Summary
referring to the expression as the "pooled variance" because and are the variances of the prices in the two groups
Summary of this case from Mid Continent Steel & Wire, Inc. v. United StatesOpinion
2018-1229 2018-1251
10-03-2019
MID CONTINENT STEEL & WIRE, INC., Plaintiff-Appellant v. UNITED STATES, Defendant-Appellee PT Enterprise Inc., Pro-Team Coil Nail Enterprise Inc., Unicatch Industrial Co., Ltd., WTA Interntional Co., Ltd., Zon Mon Co., Ltd., Hor Liang Industrial Corporation, President Industrial Inc., Liang Chyuan Industrial Co., Ltd., Defendants-Cross-Appellants
Adam H. Gordon, The Bristol Group PLLC, Washington, DC, argued for plaintiff-appellant. Also represented by Ping Gong. Mikki Cottet, Appellate Staff, Civil Division, United States Department of Justice, Washington, DC, argued for defendant-appellee. Also represented by Jeanne Davidson, Joseph H. Hunt, Patricia M. McCarthy. Andrew Thomas Schutz, Grunfeld, Desiderio, Lebowitz, Silverman & Klestadt LLP, Washington, DC, argued for defendants-cross-appellants. Also argued by Ned H. Marshak, New York, NY. Also represented by Max Fred Schutzman, New York, NY; Kavita Mohan, Washington, DC.
Adam H. Gordon, The Bristol Group PLLC, Washington, DC, argued for plaintiff-appellant. Also represented by Ping Gong.
Mikki Cottet, Appellate Staff, Civil Division, United States Department of Justice, Washington, DC, argued for defendant-appellee. Also represented by Jeanne Davidson, Joseph H. Hunt, Patricia M. McCarthy.
Andrew Thomas Schutz, Grunfeld, Desiderio, Lebowitz, Silverman & Klestadt LLP, Washington, DC, argued for defendants-cross-appellants. Also argued by Ned H. Marshak, New York, NY. Also represented by Max Fred Schutzman, New York, NY; Kavita Mohan, Washington, DC.
Before Newman, O’Malley, and Taranto, Circuit Judges.
Taranto, Circuit Judge. The United States Department of Commerce found that certain foreign producers and exporters were dumping certain products into the United States market, and it imposed a small antidumping duty on their imports. A domestic company argues that Commerce should have imposed a higher duty. The foreign producers and exporters argue that Commerce made methodological errors, the correction of which would reduce any dumping margin to a de minimis level, so that no duty would be imposed. We reject the domestic firm’s challenge. We partly reject the foreign firms’ challenge, and we remand to secure further explanation from Commerce about one issue.
I
Based on a petition from appellant Mid Continent Steel & Wire, Inc., Commerce initiated an antidumping duty investigation into steel nail products from Taiwan and certain other places. Certain Steel Nails from India, the Republic of Korea, Malaysia, the Sultanate of Oman, Taiwan, the Republic of Turkey, and the Socialist Republic of Vietnam , 79 Fed. Reg. 36,019 (Dep’t of Commerce June 25, 2014). Commerce separated the Taiwanese investigation into its own proceeding and named Taiwanese exporter PT Enterprise Inc. and its affiliated nail producer Pro-Team Coil Nail Enterprise Inc. as mandatory respondents. See Certain Steel Nails from Taiwan: Negative Preliminary Determination of Sales at Less Than Fair Value and Postponement of Final Determination , 79 Fed. Reg. 78,053, 78,054 (Dep’t of Commerce Dec. 29, 2014) (Preliminary Determination). Those firms are the cross-appellants before us, along with other Taiwanese producers of nails. We hereafter use "PT" to refer sometimes to the cross-appellants collectively, sometimes just to PT Enterprise and Pro-Team.
