MID CONTINENT STEEL & WIRE, INC. ET AL., Plaintiff and Consolidated Plaintiffs, v. UNITED STATES, Defendant, and PT ENTERPRISE INC. ET AL., Defendant-Intervenors and Consolidated Defendant-Intervenors.
Consol. Court No. 15-00213
UNITED STATES COURT OF INTERNATIONAL TRADE
January 8, 2021
Claire R. Kelly, Judge
Slip Op. 21-4
OPINION AND ORDER
[Sustaining Commerce‘s decision to use a simple average to calculate a pooled standard deviation as part of its differential pricing analysis in its antidumping duty investigation of certain steel nails from Taiwan.]
Dated: January 8, 2021
Adam H. Gordon and Ping Gong, The Bristol Group PLLC of Washington, DC, for plaintiff and consolidated defendant-intervenor Mid Continent Steel & Wire, Inc.
Bruce M. Mitchell, Ned H. Marshak, Andrew T. Schutz, and Dharmendra Choudhary, Grunfeld Desiderio Lebowitz Silverman & Klestadt LLP of Washington, DC, and New York, NY, for consolidated plaintiffs and defendant-intervenors PT Enterprise Inc., Pro-Team Coil Nail Enterprise Inc., Unicatch Industrial Co., Ltd., WTA International Co., Ltd., Zon Mon Co., Ltd., Hor Liang Industrial Corp., President Industrial Inc., and Liang Chyuan Industrial Co., Ltd.
Mikki Cottet, Senior Trial Counsel, Commercial Litigation Branch, Civil Division, U.S. Department of Justice, of Washington, DC, for defendant. Also on the brief were Jeffrey Bossert Clark, Acting Assistant Attorney General, Jeanne E. Davidson, Director, and Patricia M. McCarthy, Assistant Director. Of counsel on the brief was Vania Wang, Attorney, Office of the Chief Counsel for Trade Enforcement & Compliance, U.S. Department of Commerce of Washington, DC.
The court remanded to Commerce for further proceedings in conformity with Mid Continent III, see Order, where the Court of Appeals found wanting, for several reasons, Commerce‘s explanation of its use of a simple average when determining the pooled standard deviation in its Cohen‘s d test. See Mid Continent III, 940 F.3d at 673-75. The Court of Appeals observed that: Commerce failed to address that the relevant part of the literature Commerce itself cites, calls for the use of weighted averages; Commerce‘s statement that simple averaging will not “skew” the outcome of its analysis, was conclusory; Commerce neither supported its dismissal of PT‘s charge that simple averages would distort the outcome of its Cohen‘s d test, nor explained why simple averaging was preferable; and, Commerce appeared to assume that weighted averaging must be done by counting the number of transactions, as opposed to quantities, sold within every transaction. See id.
On remand, Commerce reconsiders whether its “calculation of the pooled standard deviation based on a simple average of the variances determined for the test and comparison groups was appropriate,” and finds that it is. Second Remand Results at 4-17. Commerce explains that a simple average of the variances for each group accurately represents pricing behaviors within each group because it is the group pricing that matters, and not the individual pricing amongst all sales, for the purposes of its analysis. See id. at 15-17, 35-36. Thus, using a simple average of the variances within each group in the pooled standard deviation, as the denominator in the Cohen‘s d analysis, will not mask possible targeted dumping that only exists within one group. See id. Conversely, a weighted average, based either on the number of transactions, sales volume or value, would skew the results towards the group with greater transactions or sales volume or value (and thus dilute the results from the other group). See id. at 2, 15-17. Commerce, therefore explains that a simple average is preferable because the purpose of the analysis is to identify masked dumping within the test group. See id. at 8, 14-17. For the following reasons, the court sustains Commerce‘s decision to use a simple average to calculate the pooled standard deviation.
