to the merits of the case. Ellwood City argues that Commerce’s decision is unsupported by substantial evidence. Pls.’ Comments at 2–3, ECF No. 69. It rests its argument on Commerce’s identification of two errors in Bharat’s data, which it claims raise broader reliability Court No. 1:21-cv-00007 (SAV) Page 13 concerns. Those errors are (1) errors in Bharat’s mill certificates, which misreported the molybdenum content of some in-scope merchandise, and (2) misreported parts cost for some products. Id. at 5–11. Plaintiffs argue that Commerce failed to address the implications of these errors. Id. at 5. Commerce responds that it reasonably concluded that these errors were isolated, did not raise broader reliability concerns, and required drawing only partial adverse inferences. Def.’s Comments at 9–15, ECF No. 65. After discussing the relevant law, the Court takes each of these arguments in turn. A. Commerce “shall verify all information relied upon in making … a final determination in an investigation[.]” 19 U.S.C. § 1677m(i)(1). But the law does not require Commerce to turn over every stone. See PMC Specialties Grp., Inc. v. United States, 20 CIT 1130 , 1134 (1996) (quoting Monsanto Co. v. United States, 12 CIT 937 , 944 (1988)) (“[V]erification … is not intended to be an exhaustive examination of the respondent’s business.”). Instead, “verification is a spot check” in which Commerce “has considerable latitude in picking and choosing which items it will examine in detail.” Id. The Federal Circuit has found that Commerce enjoys “wide latitude in its verification procedures.” Am. Alloys, Inc. v. United States, 30 F.3d 1469 , 1475 (Fed. Cir. 1994) (citations omitted). A gap in the record exists when necessary information “cannot be verified….” 19 U.S.C. § 1677e(a)(2)(D). The Tariff Act provides a process to identify and fill these gaps. See 19 U.S.C. § 1677e(a)–(b). The statute allows Commerce to use “facts Court No. 1:21-cv-00007 (SAV) Page 14 otherwise available” in place of the missing information when an interested party “withholds information that has been requested” or the relevant information “cannot be verified.” Id. § 1677e(a)(2)(A),(D). Separately, 19 U.S.C. § 1677e(b) permits those facts otherwise available to be selected based on an adverse inference if “an interested party has failed to cooperate by not acting to the best of its ability to comply with a request for information from [Commerce].…” Id. § 1677e(b)(1). Although § 1677e(a) and § 1677e(b) are often collapsed into the term “adverse facts available” or “AFA,” the two statutory processes require distinct analyses rather than the single analysis implied by the term “AFA.” Commerce first must determine that it is missing necessary information. And if it wishes to fill the resulting gap with facts adverse to an interested party, Commerce must separately determine that the party has failed to cooperate by not acting to the best of its ability. See Zhejiang DunAn Hetian Metal Co. v. United States, 652 F.3d 1333, 1346 (Fed. Cir. 2011). Substantial evidence must support Commerce’s decision to draw an adverse inference. See Diamond Sawblades Mfrs. Coal. v. United States, 986 F.3d 1351, 1366 (Fed. Cir. 2021). Commerce does not have a blank check to disregard all of a party’s information just because it finds a gap in the record. Id. at 1366–67. It must balance the error’s effect with the fact that the “overriding purpose of Commerce’s administration of antidumping laws is to calculate dumping margins as accurately as possible.” Yangzhou Bestpak Gifts & Crafts Co. v. United States, 716 F.3d 1370, 1379 (Fed. Cir. 2013). The agency cannot use an adverse inference to “impose punitive, aberrational, or uncorroborated margins.” F. Lii de Cecco di Filippo Fara Court No. 1:21-cv-00007 (SAV) Page 15 S. Martino S.p.A. v. United States, 216 F.3d 1027, 1032 (Fed. Cir. 2000). Instead, Commerce must use its power “within the constraints of the statute and record.” Oman Fasteners, LLC v. United States, 125 F.4th 1068 , 1086 (Fed. Cir. 2025) (citation omitted). B. Commerce’s first adverse inference concerned the molybdenum content of one steel grade Bharat sold. Bharat reported the molybdenum content for each control number. PDM at 9, J.A. at 83,295, ECF No. 71. Commerce looked at mill certificates from Bharat to verify these reports. See Second Remand Determination at 12, ECF No. 58. Mill certificates show the chemical composition of the fluid end blocks. Pls.’ Comments at 6, ECF No. 69. Bharat submitted mill certificates for two products of the same steel grade, each showing a different molybdenum content. Id. at 6; Letter from Commerce to Bharat (Sep. 2, 2020), J.A. at 83,336, ECF No. 71. The mill certificates for a particular steel grade should be identical. Pls.’ Comments at 6–7, ECF No. 69. During verification, Bharat presented updated mill certificates for this steel grade. See Second Remand Determination at 12, ECF No. 58. Four of the fourteen new certificates Bharat submitted to Commerce differed from the original mill certificates. Id. Commerce determined that Bharat’s mill certificates for this steel grade were unreliable because of these inconsistencies. Id. at 12–13. It therefore (1) applied an adverse inference, (2) found all sales of this grade to be in-scope Court No. 1:21-cv-00007 (SAV) Page 16 merchandise, and (3) assigned these sales the highest individual dumping margin calculated for Bharat’s other U.S. sales. Id. Plaintiffs raise two objections. First, they argue that the mill certificate error raises broader reliability concerns. Pls.’ Comments at 5, ECF No. 69. This error misreported all sales for one steel grade as having out-of-scope molybdenum content, and Bharat could not explain how the error occurred. Second Remand Determination at 19, ECF No. 58. Plaintiffs argue Commerce cannot be sure that other steel grades do not have the same problem. Pls.’ Comments at 7, ECF No. 69. Second, they assert that Commerce’s attempts to account for this error — by looking at mill certificates for other steel grades — fundamentally misunderstands the nature of the issue. Commerce verified the mill certificates by seeing if they matched the information Bharat provided. Second Remand Determination at 12, ECF No. 58. But Plaintiffs note that this comparison only tells Commerce whether the certificates and Bharat’s information match. See Pls.’ Comments at 7–8, ECF No. 69. It does not tell Commerce whether the certificate or Bharat’s information accurately states the molybdenum content of particular steel grades. Id.; see also Oral Arg. Tr. at 16:8–22, ECF No. 82. Ellwood City observes that, for the misreported steel grade, both Bharat’s initial disclosure and the initial mill certificates were inaccurate. Pls.’ Comments at 7–8, ECF No. 69. Simply examining the mill certificates and seeing if they match Bharat’s information cannot tell Commerce the actual molybdenum content of each product and therefore cannot support Commerce’s determination that the mill certificate issue was isolated to one steel grade. See id. Court No. 1:21-cv-00007 (SAV) Page 17 Plaintiffs’ arguments miss the mark. The steel grade at issue accounted for only a tiny percentage of Bharat’s overall sales. Second Remand Determination at 23, ECF No. 58. For Commerce to draw an adverse inference, the agency must find a gap in the record. 19 U.S.C. § 1677e(a)–(b). Commerce examined mill certificates for every steel grade sold in the U.S. during the period of investigation to see if there were further mismatches. See Sales Verification Report at 7–10, J.A. at 99,767–71, ECF No. 71; see generally Sales Verification Exs. 1–17, J.A. at 99,982–100,684, ECF No. 71. There were not. Because Commerce verified the rest of Bharat’s information, there is no gap in the record to fill. Without a gap to fill, there can be no adverse inference. See 19 U.S.C. § 1677e(a). The statute does not force Commerce to create implied gaps in the rest of the record based on narrow errors. See Oman Fasteners, 125 F.4th at 1086 (holding that small errors do not allow Commerce to ignore the reality of what is in the record). Commerce discovered the mill certificate error because some certificates for the same steel grade did not match. Pls.’ PreVerification Comments at 4–6, J.A. 99,397–99, ECF No. 71. But Plaintiffs want Commerce to assume inaccuracy even when the mill certificates do match. They therefore ask Commerce to impute a different, much more serious error onto the rest of Bharat’s reports. Plaintiffs offer only one solution for how Commerce could have verified Bharat’s information to their satisfaction — testing every steel grade to prove its molybdenum content. See Oral Arg. Tr. at 17:8–22, ECF No. 82. They misunderstand verification. Verification is not “an unending haystack search for a needle.” Bonney Court No. 1:21-cv-00007 (SAV) Page 18 Forge Corp. v. United States, 49 CIT __ , Court No. 1:20-cv-03837, 2025 Ct. Int’l Trade LEXIS 59, at *22 (May 6, 2025). The majority of the mill certificates — ten of fourteen — for the steel grade at issue accurately reported the molybdenum content. Second Remand Determination at 12, ECF No. 58. Nonetheless, as an adverse inference, Commerce assigned every sale of this steel grade the highest dumping margin in the record. Id. at 12–13. Plaintiffs ask this Court to force Commerce to assume that every mill certificate for every other steel grade was inaccurate — despite those grades’ having matching certificates. Errors found in only a subset of the data do not mandate that Commerce conclude that all the data is inaccurate. Diamond Sawblades, 986 F.3d at 1365 (explaining that the law requires “information-specific consideration … rather than any blanket disregard … whenever some of the information supplied by that person is unreliable.”). The failure of mill certificates to match does