The statute directs Commerce to determine whether the merchandise at issue is being sold or is likely to be sold in the United States "at less than its fair value," 19 U.S.C. § 1673, which the statute identifies as "dumping," id. , § 1677(34) (defining "dumping" to mean "the sale or likely sale of goods at less than fair value"). To make the required determination, Commerce must assess the difference between the "normal value" of the goods at issue (reflecting the home-market value) and the "export price or constructed export price" of those goods (reflecting the price at which they are sold into the United States). See id. , § 1677b(a) (stating that the determination of the existence of sales "at less than fair value" is to be based on a comparison of "the export price or constructed export price and normal value"); id. , § 1677a (addressing "export price" and "constructed export price"); id. , § 1677b (addressing "normal value"). That difference is the "dumping margin." Id. , § 1677(35)(A) (defining "dumping margin"). If Commerce finds the specified less-than-fair-value sales, and the International Trade Commission makes certain findings about effects on domestic industry, "there shall be imposed upon such merchandise an antidumping duty, in addition to any other duty imposed, in an amount equal to the amount by which the normal value exceeds the export price (or the constructed export price) for the merchandise," id. , § 1673, i.e. , in the amount of the dumping margin. Addressing the fact that a foreign producer or exporter often makes many sales, the statute provides certain rules and authorizations that govern Commerce’s required determinations. Id. , § 1677f-1. It defines "weighted average dumping margin" to mean "the percentage determined by dividing the aggregate dumping margins determined for a specific exporter or producer by the aggregate export prices and constructed export prices of such exporter or producer." Id. , § 1677(35)(B). The statute provides, as a general rule, that Commerce must "determine whether the subject merchandise is being sold in the United States at less than fair value" by "comparing the weighted average of the normal values to the weighted average of the export prices (and constructed export prices) for comparable merchandise" or by making the value/price comparison for each individual transaction. Id. , § 1677f-1(d)(1)(A)(i), (ii). But the statute also directs Commerce to disregard weighted average dumping margins if they are de minimis, id. , § 1673b(b)(3); and of relevance here, it provides authority to Commerce to compare average values (on the foreign side) to individual export prices or constructed export prices (on the U.S. side) in specified circumstances involving disparities among the U.S. side prices for the foreign exporter or producer. Id. , § 1677f-1(d)(1)(B).
Certain aspects of the method adopted by Commerce for calculating the dumping margin in the present matter are unchallenged. On the U.S. side of the required comparison, Commerce used the export price, rather than a constructed export price. On the foreign side, Commerce determined the Taiwanese normal value by determining a "constructed value," which required determinations about, among other things, amounts PT paid for various inputs. 19 U.S.C. § 1677b(a)(4), (e).
Although those basic choices are not in dispute, there is a dispute about how Commerce carried out its "constructed value" calculation. Among the inputs PT purchased were services from many "toll" manufacturers (or "tollers")—firms that provide limited manufacturing services using materials or other contributions supplied or owned by its customers. Mid Continent has contended that certain of PT’s tollers should be excluded from this input calculation because those tollers were affiliated with PT. The evident concern with a "transaction between affiliated entities" is that it might not "adequately represent the true amount," SKF USA Inc. v. United States , 630 F.3d 1365, 1372 (Fed. Cir. 2011) —here, that PT’s payments for tolling to an affiliate might be artificially low, with the consequence that the constructed value might be too low, thus shrinking the gap between the constructed value and the U.S. price and, in turn, reducing the dumping margin and antidumping duty.
In its Preliminary Determination, Commerce rejected Mid Continent’s affiliation claim as to a number of PT’s tollers and found no dumping. Preliminary Determination, 79 Fed. Reg. at 78,054–78,055 ; Decision Memorandum for the Preliminary Determination in the Antidumping Duty Investigation of Certain Steel Nails from Taiwan , 79 ITADOC 78053 (issued Dec. 17, 2014) (Preliminary Decision Mem.). Commerce then conducted its full investigation and analysis, including verification of key factual submissions.
In its Final Determination, Commerce continued to find non-affiliation of certain PT tollers, contrary to Mid Continent’s contentions. See Certain Steel Nails from Taiwan: Final Determination of Sales at Less Than Fair Value , 80 Fed. Reg. 28,959, 28,960 –62 (Dep’t of Commerce May 20, 2015) (Final Determination); Issues and Decision Memorandum for the Affirmative Final Determination in the Less than Fair Value Investigation of Certain Nails from Taiwan , 80 ITADOC 28959, at 47–53 (issued May 13, 2015) (Issues and Decision Mem.). But, based on some adjustments of earlier information, Commerce now found a positive dumping margin above (though not far above) the level that Commerce deems de minimis, and it imposed a duty in that amount. See Final Determination, 80 Fed. Reg. at 28,961. In reaching that result, Commerce found so-called "differential pricing" by PT among its export prices, and it rejected certain of PT’s challenges to the method for analyzing differential pricing that Commerce had set out in its preliminary decision memorandum. Issues and Decision Mem. at 15–31.
