Slip Op. 24-15
UNITED STATES COURT OF INTERNATIONAL TRADE
MID CONTINENT STEEL & WIRE, INC. ET AL.,
Plaintiff and Consolidated Plaintiffs,
v. Before: Claire R. Kelly, Judge UNITED STATES, Consol. Court No. 15-00213 Defendant,
and
PT ENTERPRISE INC. ET AL.,
Defendant-Intervenors and Consolidated Defendant- Intervenor.
OPINION AND ORDER
[Sustaining the Department of Commerce’s Fourth Remand Redetermination.]
Dated: February 12, 2024
Adam H. Gordon, The Bristol Group PLLC of Washington D.C., for plaintiff and defendant-intervenor Mid Continent Steel & Wire Inc.
Bruce M. Mitchell, Andrew T. Schutz, Dharmendra N. Choudhary, Max F. Schutzman, and Ned H. Marshak, Grunfeld Desiderio Lebowitz Silverman & Klestadt, LLP of New York 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 Corporation, President Industrial Inc., and Liang Chyuan Industrial Co., Ltd. Consol. Court No. 15-00213 Page 2
Mikki Cottet, Senior Trial Counsel, Commercial Litigation Branch, Civil Division, U.S. Department of Justice, of Washington, D.C., for the defendant United States. Also on the brief were Brian M. Boynton, Principal Deputy Assistant Attorney General, and Patricia M. McCarthy, Director. Of counsel Vania Y. Wang, Attorney, Office of the Chief Counsel Civil Division Trade Enforcement & Compliance U.S. Department of Commerce, of Washington, D.C.
Kelly, Judge: Before the Court is the U.S. Department of Commerce’s
(“Commerce”) Final Results of Redetermination Pursuant to Court Remand, Aug. 31,
2023, ECF No. 207-1 (“Fourth Remand Results”) in the antidumping duty
investigation of certain steel nails from Taiwan, following the third remand
redetermination made in accordance with the mandate of the U.S. Court of Appeals
for the Federal Circuit in Mid Continent Steel & Wire, Inc. v. United States, 31 F.4th
1367, 1381 (Fed. Cir. 2022) (“Mid Continent V”) rev’g in part 495 F. Supp. 3d 1298
(Ct. Int’l Tr. 2021) (“Mid Continent IV”). Following this Court’s fourth remand order,
see Mid Continent Steel & Wire, Inc. v. United States, 628 F. Supp. 3d 1316 (Ct. Int’l
Tr. 2023) (“Mid Continent VI”), Commerce again contends its use of simple averaging
is reasonable. For the following reasons, Commerce’s fourth remand redetermination
is sustained.
BACKGROUND
The Court presumes familiarity with the facts of this case from this Court’s
previous opinions, as well as the Court of Appeals’ decisions in Mid Continent Steel
& Wire, Inc. v. United States, 940 F.3d 662 (Fed. Cir. 2019) (“Mid Continent III”) and
Mid Continent V, and will discuss additional facts relevant to the Court’s review of
the Fourth Remand Results. On June 25, 2014, Commerce initiated an antidumping Consol. Court No. 15-00213 Page 3
duty 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). On May 20, 2015, Commerce issued its final determination, which
resulted in an antidumping duty order on subject nails from Taiwan. 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, May 13, 2015, ECF No. 17 (“Final Decision
Memo.”).
On March 23, 2017, this Court sustained Commerce’s determination, including
its decision to use a simple average of standard deviations in the denominator of
Cohen’s d test. See Mid Continent Steel & Wire, Inc. v. United States, 219 F. Supp.
3d 1326, 1351 (Ct. Int’l Tr. 2017) (“Mid Continent I”). On October 3, 2019, the Court
of Appeals vacated this Court’s judgment and remanded in part to Commerce for
further explanation of its decision to use a simple average of standard deviations in
the denominator of Cohen’s d test. See Mid Continent III, 940 F.3d at 674–75. On
remand, Commerce defended its decision to use the simple average, explaining that
its use of the simple average was both accurate and in accord with statistical
literature. See Final Results of Redetermination Purs. Ct. Remand at 4, 15–16, June
16, 2020, ECF No. 144-1 (“Second Remand Results”). On January 8, 2021, this Court Consol. Court No. 15-00213 Page 4
again sustained Commerce’s decision, concluding that Commerce had adequately
explained how its use of simple averaging was more accurate, and thus a reasonable
choice of methodology. See Mid Continent IV, 495 F. Supp. 3d at 1308.