not mandate that Commerce dig deeper into mill certificates that do match. Accord Bonney Forge Corp., 49 CIT __ , 2025 Ct. Int’l Trade LEXIS 59, at *11 (“‘[J]ust keep looking’ is not a valid objection.”). Commerce appropriately limited its drawing of an adverse inference to the data type affected by the error it found. Substantial evidence supports its decision. C. Commerce also applied an adverse inference to parts costs for two control numbers. For each control number, Bharat reported whether the product had a physical characteristic for “parts” — minor attachments added to the fluid end blocks at the request of customers. Second Remand Determination at 11, ECF No. 58. If so, Court No. 1:21-cv-00007 (SAV) Page 19 Bharat reported the costs associated with those parts. Id. at 11–12. During verification, Commerce discovered that Bharat reported two control numbers “with a physical characteristic for parts ... but … [not] any part costs.” Cost Verification Report at 18, J.A. at 99,801, ECF No 71. For one of the products, Bharat should have reported costs; and for the other, it should have reported no physical characteristic for parts. Pls.’ Comments at 8, ECF No. 69. Commerce consequently applied an adverse inference to the parts costs for the two products where they were missing. Def.’s Comments at 8, ECF No. 65. Plaintiffs try to paint these errors as tainting the rest of Bharat’s cost reporting. Because U.S. sales of each product are compared with their constructed costs, correct cost reporting is key. Pls.’ Comments at 9–10, ECF No. 69. Without it, Commerce cannot accurately determine if Bharat was dumping the products at issue. See id. Plaintiffs argue that, even if parts costs were a small share of overall costs, the importance of accurate cost reporting means that these errors call into question Bharat’s information more broadly. Id. On this issue, Commerce also acted within the bounds of its latitude. It discovered the missing parts costs at verification because it examined a variety of documentation, such as purchase orders, packing lists, and invoices. Cost Verification Report at 18, J.A. at 99,801, ECF No. 71. Commerce reviewed the cost calculations for all fluid end blocks produced during the period of investigation. See id. at 3, J.A. at 99,786; Second Remand Determination at 11–12, ECF No. 58. It only found errors in two of fifteen control numbers, which accounted for a very small share Court No. 1:21-cv-00007 (SAV) Page 20 of Bharat’s overall sales. Def.’s Comments at 3–4, 13, ECF No. 65. And parts costs were only a tiny percentage of Bharat’s overall costs. Id. at 4. The last control number digit represented parts costs. Second Remand Determination at 11, ECF No. 58. Because control number digits go from greatest relevance to least relevance, parts costs represented the least important factor in the underlying products’ overall cost structure. See id. at 11–12. Plaintiffs’ effort to broaden the scope of these isolated errors is unconvincing. The misreported data comprised a tiny fraction of the costs for two out of fifteen products. Def.’s Comments at 4, ECF No. 65. Plaintiffs attempt to ratchet up this error by noting that Commerce must ensure that Bharat’s cost reporting as a whole is accurate. Pls.’ Comments at 9, ECF No. 69. However, Plaintiffs provide no explanation for why small errors, in only a subset of control numbers, call the overall accuracy of the data into question. See id. (claiming that “discovering the underreporting of [control number] costs cannot be waived away” without providing further explanation). They simply repeat that accurate cost reporting is important. Id. But that alone does not make small errors more serious. Commerce had no reason to apply a broader adverse inference to induce cooperation. Commerce found that Bharat’s actions were not those of “a noncooperating party working to hide” sales below fair market cost. Second Remand Determination at 19, ECF No. 58. Although there is no intent requirement in the antidumping statute, see Nippon Steel Corp. v. United States, 337 F.3d 1373, 1383 (Fed. Cir. 2003), narrow errors only justify broadly applying an adverse inference Court No. 1:21-cv-00007 (SAV) Page 21 when there is “a particularly strong need to deter non-compliance, which would have to rest on a particularly serious failure to cooperate[.]” Oman Fasteners, 125 F.4th at 1087. Commerce found that no such need existed. Bharat was a first-time respondent that was not trying to frustrate Commerce’s investigation. Second Remand Determination at 19, ECF No. 58. Given the limited nature of the errors Commerce discovered, its decision to limit the application of adverse inferences to the data sets affected by the errors is well justified. D. Precedent supports Commerce’s decision to narrowly draw adverse inferences. In Diamond Sawblades, Commerce requested country-of-origin records from a Chinese manufacturer’s U.S. affiliates. 