Mid Continent filed an action in the Court of International Trade (Trade Court), seeking a higher duty by challenging Commerce’s finding of no affiliation between PT and certain of its tollers. PT also sued in the Trade Court, seeking a lower or zero duty by challenging certain aspects of Commerce’s calculation methodology. The Trade Court sustained the relevant Commerce conclusions and remanded on an unrelated issue. Mid Continent Steel & Wire, Inc. v. United States , 219 F. Supp. 3d 1326 (Ct. Int’l Trade 2017). The Trade Court subsequently entered judgment sustaining Commerce’s Final Determination after the remand was complete. J.A. 72–73.
Both Mid Continent and PT have timely appealed to this court, and we have jurisdiction pursuant to 28 U.S.C. §§ 1295(a)(5) and 2645(c). Mid Continent appeals Commerce’s determination of non-affiliation between PT and certain of its tollers. PT cross-appeals three aspects of Commerce’s method of calculating the dumping margin.
We review Commerce’s decision using the same standard of review applied by the Trade Court, while carefully considering that court’s analysis. Diamond Sawblades Mfrs. Coal. v. United States , 866 F.3d 1304, 1310 (Fed. Cir. 2017). We decide legal issues de novo and uphold factual determinations if they are supported by substantial evidence. 19 U.S.C. § 1516a(b)(1)(B)(i) ; see Diamond Sawblades , 866 F.3d at 1310 ; Dupont Teijin Films USA, LP v. United States , 407 F.3d 1211, 1215 (Fed. Cir. 2005). "A finding is supported by substantial evidence if a reasonable mind might accept the evidence to support the finding." Nobel Biocare Services AG v. Instradent USA, Inc. , 903 F.3d 1365, 1374 (Fed. Cir. 2018) ; see Consol. Edison Co. v. NLRB , 305 U.S. 197, 229, 59 S.Ct. 206, 83 L.Ed. 126 (1938). In carrying out its statutorily assigned tasks, Commerce has discretion to make reasonable choices within statutory constraints. See , e.g. , Nucor Corp. v. United States , 927 F.3d 1243, 1248–49 (Fed. Cir. 2019) ; Apex Frozen Foods Private Ltd. v. United States , 862 F.3d 1322, 1329 (Fed. Cir. 2017) ; see also Utility Air Regulatory Group v. EPA , 573 U.S. 302, 321, 134 S.Ct. 2427, 189 L.Ed.2d 372 (2014) ; City of Arlington v. FCC , 569 U.S. 290, 297, 133 S.Ct. 1863, 185 L.Ed.2d 941 (2013). But Commerce must provide an explanation that is adequate to enable the court to determine whether the choices are in fact reasonable, including as to calculation methodologies. See CS Wind Vietnam Co., Ltd. v. United States , 832 F.3d 1367, 1376–77 (Fed. Cir. 2016).
We address Mid-Continent’s challenge and PT’s challenges in turn. We reject Mid Continent’s challenge. We reject two of PT’s challenges, but we vacate the decision of the Trade Court on the third challenge and remand for that court to remand to Commerce for further explanation.
II
The statute directs Commerce, when determining a constructed value of the merchandise at issue, to examine certain transactions to gauge the costs of various inputs. 19 U.S.C. § 1677b(e). The statute provides, however, that "[a] transaction directly or indirectly between affiliated persons may be disregarded" in certain circumstances—specifically, "if, in the case of any element of value required to be considered, the amount representing that element does not fairly reflect the amount usually reflected in sales of merchandise under consideration in the market under consideration." Id. , § 1677b(f)(2). Mid Continent’s challenge focuses on the threshold question of what it means for a transaction to be between "affiliated persons."
Congress has provided a definition:
The following persons shall be considered to be "affiliated" or "affiliated persons":
(A) Members of a family, including brothers and sisters (whether by the whole or half blood), spouse, ancestors, and lineal descendants.
(B) Any officer or director of an organization and such organization.
(C) Partners.