On April 21, 2022, the Court of Appeals vacated this Court’s judgment,
remanding to Commerce for further explanation of its decision to use the simple
average. See Mid Continent V, 31 F.4th at 1381; see also Mandate, June 13, 2022,
ECF No. 177; Remand Order, June 14, 2022, ECF No. 178. The Court of Appeals held
that Commerce inadequately explained its choice of the simple average of the
standard deviations for the Cohen’s d denominator. Mid Continent V, 31 F.4th at
1378–81. The Court of Appeals rejected Commerce’s reasoning that the “equally
rational” and “equally genuine” pricing choices warranted equal weighting in the
Cohen’s d denominator. Id. at 1379. The Court of Appeals explained that “Commerce
needs a reasonable justification for departing from what the acknowledged literature
teaches about Cohen’s d.” Id. at 1381. The Court of Appeals also suggested that the
preferred way to establish the denominator was to “use the standard deviation of the
entire population.” Id. at 1377.
In the third remand redetermination, Commerce defended its decision to use
the simple average with the Cohen’s d test, explaining that its usage is consistent
with statistical literature. See Final Results of Redetermination Purs. Ct. Remand
at 42–43, 52, Nov. 10, 2022, ECF No. 186-1 (“Third Remand Results”). In Mid
Continent VI, this Court remanded Commerce’s third final results redetermination, Consol. Court No. 15-00213 Page 5
concluding that Commerce had not complied with the Court of Appeals’ mandate to
provide a reasonable justification for departing from the academic literature and to
explain its choice to rely upon a simple average of the standard deviations of the test
and control groups to determine the denominator in its Cohen’s d analysis. 628 F.
Supp. 3d at 1322–23. More specifically, this Court found unjustified Commerce’s
position that the academic literature did not support use of a weighted average,
concluding that Commerce’s explanation “appears to contradict Cohen, Ellis, and Coe
at a number of points, as the Court of Appeals has already observed.” Id. at 1325
(citing Mid Continent V, 31 F.4th at 1378). In doing so, this Court instructed
Commerce to either explain its reasoning or reconsider its choice. Id. at 1326.
Commerce issued its Fourth Remand Results on August 1, 2023. See Fourth
Remand Results at 1. In the Fourth Remand Results, Commerce continues to rely on
a simple average for the Cohen’s d test, justifying its decision by contending the
simple average incorporates equal reliability of the calculated standard deviations,
and thus can be reasonably used to calculate the denominator of the Cohen’s d
coefficient. Id. at 10–13. Commerce also concludes that the Court of Appeals’
proposed alternative, to use a single standard deviation of all sale prices in the test
and comparison groups as the denominator, would not be appropriate in the context
of its differential pricing methodology. Id. at 13–17. Consol. Court No. 15-00213 Page 6
JURISDICTION AND STANDARD OF REVIEW
The Court has jurisdiction pursuant to Section 516A of the Tariff Act of 1930, 1
as amended, 19 U.S.C. § 1516a(a)(2)(B)(i) and 28 U.S.C. § 1581(c) (2012), 2 which
grants the Court authority to review actions contesting the final determination in an
antidumping duty order. The Court will uphold Commerce’s determination 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). “The results of a redetermination pursuant
to court remand are also reviewed ‘for compliance with the court’s remand order.’”
Xinjiamei Furniture Co. v. United States, 968 F. Supp. 2d 1255, 1259 (Ct. Int’l Tr.
