In late April, the Federal Circuit issued its opinion in Marmen Inc. v. United States, a case we’ve been tracking because it attracted an amicus brief. In this case, Marmen appealed a judgment of the Court of International Trade, which sustained a final determination by the Department of Commerce of a 4.94% dumping margin for utility-scale wind towers from Canada. In an opinion authored by Judge Prost and joined by Judges Taranto and Chen, the Federal Circuit vacated the judgment and remanded the case for further proceedings. This is our opinion summary.
Judge Prost began by outlining the procedural and factual background of the case:
Wind towers “are a component of utility scale wind turbine electrical power generating units used to convert the energy from wind to electrical energy.” . . . “Wind towers are tubular steel structures upon which the other major wind turbine components i.e., rotor blades and nacelles are mounted.” . . . “Wind towers have three types of sections: the base section, one or more mid-sections, and the top section.” . . . “[S]teel plate is the primary input for wind towers,” and “the quantity and thickness of steel plate consumed in a particular tower can vary significantly depending on the specification.” . . .
In July 2019, Commerce initiated an antidumping (“AD”) investigation of wind towers from Canada in response to an AD petition filed by the Wind Tower Trade Coalition (“WTTC”). Marmen Inc. and Marmen Énergie Inc. were selected as mandatory respondents. In February 2020, Commerce issued a preliminary AD determination, assigning Marmen a 5.04% dumping margin, and in June 2020, Commerce issued its final AD determination, assigning Marmen a 4.94% dumping margin.
Marmen appealed to the CIT, making three arguments that are relevant here: Commerce erred by (1) determining that Marmen’s steel plate costs did not reasonably reflect the costs associated with the production and sale of the products, and thus it was error for Commerce to weightaverage (or smooth) the reported steel plate costs; (2) rejecting a supplemental cost-reconciliation item that was intended to correct certain purchase information that had not been properly converted from U.S. dollars (“USD”) to Canadian dollars (“CAD”); and (3) using the average-to-transaction (“A-to-T”) methodology, instead of the ordinary average-to-average (“A-to-A”) methodology, to calculate Marmen’s dumping margin based on a misapplication of Cohen’s d test.
The CIT affirmed the weight-averaging of Marmen’s steel plate costs, . . . but remanded the case to Commerce on the other two issues. For the USD-to-CAD cost reconciliation, Commerce initially rejected the submission because it “constituted untimely and unsolicited new information.” . . . The CIT explained that “Commerce has a duty ‘to determine dumping margins as accurately as possible,’” . . . and concluded that it was an abuse of discretion to deny the currency-conversion correction solely on the basis that it was untimely . . . .
As to the appropriate A-to-A or A-to-T methodology, Commerce first decides whether “a pattern of export prices (or constructed export prices) for comparable merchandise . . . differ significantly among purchasers, regions, or periods of time.” . . . If such a pattern exists, then Commerce “explains why such differences cannot be taken into account using [the A-to-A method].” . . . If the A-to-A methodology cannot account for these differences, then Commerce may apply the A-to-T methodology instead to calculate a dumping margin. In Marmen I, Marmen argued that there was no significant difference among purchasers, regions, or periods of time, and thus Commerce should have used the A-to-A method to calculate the dumping margin. According to Marmen, Commerce erroneously found a significant price difference based on a misapplication of Cohen’s d test—a statistical test which can be used in certain circumstances to determine differences between the mean averages of two data sets. This, according to Marmen, led Commerce to incorrectly use the A-to-T methodology for calculating the dumping margin. . . . While Marmen I was pending before the CIT, our court issued Stupp Corp. v. United States, which called into question the appropriateness of Commerce’s use of Cohen’s d test. . . . Based on Stupp, the CIT remanded “the issue of Commerce’s use of the Cohen’s d test for Commerce to explain further whether the limits on the use of the Cohen’s d test were satisfied.” . . .