(D) Employer and employee.
(E) Any person directly or indirectly owning, controlling, or holding with power to vote, 5 percent or more of the outstanding voting stock or shares of any organization and such organization.
(F) Two or more persons directly or indirectly controlling, controlled by, or under common control with, any person.
(G) Any person who controls any other person and such other person.
For purposes of this paragraph, a person shall be considered to control another person if the person is legally or operationally in a position to exercise restraint or direction over the other person.
Id. , § 1677(33) (emphasis added to the "control" portions that are at issue here). Commerce, for its part, has adopted an implementing rule for antidumping duty investigations:
"Affiliated persons" and "affiliated parties" have the same meaning as in [§ 1677(33) ]. In determining whether control over another person exists, within the meaning of [§ 1677(33) ], the Secretary will consider the following factors, among others: Corporate or family groupings; franchise or joint venture agreements; debt financing; and close supplier relationships. The Secretary will not find that control exists on the basis of these factors unless the relationship has the potential to impact decisions concerning the production, pricing, or cost of the subject merchandise or foreign like product. The Secretary will consider the temporal aspect of a relationship in determining whether control exists; normally, temporary circumstances will not suffice as evidence of control.
19 C.F.R. § 351.102(b)(3) (emphasis added).
In the present matter, Mid-Continent contends that Commerce mistakenly rejected its argument that PT was affiliated with certain tolling companies. We reject the contention.
Commerce made detailed findings to support its overall finding that PT lacked the ability to exercise control over those tollers—which, therefore, were not affiliated with PT. Commerce prominently found that there are many companies other than PT to whom the relevant tollers "could provide their services" if exploited by PT. Issues & Decision Mem. at 51, 52; see Mid Continent , 219 F. Supp. 3d at 1334 n.11. Commerce also found that many tollers sold to others as well as PT and had existed before selling to PT, and that the tollers that had all or most of their sales to PT were profitable, indicating lack of exploitation. Issues & Decision Mem. at 51–52; Mid Continent , 219 F. Supp. 3d at 1334 & n.12. Commerce further found that PT and the tollers had no stock ownership in each other, no shared officers or managers, and no other relationships, such as common familial ownership, that might suggest ability to control; no toller had a long-term contract with PT or a debt-financing agreement with PT. Issues & Decision Mem. at 50–52; Mid Continent , 219 F. Supp. 3d at 1334.
We see no lack of substantial evidence to support those findings. And those findings permit a reasonable mind to find, and so are sufficient to support the overall finding, that PT, while working closely with the tollers at issue, did not have the ability to control them. To the extent that Mid Continent argues that the evidence could support a contrary finding, it misidentifies the relevant question, which is only whether the evidence supports the finding that Commerce made. "[T]he possibility of drawing two inconsistent conclusions from the evidence does not prevent an administrative agency’s finding from being supported by substantial evidence." Consolo v. Federal Maritime Comm’n , 383 U.S. 607, 620, 86 S.Ct. 1018, 16 L.Ed.2d 131 (1966) ; see American Textile Mfrs. Inst., Inc. v. Donovan , 452 U.S. 490, 523, 101 S.Ct. 2478, 69 L.Ed.2d 185 (1981) ; Nobel Biocare Services , 903 F.3d at 1375.
Mid Continent suggests that Commerce used the wrong standard in answering the affiliation question. Specifically, it suggests that Commerce did not determine whether PT had the ability to control the tollers at issue, but only whether PT had in fact exercised control. We disagree. That suggestion is contrary to what Commerce did: notably, it is contrary to Commerce’s reliance on the number of other buyers available to the tollers at issue, a fact that directly bears on ability to control. The suggestion also is contrary to what Commerce said it was deciding—namely, whether PT "was able to exert restraint or direction over" the tollers at issue, whether the tollers became " ‘reliant upon’ " PT, whether PT was " ‘in a position to exercise restraint or direction,’ " whether PT had "the ability to control" the tollers, and whether there was "dependence on" PT by the tollers. Issues & Decision Mem. at 50–52 (emphases added). Because evidence of actual control is highly relevant to whether the ability to control exists, Commerce’s consideration of such evidence does not show that Commerce failed to apply an ability-to-control standard.