2014) (quoting Nakornthai Strip Mill Pub. Co. v. United States, 587 F. Supp. 2d 1303,
1306 (Ct. Int'l Trade 2008). 3
1 Further citations to the Tariff Act of 1930, as amended, are to the relevant provisions of Title 19 of the U.S. Code, 2012 edition. 2 Further citations to Title 28 of the U.S. Code are to the 2012 edition. 3 Plaintiffs argue that Commerce “is not entitled to the same deference accorded [to
it] when this Court analyzed its initial decision,” and that Commerce should “not be accorded another chance” to explain use of simple averaging if another remand is required. Consol. Pls.’ Cmts. On [Fourth Remand Results] at 2, Oct. 2, 2023, ECF No. 209 (“Pls. Cmts.”). Plaintiffs cite cases which do not support a new standard of review in this case. See, e.g., INS v. Cardoza–Fonseca, 480 U.S. 421, 446 n.30 (1987) (explaining that an agency is afforded less deference to an interpretation that conflicts with previous interpretation of the authority at issue); Good Samaritan Hosp. v. Shalala, 508 U.S. 402, 417 (1993) (affirming deference to agency decision that “closely fits the design of the statute as a whole and its object and policy” despite shifts in agency practice years prior (internal citations and quotations omitted)); Tung Mung Dev. Co. v. United States, 25 C.I.T. 752, 772 (2001) (remanding Commerce’s determination where its decision was “a clear reversal of its prior
(footnote continued) Consol. Court No. 15-00213 Page 7
DISCUSSION
In a dumping investigation, Commerce typically compares the weighted
average of normal values with the weighted average of export prices for comparable
merchandise, unless it determines another method is appropriate. 19 U.S.C. § 1677f-
1(d)(1)(A)(i); 19 C.F.R. § 351.414(c)(1). Section 1677f-1, of Title 19, however, allows
Commerce to compare “the weighted average of the normal values to export
prices . . . of individual transactions for comparable merchandise if (i) there is a
pattern of 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 [with another method].” 4 19 U.S.C. § 1677f-
practice”); Olympic Adhesives, Inc. v. United States, 899 F.2d 1565, 1574–75 (Fed. Cir. 1990) (ordering directed remand where International Trade Administration failed to comply with statutory and regulatory requirements in the interest of time, circumstances, lack of evidence and judicial economy); CS Wind Vietnam Co. v. United States, 832 F.3d 1367, 1374 (Fed. Cir. 2016) (remanding Commerce’s unsupported decision and directing it to weight calculations regarding dumping margins). Commerce has not strayed from defending application of a simple average in its Cohen’s d test and has remained consistent in its underlying reasoning. See Fourth Remand Results at 10–13; Third Remand Results at 42–43, 52; Second Remand Results at 15–16, 39–40; Def.’s Resp. To [Pls. Cmts.] at 8–9, Nov. 15, 2023, ECF No. 212; [Def.-Int.] Reply To [Pls. Cmts.] at 2–3, Nov. 15, 2023, ECF No. 213 (“Def.-Int. Reply”). To the extent that the Court instructs Commerce to correct or otherwise address a deficiency in its decisionmaking, a court’s remand order represents a course correction to which the agency’s decisionmaking must comport when rendering a new determination that accords with its statutory obligations. SEC v. Chenery Corp., 332 U.S. 194, 199–201 (1947). 4 This subsection addresses targeted dumping, which occurs when an exporter sells
(footnote continued) Consol. Court No. 15-00213 Page 8
1(d)(1)(B)(i)–(ii). To implement Section 1677f-1(d)(1)(B), Commerce performs a
“differential pricing analysis” of a respondent’s sales to determine whether a “pattern
of significantly different prices” exists. 5 See Differential Pricing Analysis; Request
for Comments, 79 Fed. Reg. 26,720, 26,722 (Dep’t of Commerce May 9, 2014). This
analysis contains three tests—the Cohen’s d test, the ratio test, and the meaningful
difference test. See id.; Mid Continent V, 31 F.4th at 1371. Only the Cohen’s d test,
which determines whether there is a “pattern of prices that differ significantly,” is at
issue in this case. See Mid Continent V, 31 F.4th at 1369–70; Differential Pricing
Analysis; Request for Comments, 79 Fed. Reg. at 26,722.
As applied by Commerce, the Cohen’s d test involves comparing the prices of
“test groups” of a respondent’s sales to a “comparison group” by region, purchaser,
and time period. See Differential Pricing Analysis; Request for Comments, 79 Fed.
Reg. at 26,722. For each category, Commerce segregates sales into subsets, with one
subset becoming the test group, and the remaining subsets being combined as the
comparison group. Id. Commerce then calculates the means and standard deviations
of the test and comparison groups. Id. Commerce finally calculates a d coefficient by
at a lower, “dumped” price to particular customers or regions, while selling at higher prices to other customers or regions, such that the higher-priced products mask the dumped products by increasing the overall average price. See Apex Frozen Foods Priv. Ltd. v. United States, 862 F.3d 1337, 1341 (Fed. Cir. 2017). 5 The Statement of Administrative Action of the Uruguay Round Agreements Act
explains that Commerce should proceed “on a case-by-case basis, because small differences may be significant for one industry or one type of product, but not for another.” Uruguay Round Agreements Act, Statement of Administrative Action, H.R. Doc. No. 103-316, vol. 1, at 842–43 (1994), reprinted in 1994 U.S.C.C.A.N. 4040, 4178. Consol. Court No. 15-00213 Page 9
dividing the difference in the groups’ means by the square root of the average of the
squared standard deviations of each group. 6 See Fourth Remand Results at 6 (citing
Cohen at 20). Commerce finds the average of the squared standard deviations by
adding them together and dividing by two, referring to the result as a “simple
average.” See id. Commerce does not account for the differences in the size of each
group, i.e., use a “weighted average.” Fourth Remand Results at 6.