BACKGROUND
On June 25, 2014, in response to a petition filed by Mid Continent, Commerce initiated an ADD investigation of certain steel nails from six countries, including Taiwan. See 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 Commerce June 25, 2014) (initiation of less-than-fair-value investigations). Commerce selected Taiwanese exporters PT Enterprise Inc. et al.‘s (“PT“)1 and its affiliated producer, Pro-Team, and Quick Advance, Inc. and its affiliated producer, Ko‘s Nails Inc., as mandatory respondents for the investigation. See Certain Steel Nails from Taiwan, 79 Fed. Reg. 78,053, 78,054 (Dep‘t Commerce Dec. 29, 2014) (preliminary determination of sales at less than fair value, postponement of final determination) (“Prelim. Results“); Decision Memorandum for the Preliminary Determination in the Antidumping Duty Investigation of
(Dec. 17, 2014) (“Prelim. Decision Memo“);2 see also Section 777A of the of the Tariff Act of 1930, as amended
On December 29, 2014, Commerce issued its negative preliminary determination. See Prelim. Results, 79 Fed. Reg. at 78,053; see also Prelim. Decision Memo at 1. Commerce applied its differential pricing analysis and determined that, although 41.73 percent of PT‘s U.S. sales passed the Cohen‘s d test, a meaningful difference did not exist in the dumping margins that would result from using the standard average-to-average (“A-to-A“) methodology and the alternate mixed methodology. See Prelim. Decision Memo at 12. Commerce accordingly applied the standard A-to-A methodology to all of PT‘s sales, and preliminarily determined that respondents’ steel nails from Taiwan “are not being, or are not likely to be, sold in the United States at less than fair value.” Id. at 1.4 Commerce preliminarily assigned
PT a weighted-average dumping margin of 0.00 percent. See Prelim. Results, 79 Fed. Reg. at 78,054.
On May 20, 2015, Commerce issued its final determination. See Certain Steel Nails from Taiwan, 80 Fed. Reg. 28,959 (Dep‘t Commerce May 20, 2015) (final determination of sales at less than fair value) (“Final Results“) and accompanying Issues and Decision Memorandum for the Affirmative Final Determination in the Less than Fair Value Investigation of Certain Steel Nails from Taiwan, May 13, 2015, ECF No. 17 (“Final Decision Memo“). As an initial matter, Commerce determined that it made ministerial errors which impacted the dumping margins it calculated in the Prelim. Results. See Final Results, 80 Fed. Reg. at 28,960. Additionally, Commerce made other adjustments, including changes to the margin calculation. See Final Decision Memo at 1; Final Results, 80
On March 23, 2017, the U.S. Court of International Trade (“CIT“) sustained Commerce‘s determination including its decision to use a simple average to calculate the pooled standard deviation. See Mid Continent I, 41 CIT at __, 219 F. Supp. 3d at 1340-43. On October 3, 2019, the Court of Appeals vacated and remanded in part the CIT‘s judgment sustaining Commerce‘s decision. See Mid Continent III, 940 F.3d at 675. The Court of Appeals disagreed with PT‘s challenge to Commerce‘s use of zeroing and Commerce‘s requirement to have a Cohen‘s d coefficient of at least 0.8, and thus sustained this court‘s ruling upholding Commerce on those issues. See Mid Continent III, 940 F.3d at 672. However, the Court of Appeals instructed the CIT to remand to Commerce to provide additional reasoning in support of its decision to use a simple average to calculate the pooled standard deviation. See id. at 673-75. On December 3, 2019, the CIT remanded to Commerce to provide further explanation in accordance with the Court of Appeals’ directive. See Order, Dec. 3, 2019, ECF No. 132.
On March 3, 2020, Commerce released its draft remand redetermination and invited the parties to comment. See Draft Results of Redetermination Pursuant to Court Order [Mid Continent III], PRRs 3-5, bar codes 3949998-01-03 (Mar. 3, 2020).6
JURISDICTION AND STANDARD OF REVIEW
The court has jurisdiction pursuant to
DISCUSSION
In the results of its second remand redetermination, Commerce again decides to calculate a simple average of the variances in its Cohen‘s d analysis, because in the present context it is more accurate to assign the test and comparison groups equal weight. See Second Remand Results. PT contends that a simple average is inconsistent with the statute and unreasonable as applied to this case, while the use of a weighted average is consistent with the statute, reasonable and produces a more accurate result. See Pl. [PT]‘s Cmts. on Remand Results, July 28, 2020, ECF No. 150 (“PT‘s Br.“). For the following reasons, Commerce‘s decision is sustained.