Finally, we see no merit in Mid Continent’s suggestion that Commerce made an affiliation determination for the tollers only on a collective basis and failed to make an individual determination for each toller. As Mid Continent itself has stated, Commerce’s decisions themselves provide no support for such a conclusion. A failure to analyze each element of a group cannot be inferred from the fact that the ultimate summary employs generalizations to recount shared facts where appropriate. And we see no basis for finding a concession to the contrary in a single remark by government counsel before the Trade Court, a remark easily understood as simply indicating that Commerce considered all the tollers, not just some.
We therefore affirm Commerce’s determination that PT and the toll manufacturers at issue in the appeal are not affiliated.
III
In its cross-appeal, PT renews challenges that it pressed before Commerce and the Trade Court to the methods Commerce used in calculating the dumping margin. The challenges now at issue focus on Commerce’s treatment of the U.S. side of the comparison required for calculating the dumping margin—that is, on Commerce’s treatment of export prices. Specifically, PT challenges three aspects of how Commerce implemented its statutory authority to address situations in which there are certain disparities among the export prices charged for the merchandise at issue.
A
Congress recognized that the comparison required by the statute typically requires consideration of multiple transactions, necessitating choices about treating them in the aggregate, one at a time, or in some other way. See 19 U.S.C. § 1677f-1. Congress provided that, "[i]n general," Commerce "shall" proceed either "(i) by comparing the weighted average of the normal values to the weighted average of the export prices ... for comparable merchandise"—the "average-to-average" (A-to-A) method—or "(ii) by comparing the normal values of individual transactions to the export prices ... of individual transactions for comparable merchandise"—the "transaction-to-transaction" (T-to-T) method. Id. , § 1677f-1(d)(1)(A). But Congress also provided an "exception":
[Commerce] may determine whether the subject merchandise is being sold in the United States at less than fair value by comparing the weighted average of the normal values to the export prices (or constructed export prices) of individual transactions for comparable merchandise, if—
(i) there is a pattern of export prices (or constructed export prices) for comparable merchandise that differ significantly among purchasers, regions, or periods of time, and
(ii) [Commerce] explains why such differences cannot be taken into account using a method described in paragraph (1)(A)(i) or (ii)[, i.e. , the A-to-A or T-to-T methods].
Id. , § 1677f-1(d)(1)(B). Thus, Commerce "may" use a mixed method—the "average-to-transaction" (A-to-T) method—if the two conditions are met. See Apex , 862 F.3d at 1326–27, 1334.
A regulation adopted by Commerce gives the identified names to the three methods. 19 C.F.R. § 351.414(b). It also states that Commerce will use the A-to-A method "unless [it] determines another method is appropriate in a particular case," id. , § 351.414(c)(1), and that it will use the T-to-T method "only in unusual situations, such as when there are very few sales of subject merchandise and the merchandise sold in each market is identical or very similar or is custom-made," id. , § 351.414(c)(2). The regulation does not specify details for implementing the A-to-T method except in a respect not at issue here (concerning use of "the contemporaneous month" for normal-value averaging). Id. , § 351.414(e), (f).
B
In the present matter, Commerce adopted the A-to-T method by determining that the statutory conditions for its use were met. Specifically, it determined that there was a "pattern of export prices" for "comparable merchandise that differ[ed] significantly among purchasers, regions, or periods of time"—i.e. , that there was so-called "differential pricing"—and that those differences could not be "taken into account" by using the other two methods. See Issues and Decision Mem. at 19–20; J.A. 2179. In so finding, Commerce relied on a method using the "Cohen’s d test" borrowed from statistics literature, which it had only recently adopted. See Issues & Decision Mem. at 25 & n.111 (citing Issues and Decision Memorandum for the Final Determination of the Antidumping Duty Investigation of Xanthan Gum from the People’s Republic of China , 78 ITADOC 33351 (June 4, 2013), and that ruling’s reliance on Robert Coe, It’s the Effect Size, Stupid: What Effect Size Is and Why It Is Important , Paper presented at the Annual Conference of the British Educ. Research Ass’n (Sept. 2002), www.leeds.ac.uk/educol/documents/00002182.htm)). Commerce explained aspects of its overall method in the Preliminary Decision Mem. at 10–12, and it elaborated on certain details in the Issues & Decision Mem. at 15–31. See Mid Continent , 219 F. Supp. 3d at 1337–44.