Commerce tests each subset against the remaining subsets across each
category and assigns a d coefficient. If the d value of a test group is equal to or greater
than the “large threshold,” or 0.8 (the difference in the means was at least 80% of the
pooled standard deviation), the observations within that group are said to have
“passed” the Cohen’s d test. Differential Pricing Analysis; Request for Comments, 79
Fed. Reg. at 26,722. If a sufficient quantity of sales by volume pass Cohen’s d test,
Commerce may compare the export prices of individual transactions to normal value,
instead of comparing the average export prices to normal value. Id. at 27,622–23.
The Court determines whether Commerce’s methodology is reasonable in light
of considerations that run counter to its decision. See Motor Vehicle Mfrs. Ass’n of
U.S. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983); Ceramica
6 Thus, d = |݉ െ ݉ | / ඥ(ɐଶ + ɐଶ )Τ2, where |݉ െ ݉ | is the absolute value of the difference in means between the test and comparison groups, and ɐଶ + ɐଶ is the sum of the squared standard deviation of both groups. Standard deviation squared (ɐଶ ) is also referred to as “variance.” Commerce’s formulation of what it calls the Cohen’s d test is also known as Cohen’s equation (2.3.2). See Cohen, Jacob, Statistical Power Analysis for the Behavioral Sciences, 44, (2d ed. 1988), A-580-876, PRRD 8, bar code 4181776-01 (Nov. 12, 2021) (“Cohen”). Consol. Court No. 15-00213 Page 10
Regiomontana, S.A. v. United States, 636 F. Supp. 961, 966 (Ct. Int’l Tr. 1986), aff’d,
810 F.2d 1137, 1139 (Fed. Cir. 1987); see also, e.g., Stupp Corp. v. United States, 5
F.4th 1341, 1354 (Fed. Cir. 2021) (stating the standard of review for components of
Commerce’s differential pricing methodology is reasonableness) (citing Mid
Continent III, 940 F.3d at 667).
In the Fourth Remand Results, Commerce explains its choice to employ a
simple average in the Cohen’s d denominator, acknowledging as it must, the Court of
Appeals’ holding that the academic literature surrounding Cohen’s d relies upon a
weighted average. Fourth Remand Results at 9 (accepting the Court of Appeals
finding that in the Cohen’s d literature, simple averaging applies only when the
sample sizes are equal); Mid Continent V, 31 F.4th at 1378 (“In making [its] choice to
use simple averaging . . . Commerce departed from, rather than followed, the cited
statistical literature”). Nonetheless, Commerce maintains the reasonableness of its
use of a simple average for the Cohen’s d denominator. To support its determination,
Commerce explains that although the academic literature most often employs a
weighted average when pooling the standard deviations of two samples, the literature
uses a simple average when the sample sizes are equal. Fourth Remand Results at
12–13. Commerce reasons that the use of a simple average where sample sizes are
equal stems from the equal reliability of standard deviations in samples of equal
sizes. Id. at 13. Consol. Court No. 15-00213 Page 11
Commerce’s focus on reliability stems from the use of samples in the literature.
Where samples are compared and a standard deviation for each sample is an
approximate, the actual standard deviation for the group represented by the sample
is not known. Fourth Remand Results at 10 (citing Cohen at 6); see also Mid
Continent V, 31 F.4th at 1377. However, the larger the sample size, the more reliable
that approximate. Fourth Remand Results at 10 (citing Cohen at 6). Thus, where
two samples are compared, the value of the standard deviation as an approximate is
necessarily a function of the sample size. Id. at 11–12. The larger sample size will
be more reliable, and thus should play a greater role, in evaluating the difference
between the means. Id. 7
Using this reliability framework Commerce reasons that just as sample sizes
of the same size share the same level of reliability, so do any two full populations.