In its second remand redetermination, Commerce again examines whether PT‘s sales reflect a “pattern of export prices” for “comparable merchandise that differ[ed] significantly among purchasers, regions, or periods of time” and whether the differences can be “taken into account” using an A-to-A or A-to-T methodology. See Second Remand Results at 5. Commerce determines whether there is a pattern
In the Second Remand Results, Commerce further explains that it uses a simple average, rather than a weighted average, for the pooled standard deviation to calculate an average of the comparison and test group‘s pricing behaviors that equally reflects both groups:
[T]he purpose of Commerce‘s Cohen‘s d test is to determine whether U.S. prices differ significantly among purchasers, regions, or time periods - i.e., do prices to each purchaser, region, or time period differ significantly from all other prices of the comparable merchandise. Although these are all prices in the U.S. market made by the respondent, this analysis requires that these prices be subdivided into separate distinct groups to consider separately whether the respondent‘s pricing behavior for sales to one specific group differs from its pricing behavior for all other sales. In other words, these prices, all of which are used to evaluate: 1) a respondent‘s pricing behavior in the U.S. market; and 2) whether the respondent is dumping, are now considered to represent two distinct pricing behaviors which may differ significantly. For the purpose of this particular analysis, Commerce finds that these two distinct pricing behaviors are separate and equally rational, and each is manifested in the individual prices within each group.
Second Remand Results at 8-9, 32. Commerce reasons that the appropriate yardstick by which to gauge the difference in the means between the groups is a simple average of each group‘s standard deviation, rather than one that weights the deviation based on the individual sales. See id. at 9-10. Commerce explains, given the objective of comparing pricing behavior of two distinct groups of sales, weight-averaging by sales volume or value (or number of transactions), would inappropriately move the pooled standard deviation toward the pricing behavior of either the test or comparison group. See Second Remand Results at 7-9, 14-16, 32. Thus, in its Second Remand Results, Commerce continues to find that using a simple average is preferable to a weighted average, and Commerce further addresses the specific points for which the Court of Appeals ordered further explanation. See generally id.
In using a simple average, Commerce acknowledges that the Cohen‘s d literature supports the use of a weighted average, but further explains that the literature varies depending on the given context. See
and why a simple average is more accurate than a weighted average.10 See Mid Continent III, 940 F.3d at 674.
Commerce also confronts the examples PT provided to the Court of Appeals, and the Court of Appeals referenced in its holding, which PT argues prove that simple averaging to get the pooled variance distorts the results of the Cohen‘s d test. See Second Remand Results at 10-14; Mid Continent III, 940 F.3d at 674. As a preliminary matter, Commerce claims, as it did in the Final Decision Memo, that PT relies upon a flawed assumption in the use of its examples. Namely, Commerce claims that PT‘s examples rest on the assumption that the standard deviation will increase with the group size.11 The Court of Appeals noted that Commerce‘s determination that the assumption PT was purportedly making—that standard deviation will increase with the group size—is not always true was unsupported and, in any event, did not substitute for an explanation of why using the weighted-average methodology would distort the results. See Mid Continent III, 940 F.3d at 674. PT refutes Commerce‘s characterization of its argument and states unequivocally “that
PT‘s proposed methodology has nothing to do with ‘the assumption that the standard deviation increases as the size of the group increases.‘” PT‘s Br. at 34. Indeed, PT provides other examples to illustrate its point that Commerce‘s characterization of its argument is incorrect and that its approach
Nonetheless, PT‘s examples do not demonstrate that Commerce‘s approach is unreasonable. PT claims its examples show that using a simple average is objectively unreasonable, but it offers no explanation in support of its assertion, other than pointing out that the use of a simple average results in a dumping margin for PT, while the use of a weighted average produces a de minimis rate. See PT‘s Br. at 3, 8-12. In response to the Court of Appeals concerns, Commerce explains how its approach effectuates its statutory objective. Commerce further states that PT‘s argument that Commerce‘s choice of methodology may produce different results does not on its own demonstrate that Commerce‘s chosen methodology is unreasonable. See Fujitsu Gen. v. United States, 88 F.3d 1034, 1043 (Fed. Cir. 1996) (“To survive judicial scrutiny, an agency‘s construction need not be the only reasonable interpretation or even the most reasonable interpretation.“).
Finally, although the Court of Appeals found that Commerce‘s view—that simple averaging was more predictable, while weighted averaging can conceal manipulation—appeared to be based on the assumption that weight averaging must be done by counting the number of transactions, as opposed to quantities, see Mid Continent III, 940 F.3d at 674, Commerce now explains that that distinction is immaterial. See Second Remand Results at 15-16. Commerce states that no matter how it weights the smaller group, either by volume, value or number of transactions, using a weighted average “would improperly give preference to one pricing behavior over another.”12 See id. at 16.