To determine whether the required "pattern" existed, Commerce applied what it identified as "a generally recognized" statistical test called the "Cohen’s d test" for testing differences among subsets within an overall set of data—here, for testing subsets of PT’s export prices defined by purchasers, regions, or periods of time (each a "test group") by comparing them to the export prices outside the test group (the "comparison group") to see if they differed significantly. See Preliminary Decision Mem. at 11. It calculated a ratio called a "Cohen’s d coefficient" for each comparison. The numerator of that ratio is the difference between the mean of the test group’s prices and the mean of the comparison group’s prices; the denominator is a measure of dispersion when pooling the test and comparison groups—here, Commerce used the "pooled" variance (the square of the standard deviation). If the mean-to-mean difference in the numerator was at least four fifths of the pooled figure, i.e. , the Cohen’s d coefficient was 0.8 or greater, Commerce deemed the test group’s pricing significantly different from the comparison group’s pricing for purposes of meeting the "pattern" condition. Id. at 11. Of significance to the issues presented on appeal, in arriving at the pooled dispersion figure used as the denominator of the ratio forming the d coefficient, Commerce used a "simple average" of the variance of the test group and the variance of the comparison group, disregarding the comparative sizes of the two groups; it did not weight the variances being pooled by the volume (or value or other characteristic) of goods sold within each group. See Issues & Decision Mem. at 28–29; Mid Continent , 219 F. Supp. 3d at 1338–39.
Based on its calculation of Cohen’s d coefficients for sales by customer, region, or time segment, Commerce proceeded to decide whether it would use an A-to-T comparison rather than an A-to-A comparison, for some or all of the PT’s sales, as it "may" do under the statute if the specified pattern is present and an A-to-A (or T-to-T) comparison cannot account for that pattern. To make this decision, Commerce applied a "ratio test"—examining the percent of sales (by value) that have coefficients of at least 0.8—combined with a "meaningful differences" test.
Thus, Commerce said that, in cases where the percent is less than 33, it would use the A-to-A method in toto, whereas in cases where the percent is above 66, Commerce would use the A-to-T method in toto. But if, as in this matter, the percent is from 33 to 66, Commerce said, its approach was the following: (a) use the A-to-T method to calculate a dumping margin for the sales having coefficients of at least 0.8, but do so by zeroing out the non-dumped sales so that they do not offset dumping margins from the dumped sales; (b) use the A-to-A method to calculate a dumping margin for the other sales, with no zeroing of non-dumped sales; and (c) arrive at an overall weighted-average dumping margin for all of the respondent’s sales by combining the two results in a way that zeroes out negative dumping margins among the A-to-A group so that they do not offset positive margins among the A-to-T group. See Issues & Decision Mem. at 26. Finally, Commerce stated that it uses the resulting aggregate figure if there is a meaningful difference between doing so and simply using the A-to-A method, which, Commerce said, includes (though is not limited to) a situation in which the former moves the dumping margin above the de minimis threshold, as it did here. Preliminary Decision Mem. at 12. See generally Mid Continent , 219 F. Supp. 3d at 1343–44.
The parties have not provided us with or pointed to a clear statement of the specific arithmetic steps that Commerce used for the combination. Such a clear statement is commonly necessary to enable us to understand, and hence to review, an action taken by Commerce; when there is no such clear statement, a remand for further explanation becomes more likely. In this matter, however, it is undisputed that Commerce used some kind of zeroing of the negative-margin sales in the group to which it applied an A-to-A method when taking the next step of arriving at an overall dumping margin. That is enough for review of the limited (statutory) challenge presented to us concerning this stage of Commerce’s calculation.
--------
C
We now address PT’s three challenges.
1
PT first challenges an aspect of Commerce’s method at the aggregation stage—specifically, its zeroing of negative dumping margins for sales in the A-to-A group when arriving at the overall weighted-average dumping margin. PT states that the zeroing at that stage of the calculation applies, within the broad class of merchandise at issue, to at least some control-number subcategories (CONNUMs) that contain no sales having a d coefficient of at least 0.8. PT objects to such zeroing.