See id. Where a full population is examined, the standard deviation is not an
approximate. Id. at 12. The standard deviation of a full population is in fact the
7 Logically, where there is more data upon which an estimate is based, the estimate should be more accurate. Yet, Plaintiffs reject Commerce’s reference to the size of a sample in its reasoning, because in the academic literature, the size of the sample refers to counts, typically of people. Pls. Cmts. at 9–10. However, Commerce’s practice is to base its analysis not on the number of transactions, but on the weights in kilograms of the product. Id. at 10. It is unclear to the Court the basis of Plaintiffs’ argument given that Commerce’s reference to counts is simply an example to illustrate its analysis. Commerce could easily have used weights rather than counts in explaining its reasoning. Commerce’s point is that when the size of two samples is the same, whether by weight or count, the two samples will have equal reliability. Consol. Court No. 15-00213 Page 12
actual standard deviation—it has 100% reliability. 8 Thus, comparing the two
standard deviations of two full populations is the same as comparing the standard
deviations of two samples of equal size. Id. at 11. The reliability of equal sample
sizes is the same and the reliability of two full populations is the same. Id. Although
it is true that the academic literature does not support the use of a simple average
for unequal sample sizes, Mid Continent V, 31 F.4th at 1378; Pls. Cmts. at 9 (arguing
that the availability of the simple average mechanism when the groups are the same
size does not support the use of the simple average when they are not), the Court of
Appeals explicitly instructed Commerce that it is not limited to the literature in
supporting its determination. Mid Continent V, 31 F.4th at 1381. Its methodology
must be reasonable. Id. (“Commerce needs a reasonable justification for departing
from what the acknowledged literature teaches about Cohen’s d”).
8 Plaintiffs argue that the reliability of data does not control Commerce’s decision regarding the Cohen’s d denominator. Pls. Cmts. at 9. Plaintiffs invoke this Court’s prior rationale with respect to weight averaging, namely, that just because weight averaging is supported in sampling does not mean it is unsupported when sampling is absent. Pls. Cmts. at 9 (citing Mid Continent VI, 628 F. Supp. 3d at 1324). Plaintiffs use this rationale to argue that equality in size or reliability is not indicative of whether the denominator should be based upon a weighted average, or a simple average. Id. Plaintiffs are correct that this Court previously faulted Commerce’s logic in that its conclusion did not follow from its premise. Mid Continent VI, 628 F. Supp. 3d at 1324 (“Commerce’s premise does not lead to its conclusion. That weighted averaging is supported when sampling is present does not mean that it is unsupported when sampling is absent”). Here Commerce’s logic is sound. It assumes that simple averaging is appropriate where there is equal reliability; and therefore, concludes that because full populations have equal reliability that simple averaging is appropriate for full populations. Consol. Court No. 15-00213 Page 13
Responding to the Court of Appeals, Commerce has provided an explanation
that logically connects the relevance of full populations to the use of simple averaging.
Commerce is not relying solely upon the academic literature to support its choice, but
rather argues that the principle it derives from the academic literature leads to a
logical conclusion that simple averaging in this case is a reasonable choice. Fourth
Remand Results at 12–13, 22–25. Commerce identifies where simple averaging is
supported by the literature, extrapolates a rationale for why simple averaging is
appropriate, and then applies that rationale to the circumstances before Commerce.
Although there may be other reasonable alternatives, the Court cannot find fault with
Commerce’s logic here. Commerce’s reliability analysis is reasonable.
Plaintiffs argue that the use of a simple average is not reasonable and suffers
from the same defect as Commerce’s reasoning in Mid Continent V, 31 F.4th at 1379,
in which it argued that the standard deviation of each group was equally rationale
and thus should be given equal weight. 9 Pls. Cmts. at 11. The Court of Appeals
rejected that explanation because:
The fact that the seller is acting rationally and genuinely in its pricing choices in both the test and comparison groups provides no apparent reason for assigning equal weight to each group's standard deviation when computing the pooled standard deviation. The rationality and genuineness of the seller's pricing choices have no evident connection to the undisputed purpose of the denominator figure—to provide a dispersion figure for the more general pool that serves as a yardstick for
9 Plaintiffs cite to Mid Continent III in their comments to support their position. Pls. Cmts. at 11. However, the quoted portion of the cited opinion and the reporter number and abbreviation are to Mid Continent V. See 31 F.4th at 1379; Pls. Cmts. at 11. Consol. Court No. 15-00213 Page 14
deciding on the significance of the difference in mean prices of the two groups. Both the numerator and denominator take the behavior as a given and form certain statistical measures from the objective data that are then related in the ratio that is Cohen's d. Commerce has not identified anything in the statistical measure at issue that depends on considerations of rationality and genuineness of the conduct that gave rise to the objective data. Indeed, Commerce has not shown that the numerous real-world examples used in Cohen to illustrate the methods taught are different in the respect Commerce now features, i.e., Commerce has not shown that the Cohen examples (generally or, perhaps, ever) involve sampled groups of data that reflect behavior that is not “rational” and “genuine.” Thus, Commerce has not adequately justified, through its central rationale, its departure from the statistical literature's description of the Cohen's d coefficient.