Before the court, PT supplies various examples as to why, in its view, weighted-averaging leads to reasonable results and a simple average does not. See PT‘s Br. at 8-16. Defendant argues that PT‘s examples and arguments relying on these examples are barred by the doctrine of exhaustion. See Second Remand Results at 18-25. “[T]he Court of International Trade shall, where appropriate, require the exhaustion of administrative remedies.”
Much of the information to which Defendant objects further illustrates arguments that PT already made.13 To the
Nonetheless, as further examples they do no more than reveal that the results of a simple average and weighted average pooled standard deviation are different. See PT‘s Br. at 17-18 (presenting and explaining Table 1, which shows the number of times that the simple average and weighted average produced different results). Although PT repeatedly concludes that the weighted average results are reasonable, while the simple average results are not, it offers no analysis of why its
pronouncements are accurate. See, e.g., PT‘s Br. at 19 (“examples of [simple average] results being more reasonable than [weighted average] results, if they exist at all, are much harder to find“); see also, e.g., id. at 23 (“for certain scenarios Commerce‘s [simple average] methodology does not work“). The one point that PT appears to make with these visuals is that the test group may have a similar price spread as the comparison group in cases where the test group nonetheless passes the Cohen‘s d test using a simple average. See PT‘s Br. at Exs. 2, 3 & 4. Implicit in this point is a view that the test group should have a different spread from the comparison group in order to have a passing score for Cohen‘s d. It is unclear to the court why PT‘s implicit assumption should be true. The very point that Commerce makes to support its use of a simple average is that the test group and the comparison group are two distinct groups and the price variances within them need to be considered independently. The statute tasks Commerce with determining whether there are U.S. prices that differ significantly among purchasers, regions or periods of time; and therefore, Commerce examines the pricing behaviors for those distinct groups. See Second Remand Results at 8-9, 32; Def.‘s Resp. to Cmts. on Remand Redetermination at 15-16, Sept. 10, 2020, ECF No. 155 (“Def.‘s Resp.“). The court fails to see how the existence of similar price spreads among sales within the two groups necessarily means that the overall pricing behaviors of the two groups are the same.
For similar reasons PT‘s arguments regarding analysis of variance or “ANOVA” fail to persuade the court that Commerce‘s approach is unreasonable. Again, in its comments before this court PT includes arguments that Defendant claims PT failed to exhaust before the agency. See PT‘s Br. at 34-39. PT argues that the statistical principles upon which the Cohen literature relied support the use of a weighted average for the pooled standard deviation. See id. at 34-36. More specifically PT argues that the statistical concept ANOVA “establishes universal relationships between the spread of a batch of data . . . and the spreads within subgroups of that batch.” See id. at 35. Thus PT contends ANOVA
But again, even allowing PT to make the more detailed argument in this court does not convince the court that Commerce‘s approach is unreasonable. Commerce considered the arguments for using a weighted average, including the literature cited by PT explaining that statistical principles supported the use of a weighted average. See generally Second Remand Results; see also Fujitsu, 88 F.3d at 1039 (noting the deference due to Commerce‘s expertise for determinations involving “complex economic and accounting decisions of a technical nature“). Commerce concluded that, given the task before it as provided in the statute, it would rely on the simple average because in the context of the Cohen‘s d test, the U.S. prices to each purchaser, region or time period separately and equally represent the respondent‘s pricing behavior, which itself is determined by the respondent‘s rational economic goal of maximizing the benefits accruing to the respondent. Second Remand Results at 36. Therefore, PTs arguments concerning ANOVA as support for the use of a weighted average have been addressed by Commerce. For the foregoing reasons, Commerce‘s choice to use a simple average for the pooled standard deviation is reasonable.
CONCLUSION
For the foregoing reasons, it is
ORDERED that Commerce‘s Second Remand Results is sustained. Judgment will enter accordingly.
Dated: January 8, 2021
New York, New York
/s/ Claire R. Kelly
Claire R. Kelly, Judge
Notes
The administering authority 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) the administering authority explains why such differences cannot be taken into account using a method described in paragraph [
19 U.S.C. § 1677f-1(1)(A)(i) or (ii)].