Commerce justified such zeroing on the ground that it served to avoid "reduc[ing] or completely negat[ing] the results of the A-to-T method" and "continu[ing] to mask dumping." Issues & Decision Mem. at 26–27. The Trade Court upheld that choice as reasonable. Mid Continent , 219 F. Supp. 3d at 1344 ("Commerce’s decision to effectuate the statute by applying the mixed methodology, without offsets during the aggregation stage, to both preserve the masked dumping uncovered with A-to-T and achieve a proportionate remedy is reasonable."). In this court, PT expressly disclaims any challenge to that rationale as failing "a reasonable or arbitrary/capricious standard." PT Br. at 30. PT’s sole contention is that this zeroing is "contrary to the statutory language." Id. We reject that contention.
The statute prescribes preconditions for Commerce to use an A-to-T method in making the less-than-fair-value determination, but it does not specify the implementation choice at issue here. It says that, if the preconditions are met, Commerce "may determine whether the subject merchandise is being sold in the United States at less than fair value by comparing the weighted average of the normal values to the export prices (or constructed export prices) of individual transactions for comparable merchandise." 19 U.S.C. § 1677f-1(d)(1)(B). That language does not tell Commerce that, if it finds the preconditions present, it must use an A-to-T comparison for all of the exporter’s/producer’s transactions or, if not, how it may relate the results for transactions to which it applies the A-to-T method to transactions to which it applies another (here, A-to-A) method. The statute leaves a choice for Commerce to make reasonably.
Our precedents support this conclusion. In Apex , we concluded that § 1677f-1(d)(1)(B) "defines the preconditions for applying the A-T methodology, but it does not limit in any way the application of the A-T methodology, should the preconditions be met." 862 F.3d at 1334. And we upheld application of the A-to-T method even to sales outside the pattern establishing the statutory "pattern" precondition. Id. at 1334–36. In an earlier case, we explained that "[o]ur case law has repeatedly examined the antidumping statute and found it to be ‘silent or ambiguous’ as to zeroing methodology," and we held that § 1677f-1(d)(1)(B) does not require zeroing. U.S. Steel Corp. v. United States , 621 F.3d 1351, 1361 (Fed. Cir. 2010). We see no better basis for a statutory answer to the specific question presented here. We therefore reject the only challenge PT makes to Commerce’s use of zeroing in the identified part of its calculation.
2
PT next challenges Commerce’s reliance on a d ratio of at least 0.8 as a rigid measure of significance of the difference measured by the Cohen’s d test. PT argues that Commerce must instead use a "flexible" threshold to deem insignificant certain above-0.8 differences that reflect price differences that are small in dollar value—such as, PT says, one difference in this matter that is less than half a penny per kilogram (the unit for pricing the nails at issue). This is a challenge to the reasonableness of Commerce’s choice of one part of the overall analysis of differential pricing, explained by Commerce in the Issues & Decision Mem. at 24–26.
The Trade Court described Commerce’s rationale for adhering to the 0.8 line and explained why that rationale is reasonable. Mid Continent , 219 F. Supp. 3d at 1337–40. In particular, Commerce reasoned that even a small absolute difference in the means of the two groups can be significant (for the present statutory purpose) if there is a small enough dispersion of prices within the overall pool as measured by a proper pooled variance or standard deviation; the 0.8 standard is "widely adopted" as part of a "commonly used measure" of the difference relative to such overall price dispersion; and it is reasonable to adopt that measure where there is no better, objective measure of effect size. Issues & Decision Mem. at 25–26 (internal quotation marks omitted). We agree with the Trade Court that this rationale adequately supports Commerce’s exercise of the wide discretion left to it under 19 U.S.C. § 1677f-1(d)(1)(B). We therefore reject PT’s challenge.
3
Finally, PT challenges Commerce’s use of a simple average, rather than a weighted average, to calculate the pooled variance used in the Cohen’s d calculation. On this issue, we agree with PT that Commerce’s explanation is wanting. And we remand for further explanation.
