Slip Op. 23-181
UNITED STATES COURT OF INTERNATIONAL TRADE
NEXTEEL CO., LTD., Plaintiff, and SEAH STEEL CORPORATION,
Consolidated Plaintiff, Before: Jennifer Choe-Groves, Judge v.
Consol. Court No. 18-00083 UNITED STATES,
Defendant, and
UNITED STATES STEEL CORPORATION, MAVERICK TUBE CORPORATION, AND TENARISBAYCITY,
Defendant-Intervenors.
OPINION
[Sustaining the U.S. Department of Commerce’s fourth remand redetermination following the 2015–2016 administrative review of the antidumping duty order on oil country tubular goods from the Republic of Korea.]
Dated: December 18, 2023
Jeffrey M. Winton, Amrietha Nellan, and Jooyoun Jeong, Winton & Chapman PLLC, of Washington, D.C., for Consolidated Plaintiff SeAH Steel Corporation.
Claudia Burke, Deputy Director, and Hardeep K. Josan, Trial Attorney, Commercial Litigation Branch, Civil Division, U.S. Department of Justice, of New York, N.Y., for Defendant United States. With them on the brief were Brian M. Boynton, Principal Deputy Assistant Attorney General, and Patricia M. McCarthy, Director. Of counsel was Mykhaylo Gryzlov, Office of the Chief Counsel for Trade Enforcement & Compliance, U.S. Department of Commerce, of Washington, D.C.
Choe-Groves, Judge: Before the Court is the U.S. Department of Commerce’s (“Commerce”) fourth remand redetermination in the administrative review of the antidumping duty order on oil country tubular goods (“OCTG”) from the Republic of Korea (“Korea”) covering the period from September 1, 2015 to August 31, 2016. See Commerce’s Final Results of Redetermination Pursuant to Court Remand (“Fourth Remand Redetermination”), ECF No. 140-1, pursuant to Order, ECF No. 130; see also Certain Oil Country Tubular Goods From the Republic of Korea, 83 Fed. Reg. 17,146 (Dep’t of Commerce Apr. 18, 2018) (final results of antidumping duty administrative review and final determination of no shipments; 2015–2016) (“Final Results”), and accompanying Issues and Decision Memorandum for the Final Results of the 2015–2016 Administrative Review of the Antidumping Duty Order on Certain Oil Country Tubular Goods from the Republic of Korea (Apr. 11, 2018) (“Final IDM”), PR 368.1
1 Citations to the administrative record reflect the public administrative record (“PR”) and public fourth remand record (“PRR”) document numbers. ECF Nos. 60, 94, 129, 148.
3
In NEXTEEL Co. v. United States (“NEXTEEL V”), 47 CIT __, 633 F.
Supp. 3d 1190 (2023), the Court remanded for Commerce to reconsider or further discuss the issue of Commerce’s calculation and application of the 0.8 threshold in the Cohen’s d analysis. NEXTEEL V, 47 CIT at __, 633 F. Supp. 3d at 1201.
For the following reasons, the Court sustains Commerce’s Fourth Remand Redetermination.
BACKGROUND
The Court presumes familiarity with the facts and procedural history of this case and recites the facts relevant to the Court’s review of the Fourth Remand Redetermination. See NEXTEEL Co. v. United States (“NEXTEEL I”), 43 CIT __, __, 392 F. Supp. 3d 1276, 1283–84 (2019); NEXTEEL Co. v. United States (“NEXTEEL II”), 44 CIT __, __, 450 F. Supp. 3d 1333, 1336–37 (2020); NEXTEEL Co. v. United States (“NEXTEEL III”), 44 CIT __, __, 475 F. Supp. 3d 1378, 1379–80 (2020); NEXTEEL Co. v. United States (“NEXTEEL IV”), 28 F.4th 1226, 1231–33 (Fed. Cir. 2022); NEXTEEL V, 47 CIT at __, 633 F. Supp. 3d at 1192–93.
In this administrative review of OCTG from Korea, Commerce selected Plaintiff NEXTEEL Co., Ltd. (“NEXTEEL”) and Consolidated Plaintiff SeAH Steel Corporation (“SeAH”) as mandatory respondents for individual examination and determined that the Government of Korea’s involvement in the Korean
4
electricity market contributed to a particular market situation in Korea during the period of review. See Resp. Selection Mem. (Jan. 12, 2017), PR 28; Final IDM at 16–23.
In NEXTEEL I, the Court sustained in part and remanded in part the Final Results. 43 CIT at __, 392 F. Supp. 3d at 1297; see Commerce’s Final Results of Redetermination Pursuant to Court Remand (“Remand Redetermination”), ECF No. 81-1, pursuant to Order, ECF No. 73. Among the issues that the Court ordered Commerce to reconsider or further explain was the finding of a particular market situation in Korea. NEXTEEL I, 43 CIT at __, 392 F. Supp. 3d at 1286–88, 1292– 94. In the Remand Redetermination, Commerce reviewed the record de novo, provided more explanation, and again determined that a particular market situation in Korea distorted the cost of producing OCTG. NEXTEEL II, 44 CIT __, 450 F. Supp. 3d at 1336.
In NEXTEEL II, the Court sustained in part and remanded in part the Remand Redetermination. NEXTEEL II, 44 CIT at __, 450 F. Supp. 3d at 1346– 47; see Commerce’s Final Results of Redetermination Pursuant to Court Remand (“Second Remand Redetermination”), ECF No. 96-1, pursuant to Order, ECF No. 95. The Court concluded that Commerce’s particular market situation determination was not supported by substantial evidence. NEXTEEL II, 44 CIT at __, 450 F. Supp. 3d at 1343.