Mid Continent V, 31 F.4th at 1379. Here, Plaintiffs aver the arguments regarding
reliability—similar to arguments about rationality—fail to justify giving equal
weighting. Pls. Cmts. at 11–12. Although Defendant rejects the comparison, Fourth
Remand Results at 6–7, there is a similarity between Commerce’s earlier explanation
and this one, but only insofar as each explanation stems from the fact that the
standard deviation in the test and control group is drawn from a full population, and
therefore is not an approximate. Id. at 12; Second Remand Results at 39–40.
Commerce previously explained that the pricing behavior in each group was equally
genuine, it was the separate, distinct, and rational pricing for that group and thus
should be weighted equally. Second Remand Results at 8.
The point made by Commerce here is related but distinct. The pricing at issue
reveals a standard deviation that is not an approximate because it is based upon the
full population. Fourth Remand Results at 12. As Commerce elucidates, if the
standard deviation was a guess, then the literature would dictate a weighted average Consol. Court No. 15-00213 Page 15
because the guess would be dependent on the size of the sample. Id. at 14–16. Here,
Commerce addresses the Court of Appeals’ mandate to provide a “connection to the
undisputed purpose of the denominator figure.” Mid Continent V, 31 F.4th at 1379.
It premises the use of a simple average where there are equal sized samples on the
equal reliability of those samples, Fourth Remand Results at 12–13, a premise
Plaintiffs do not refute. It explains that the use of weighted average is reasonable
when sampled groups have unequal sizes because the standard deviation is simply
an estimate, and therefore weighting the sample size is appropriate (the larger
sample size would likely be more reliable than the smaller and therefore should be
weighted more). Id. at 10 (citing Cohen at 6). But when each group is not a sample,
but rather a full population, reliability concerns would not support greater weight to
the deviation found in the larger size group. Id. at 23–24.
Plaintiffs do not challenge the premise upon which Commerce relies, i.e., that
it is appropriate to use a simple average for equal sample sizes because the two
samples have equal reliability. See generally Pls. Cmts. Rather, Plaintiffs argue that
Commerce’s “analysis proves nothing.” Id. at 10. Plaintiffs state that reliability or
precision is dependent on a number of factors, at least with respect to samples. Id.
(“precision depends on multiple factors, including sample size, the amount of
variation in the population, the method by which the sample was obtained, the
method used to estimate the population property from the sample property, and other
factors”). Plaintiffs contend that the reliability of a sample cannot be compared to Consol. Court No. 15-00213 Page 16
the reliability of a full population. 10 Id. at 10–11. However, Commerce is not
comparing the reliability of a sample to the reliability of a full population, rather
Commerce argues that samples of equal sizes have equal reliability and full
populations have equal reliability. Fourth Remand Results at 12–14. Therefore,
Commerce reasons that if it is appropriate to use a simple average where sample sizes
are equal, because of the equal reliability, then it is appropriate to use a simple
average where full populations are being used. Id. at 13.
Plaintiffs assert that Commerce’s past practice supports use of a weighted
average in its differential analysis. Pls. Cmts. at 13. Specifically, Plaintiffs argue
that Commerce uses a weighted average when evaluating home market and U.S.
markets to calculate a respondent’s dumping margin. Id. at 13–14. This similarity
in calculation, Plaintiffs reason, supports use of a weighted average in Commerce’s
differential pricing analysis, rather than the simple average used here. Id. at 14.