Commerce adopted the Cohen’s d test as a "generally recognized" test, Preliminary Decision Mem. at 11, and it later cited the 2002 Coe article as a source for that test, Issues & Decision Mem. at 25 n.111. In this court, the government reiterates, citing Coe, that "the Cohen’s d coefficient is a prominent and widely-accepted statistical measure that has been developed to evaluate practical significance." U.S. Br. at 55–56 (footnote at end of sentence relying on Coe). Before Commerce, PT relied directly on Coe, which clearly shows weighted averaging in calculating the pooled standard deviation where test and comparison groups are different in quantity and non-overlapping. J.A. 2097 (giving formula taken from Coe, with variable names different); Coe, supra (question 7 and equation 4). There has been no dispute in this matter that the argument about using simple averages or weighted averages is the same whether the pooled figure is a pooled standard deviation or a pooled variance.
Yet Commerce rejected use of the weighted average. Issues and Decision Mem. at 28–29. It said that the statute does not answer the question, id. at 28, and PT does not dispute that premise. It then stated that simple averaging was "the best way to accomplish" the goal of having a "reasonable approach that affords predictability." Id. After making the essentially definitional point that with simple averaging "the respondent’s pricing practice to each group will be weighted equally," Commerce added that with simple averaging "the magnitude of the sales to one group will not skew the outcome," without elaborating on what "skew[ing]" meant. Id. at 28–29. Commerce rejected PT’s argument that simple averaging "distorts the results of the Cohen’s d test" as resting on an assumption that test groups are smaller and have lower variances than comparison groups. Id. at 29. Finally, Commerce reasoned that "by weighting the variances of the test and comparison groups by the number of observations, this in itself would open up the analysis to distortion since how the U.S. sales data is reported, i.e. , how each observation is determined, is completely under the control of the respondent." Id.
Although the Trade Court upheld this explanation, Mid Continent , 219 F. Supp. 3d at 1340–43, we do not find the explanation sufficient, for several reasons. First, Commerce said that it was simply using a widely accepted statistical test; yet it did not acknowledge that the only cited literature source for the relevant aspect of the test itself calls for the use of weighted averages. Second, Commerce’s language of "skew[ing]" is a mere conclusion where, as here, it is unaccompanied by an explanation of why the right result, consistent with the relevant statutory purpose, should be different. Third, although Commerce determined that PT’s charge that simple averaging "distorts" the outcome rests on an assumption that is not always true, that determination is both unsupported and, in any event, not itself an explanation of why weighted averaging is actually distortive in a relevant sense or, more affirmatively, why simple averaging is preferable. Fourth, Commerce asserted that simple averaging was more "predictab[le]" than weighted averaging, but the only expressed reason seems to be a concern about manipulation in how sales are reported, and that concern seems to assume that weighted averaging must be done by counting numbers of transactions, rather than quantity sold within transactions. The Trade Court recognized that both types of weighted averaging were legitimately in play in this matter. Mid Continent , 219 F. Supp. 3d at 1342. Yet Commerce gave no explanation of why weighting should not be done by "quantity" of units sold—here measured in kilograms—and what if any manipulation/predictability concern there would be if that path were followed. We note that, in this court, PT set forth a numerical illustration of weighted averaging by kilogram , reflecting the apparently undisputed fact that quantity is measured in kilograms, and price is stated in per-kilogram terms, when the nails at issue are sold.
On the record before us, we cannot conclude that Commerce’s methodology was a reasonable exercise of its agency discretion in light of the statutory constraints and policies. We therefore vacate Commerce’s determination and remand for further proceedings on this issue. In so doing, we are not holding that Commerce cannot justify the choice it has made. What is needed is a more thorough consideration by Commerce of the issue than is given in the current record and, upon such consideration, a clear explanation of the choices that Commerce makes on the arguments and evidence presented to it. See CS Wind Vietnam , 832 F.3d at 1380–81 ("Nor are we ruling that Commerce’s current result is incorrect—that it cannot be properly justified. We are remanding because we conclude that Commerce has not explained its determination sufficiently to allow us to conduct the judicial review to which CS Wind is entitled to ensure that the agency’s exercise of power adheres to the authorizing law and respects the record evidence.").
IV
For the reasons we have stated, we affirm the Trade Court’s judgment as to the issues in Mid Continent’s appeal. We also affirm the Trade Court’s judgment as to the issues raised in PT’s cross-appeal, with one exception: we vacate the Trade Court’s judgment upholding Commerce’s choice of a simple averaging in calculating the pooled variance, and we remand the case to that court for it to remand to Commerce for further proceedings on that issue.
The parties shall bear their own costs.