On remand from Marmen I, Commerce again rejected the USD-to-CAD cost reconciliation because it allegedly would double count an exchange-rate adjustment already reflected in the cost of goods sold and reported cost of production. . . . As to Commerce’s application of Cohen’s d test, and despite concerns with its use expressed by our court in Stupp, Commerce determined that “‘the assumptions of normality and roughly equivalent variances’ are not relevant to Commerce’s application of the Cohen’s d test” because Commerce’s use of the test is based on an entire population of data and not a sample. . . . The CIT, in Marmen II, agreed with Commerce on both issues and sustained Commerce’s determination of a 4.94% dumping margin.
Marmen timely appeals. We have jurisdiction under 28 U.S.C. § 1295(a)(5).
Before beginning her analysis, Judge Prost explained that the Federal Circuit reviews a decision of the Court of International Trade “de novo, applying anew the same standard used by that court in its consideration of Commerce’s determination.”
Turning to the merits, Judge Prost briefly summarized the process Commerce uses to calculate its dumping margin and how Commerce sorts “merchandise into groups to be compared based on a hierarchy of similar characteristics,” with each group being assigned a “control number (‘CONNUM’).” Judge Prost explained that Commerce weight-averaged the steel plate costs here by making a comparison between “the price of Marmen’s wind towers sold in the U.S. with those sold in its home market of Canada.” Judge Prost explained that, because “[t]here were no identical products,” when making product comparisons Commerce “matched foreign like products” based on physical characteristics reported by Marmen. Judge Prost explained that, because there was “little difference” in plate costs for different dimensions, Commerce “weight-averaged the reported steel plate costs for all reported CONNUMs, except the CONNUM for the thickest plate.” While Marmen argued this was “improper,” the court disagreed.
Judge Prost then addressed Marmen’s argument that “Commerce arbitrarily disregarded its standard test for when weight-averaging may be used,” which examines whether significant cost differences between “nearly identical” CONNUMs are unrelated to physical characteristics. Judge Prost highlighted Commerce’s response that its “cost-smoothing practice is not limited solely to instances in which the products are ‘identical’ or ‘very similar,’ and agreed with Commerce, concluding that “Commerce did not arbitrarily disregard its standard practice.”
Judge Prost next addressed Marmen’s argument “that Commerce’s factual findings are not supported by substantial evidence.” Judge Prost noted that Marmen itself had “first argued that ‘[t]hickness, in particular, affects steel plate cost,'” but that Marmen “did not identify surcharges for the smaller thicknesses” or “provide evidence of cost differences based on other physical characteristics.” In short, Judge Prost concluded that “Commerce’s findings are supported by substantial evidence.”
Judge Prost then addressed Marmen’s argument that “substantial evidence does not support Commerce’s decision to reject a minor correction to the cost-reconciliation worksheet to account for an exchange rate.” On this issue, she agreed with Marmen, explaining that neither of Commerce’s arguments for why it rejected this exchange rate “had merit.”
Lastly, Judge Prost addressed “Commerce’s use of Cohen’s d test.” She began with “a discussion of how Cohen’s d test is used in the process of calculating a dumping margin” and then turned to “why normal distributions, equal variability, and equally and sufficiently numerous data are necessary to achieving a meaningful Cohen’s d coefficient.” After briefly summarizing the “relevant conclusions” from the Federal Circuit’s “recent opinion in Stupp,” which provided a “thorough explanation of Cohen’s d test and its relevance to calculating a dumping margin,” Judge Prost determined that this case “begins where Stupp ended.” She indicated that the ultimate question to ask is whether “it was unreasonable for Commerce to use Cohen’s d test as part of its differential pricing analysis when the test is applied to data sets that do not satisfy the statistical assumptions (normal distribution, equal variability, and equally and sufficiently numerous data.)” Judge Prost concluded that it was unreasonable for Commerce to apply Cohen’s d test to “determine whether prices differ significantly when the underlying data is not normally distributed, equally variable, and equally and sufficiently numerous.”
As a result of Judge Prost’s analysis, the Federal Circuit “vacate[d] Commerce’s calculated dumping margin based on the unreasonable use of Cohen’s d test,” and remanded the matter for recalculation. The court indicated Commerce “may re-perform a different pricing analysis,” but “that analysis may not rely on Cohen’s d test for data sets like those here.”