Plaintiffs contend that Commerce relies on weighted average for all phases of pricing
calculations “until the very end, at which point it inexplicably relies on simple
averaging of two groups of data which have been obtained by weighted average prices
10 Plaintiffs attempt to cast doubt on Commerce’s reliability framework, asserting that Commerce incorrectly claims “a perfectly reliable full population is 100% reliable.” Pls. Cmts. at 11. Instead, Plaintiffs contend that perfect reliability “should be expressed as having zero errors.” Id. However, Plaintiffs fail to explain in any further detail any actual distinction between the two descriptions. Moreover, Plaintiffs’ distinction does not undermine Commerce’s analysis, as Plaintiffs further fail to explain how the characterization of a perfectly reliable full population as having zero errors meaningfully alters the results. Consol. Court No. 15-00213 Page 17
and weighted standard deviations of prices.” Id. Moreover, Plaintiffs state that
substitution of simple averaging for weighted averaging at this phase of the
calculations “skews the results by according more weight to certain sales (and less
weight to others) than they previously had accorded throughout the analysis.” Id.
Plaintiffs’ argument is inapposite. Plaintiffs argue that because Commerce
weight averages to determine dumping margins, that it should weight average in its
differential pricing methodology. Id. at 15. Plaintiffs fail to acknowledge that
Commerce’s task in its differential pricing methodology serves a diagnostic purpose.
Fourth Remand Results at 55; 19 U.S.C. § 1677f-1(d)(1)(B). Congress’ grant of
authority to Commerce dictates that diagnostic purpose. 19 U.S.C. § 1677f-1(d)(1)(B)
(“[Commerce can compare] the weighted average of the normal values to export
prices . . . of individual transactions for comparable merchandise if (i) there is a
pattern of 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 [with another method]”). Moreover,
Commerce has significant discretion to establish a reasonable methodology. Mid
Continent V, 31 F.4th at 1376 (“Commerce has discretion to make reasonable choices
within statutory constraints” (citing Mid Continent III, 940 F.3d at 667)). Dumping
margin calculations simply do not determine whether the difference in prices between
the two groups is significant or “the degree to which the phenomenon is present in Consol. Court No. 15-00213 Page 18
the population,” but rather the potential uncollected dumping duty due. See Fourth
Remand Results at 55; Pls. Cmts. at 13.
Plaintiffs also point to a handful of examples they claim refute Commerce’s
justification for use of simple averaging in its calculation.11 Pls. Cmts. at 19–24.
Plaintiffs claim the data in the examples, including both hypothetical numbers and
sales from Plaintiff PT’s database, exhibit how the simple average skews the results
by “over-weigh[ing] the smaller group,” causing “a low ‘no-pass’ value of d to exceed
Commerce’s threshold of 0.80” and thus a false “pass” under Cohen’s d. Id. at 25.
However, and as Commerce explains, Plaintiffs examples are inapposite. Fourth
Remand Results at 41–43. Plaintiffs’ examples illustrate that when the averaging of
two values changes from an identical average (with equal weights) to a weighted
average (with unequal weights), the results will invariably change. 12 Id. at 52.
11 Defendant-Intervenor offers its own example of the dangers entailed by Plaintiffs’ suggestion of use of a weighted average in Cohen’s d. See Def.-Int. Reply at 7–8. Defendant-Intervenor claims use of a “weighted average based on the physical weights of sales within each group as the denominator of [Cohen’s] d,” as Plaintiffs suggest, “opens the door to manipulation.” Id. at 7. This approach gives more weight to the standard deviation from smaller groups when those smaller groups are from larger sales, and Defendant-Intervenor argues that a supplier can manipulate the measure of d by changing the relative volume even if the mean difference between the groups is relatively large. Id. at 7. Defendant-Intervenor argues that there is potential for manipulation “[g]iven the prevalence and sophistication of many respondents’ ‘dump-proofing’ activities.’” Id. at 8. 12 Plaintiffs’ five provided examples, which involve both hypothetical and discretely
selected datasets, do nothing to undermine the reasonableness of Commerce’s use of simple averaging as a general practice. Plaintiffs’ examples show how the use of a weighted average lead to different results for these examples. Plaintiffs seem to
(footnote continued) Consol. Court No. 15-00213 Page 19
Plaintiff’s examples serve to illustrate how weighting would work; they do not
undermine the reasonableness of Commerce’s use of simple averages. Id. at 53. 13
Finally, Commerce addresses the Court of Appeals suggestion that it could
consider using the standard deviation of the full population. Commerce reasons that
“the single standard deviation causes the denominator of the Cohen’s d coefficient to
reflect not just the dispersion of the data within each group, but also the dispersion