986 F.3d at 1354 . The company did not keep detailed records on the origin of its merchandise. Instead, it used a variety of methods to identify the country of origin for Commerce. Id. One of these accounting methods, known as first-in, first-out, categorized merchandise based on when it arrived in the United States and used those arrival dates to determine the country of origin. Id. at 1358 . Commerce found this method to be unreliable and drew an adverse inference against all country-of-origin data, not just the subset that used the first-in, first-out methodology. Id. at 1354 . The Federal Circuit held that Commerce failed to justify disregarding all the company’s country of origin data. See id. at 1354-55 . As the appellate court explained, it is not enough that a subset of information is unreliable. If Commerce wishes to draw a broader adverse inference, it must provide a Court No. 1:21-cv-00007 (SAV) Page 22 “comprehensible explanation for why ... the errors found ... have a reliabilityundermining effect outside the category of sales” identified as unreliable. Id. at 1366 . Government of Quebec v. United States is not to the contrary. See Pls.’ Comments at 11-12, ECF No. 69 (citing Gov’t of Que. v. United States, 46 CIT __ , 567 F. Supp. 3d 1273 (2022)).2 There, Commerce examined the total sales value of a company’s goods to calculate its countervailable subsidy rate. Gov't of Que. v. United States, 105 F.4th 1359 , 1363–64 (Fed. Cir. 2024). The Canadian exporter of utilityscale wind towers wanted Commerce to account for an “exchange rate adjustment” its auditor performed. Id. This adjustment converted sales made in foreign currencies back into Canadian Dollars. Id. at 1364 . At verification, Commerce discovered that some of the “converted” sales were classified in U.S. Dollars but were recorded in Euros. Id. And two of these “Euro” transactions were actually made in Canadian Dollars so that no conversion was needed at all. Id. Commerce declined to apply the requested adjustment because it determined that the errors raised broader reliability concerns about the adjustment as a whole, and it could not verify whether the remaining sales were accurately recorded. Id. at 1365 . Contrary to Plaintiffs’ request here, in Government of Quebec, Commerce only applied an adverse inference to one category of data — sales with an auditor’s adjustment. Id. It did not discredit other data that it verified and determined to be accurate. See id. (noting that Commerce still used the rest of the respondent’s sales 2 Since the parties here filed their briefs, the Federal Circuit has affirmed this Court’s decision in Government of Quebec. See Gov't of Que. v. United States, 105 F.4th 1359 (Fed. Cir. 2024). The Court therefore will rely on the precedential Federal Circuit decision. Court No. 1:21-cv-00007 (SAV) Page 23 data). Here, Commerce similarly chose to use the portion of Bharat’s data that passed verification and only drew an adverse inference against the data sets that failed verification. Second Remand Determination at 22–23, ECF No. 58. And unlike in Government of Quebec, Commerce went the extra mile and verified the mill certificates for every steel grade and the cost calculations for every fluid end block produced during the period of investigation. Compare Gov't of Que., 105 F.4th at 1370 (“[Commerce] lacked the ability to verify each sale and exhaustively examine all underlying sales documentation.”), with Second Remand Determination at 12, ECF No. 58 (“This problem was only found with respect to [one] steel grade …. We reviewed mill test certificates for other steel grades and did not find this problem ….”), and Cost Verification Report at 3, J.A. at 99,786, ECF No 71 (“[W]e examined the cost buildup worksheets that Bharat Forge used to prepare the reported costs for all [fluid end blocks][.]”). Plaintiffs ask that Commerce apply an adverse inference to whole categories of information that the agency found to be error-free. See Pls.’ Comments at 3, ECF No. 69. When information passes verification, there is no gap to fill nor any adverse inference to draw. 19 U.S.C. § 1677e(a)–(b). Government of Quebec does not require otherwise. See Gov't of Que., 105 F.4th at 1365 (noting that Commerce “relied on … reported sales information” other than the specific data that could not be verified). CONCLUSION Once again, the third time proved to be the charm. Commerce satisfied Plaintiffs’ reliance interests by completing an in-person verification. Although Court No. 1:21-cv-00007 (SAV) Page 24 Plaintiffs complain that Commerce should have more broadly drawn adverse inferences based on the small errors it found at verification, neither statute nor case law requires such an inequitable result. Neither will this Court. Commerce’s Second Remand Determination is SUSTAINED. /s/ Stephen Alexander Vaden Stephen Alexander Vaden, Judge Dated: June 2, 2025 New York, New York