of the data between the two groups.” Id. at 17. However, Commerce views effect size,
contend that the visualizations of the data they provide in their five examples illustrate that their approach is correct, and that Commerce’s use of a simple average is incorrect. Pls. Cmts. at 19–27. However, Commerce’s use of Cohen’s d in differential pricing calculations is not a visual analysis, but rather is a statistical methodology. See Differential Pricing Analysis; Request for Comments, 79 Fed. Reg. at 26,722. That Plaintiffs can identify five examples that do not correspond to what they intuitively believe should be a visual representation of “a pattern of significant price differences” is of little analytical value. Pls. Cmts. at 28. Even assuming Plaintiffs’ intuitive belief regarding an appropriate visual representation of “a pattern of significant price differences” is correct; Commerce is not tasked with developing a perfect methodology. It is tasked with developing a reasonable methodology. Furthermore, Commerce is not relying on a visual analysis to support the reasonableness of its methodology. It relies upon principles taken from the literature and logic. 13 Plaintiffs submit that even if Commerce’s choice of methodology is reasonable, its
determination in this case is unsupported by substantial evidence. More specifically, they argue the facts of this case warrant departure from the methodology because using it would lead to unreasonable results “contrary to economic reality.” Pls. Cmts. at 29–30. However, and as Commerce explains, Plaintiffs fail to expound upon precisely what the economic reality is that warrants departure from simple averaging. Fourth Remand Results at 53. Without further explanation or record support, Plaintiffs’ argument is unpersuasive. Consol. Court No. 15-00213 Page 20
i.e., the d coefficient, as meant to quantify the difference in the mean prices of each
group relative to the dispersion of prices within each group. 14 Id. at 17.
The question before this Court is not whether the Court of Appeals’ proposal is
a reasonable one, as it would appear to be given the literature, but whether it detracts
from the reasonableness of Commerce’s proposal. Commerce has explained its
rationale as based on the equal reliability of both full populations and equal sized
samples. It has also explained that standard deviation is specific to the mean to
which it relates. Id. at 14 (“[the standard deviation] in Dr. Cohen’s equations 2.2.1
and 2.2.2, is either the standard deviation of population A or the standard deviation
of population B, but it is not the standard deviation of populations A and B combined
together”). Because it is evaluating full populations, Commerce explains that using
14 Plaintiffs reject the independent nature of these two groups. Pls. Cmts. at 12 (arguing that the test in comparison groups “do not have independent existences”). Plaintiffs make this point by noting that any sale might be in either a test group or control group depending on Commerce’s focus. Id. at 12; see Fourth Remand Results at 5 (explaining that in its differential pricing analysis, Commerce uses the Cohen’s d test to measure “whether the sale prices to a given purchaser, region, or time period differ significantly from the sale prices of comparable merchandise to other purchasers, regions, or time periods, respectively”). Plaintiffs argue that it is illogical for any sale to receive more weight depending upon whether it is in the test or comparison group, as it necessarily does if Commerce uses a simple average. See Pls. Cmts. at 12–13 (“how can the essay methodology lead to reliable results when each sale has a different effect on the result, depending upon the group in which it falls?”). Plaintiff’s argument is without merit. Commerce explains it is comparing the prices to a given purchaser, region or time. The statute identifies these grouping as distinct. See 19 U.S.C.§ 1677f-1(d)(B)(i) (instructing Commerce to determine whether “there is a pattern of export prices (or constructed export prices) for comparable merchandise that differ significantly among purchasers, regions, or periods of time”); Fourth Remand Results at 55. Consol. Court No. 15-00213 Page 21
the dispersion of the group as a whole would eliminate the relevancy of each
individual standard deviation much in the same way that weighting the standard
deviations would diminish the relevancy of one of the standard deviations. See id. at
14–18. Thus, Commerce has explained how its choice is reasonable and has
addressed any evidence or arguments that might detract from the reasonableness of
its choice. See Mid Continent V, 31 F.4th at 1381 (“Commerce must either provide
an adequate explanation for its choice of simple averaging or make a different choice,
such as use of weighted averaging or use of the standard deviation for the entire
population”).
CONCLUSION
Commerce has provided a reasonable explanation for its use of a simple
average as instructed by the Court of Appeals and this Court and its determination
is sustained. Judgment will enter accordingly.
/s/ Claire R. Kelly Claire R. Kelly, Judge
Dated: February 12, 2024 New York, New York