5
In NEXTEEL III, the Court sustained the Second Remand Redetermination, in which Commerce reversed its particular market situation determination and recalculated the margins of NEXTEEL and SeAH without a particular market situation adjustment. NEXTEEL III, 44 CIT at __, 475 F. Supp. 3d at 1380.
In NEXTEEL IV, the U.S. Court of Appeals for the Federal Circuit (“CAFC”) directed the Second Remand Redetermination to be remanded for Commerce to further consider whether a particular market situation could be found based on any subset of the factors or other reasoning, and for proceedings to be consistent with the CAFC’s decision in Stupp Corp. v. United States (“Stupp III”), 5 F.4th 1341 (Fed. Cir. 2021). NEXTEEL IV, 28 F.4th at 1231 (“[W]e vacated aspects of Commerce’s differential pricing analysis [in Stupp III] over concerns about Commerce’s use of statistical methodologies when certain preconditions for their use are not met. Commerce’s analysis here raises identical concerns, so we vacate the trial court’s decision upholding the methodology and remand for reconsideration in view of [Stupp III].”) (internal citation omitted).
In the Third Remand Redetermination, Commerce reconsidered the record and determined that substantial evidence did not support the conclusion that a particular market situation existed in Korea during the period of review. Third Remand Redetermination, ECF No. 119-1, pursuant to Order, ECF No. 114. Commerce reconsidered the differential pricing analysis, provided further
6
explanation regarding Commerce’s application of the Cohen’s d test to SeAH’s U.S. sales, and determined that the weighted-average dumping margins calculated in the Second Remand Redetermination would remain the same. Id. at 74.
In NEXTEEL V, the Court sustained Commerce’s determination that the alleged particular market situation did not exist during the period of review in Korea, but remanded Commerce’s determination of certain aspects of application of the Cohen’s d test in light of Stupp III because Commerce’s explanation did not resolve the CAFC’s concerns pertaining to the use of the 0.8 threshold when the statistical assumptions are not observed. NEXTEEL V, 47 CIT at __, 633 F. Supp. 3d at 1200–01. The Court also directed SeAH to place on the record the academic literature cited by Commerce in the Final IDM. Id. at __, 633 F. Supp. 3d at 1203.
In the Fourth Remand Redetermination, Commerce determined that the statistical criteria (i.e., normality of the distribution, equal variances, and roughly the same number of observations) discussed by the CAFC in Stupp III do not need to be observed in its application of the Cohen’s d test as part of its differential pricing analysis. Fourth Remand Redetermination at 5–7. Commerce explained that the academic literature included by Commerce in the Final IDM does not support the claims that Professor Cohen’s 0.8 threshold was derived based on the statistical criteria or that the use of the threshold should only be limited to situations when sampled data exhibit a normal distribution or similarly equal
7
variances because there is no express mention that these criteria or assumptions are necessary when examining an entire population, and that any adjustment to the population data would distort the actual population’s parameters and is no longer reflective of the whole population. See id. at 6–10. Commerce determined that it would not revise the calculation of the weighted-average dumping margins. Id. at 22.
Consolidated Plaintiff SeAH filed its comments in opposition, challenging Commerce’s application of the Cohen’s d test in the differential pricing analysis. Cmts. SeAH Steel Corp. Opp’n Commerce’s June 27, 2023 Redetermination (“SeAH’s Br.”) at 5–16, ECF No. 143. Defendant filed its response and argued that the Court should affirm the Fourth Remand Redetermination. Def.’s Resp. Cmts. Regarding Remand Redetermination (“Def.’s Br.”), ECF No. 146.
JURISDICTION AND STANDARD OF REVIEW The Court has jurisdiction under 19 U.S.C. § 1516a(a)(2)(B)(i) and 28 U.S.C. § 1581(c), which grant the Court authority to review actions contesting the final results of an administrative review of an antidumping duty order. The Court will hold unlawful any determination found to be unsupported by substantial record evidence or otherwise not in accordance with law. 19 U.S.C. § 1516a(b)(1)(B)(i). The Court reviews determinations made on remand for compliance with the Court’s remand order. Ad Hoc Shrimp Trade Action Comm.
8
v. United States, 38 CIT 727, 730, 992 F. Supp. 2d 1285, 1290 (2014), aff’d, 802 F.3d 1339 (Fed. Cir. 2015).
DISCUSSION
I. Administrative Exhaustion Defendant contends that Commerce did not have time to issue draft results of redetermination for SeAH’s review and comment because of Commerce’s unexpected workload in the months of April and May 2023. Def.’s Br. at 2, 4; see also Def.’s Unopposed Mot. Extension Time, ECF No. 138; Order (June 20, 2023), ECF No. 139. SeAH asserts that its brief to the Court is its first opportunity to comment on the remand results and argues that the doctrine of administrative exhaustion is inapplicable because Commerce did not provide a draft of the Fourth Remand Redetermination for its review and comment at the administrative level. SeAH’s Br. at 1 n.1.
Before commencing suit in the U.S. Court of International Trade, an aggrieved party must exhaust all administrative remedies available to it. “In any civil action . . . the Court of International Trade shall, where appropriate, require the exhaustion of administrative remedies.” 28 U.S.C. § 2637(d). The court “generally takes a ‘strict view’ of the requirement that parties exhaust their administrative remedies.” Yangzhou Bestpak Gifts & Crafts Co. v. United States, 716 F.3d 1370, 1381 (Fed. Cir. 2013) (citations omitted). There are limited
9
exceptions to the exhaustion requirement. See Pakfood Pub. Co. v. United States, 34 CIT 1122, 1145–48, 724 F. Supp. 2d 1327, 1351–53 (2010) (listing “futil[ity] for the party to raise its argument at the administrative level” and issues “fully considered by Commerce” as two generally recognized exceptions to the exhaustion doctrine); see also Holmes Prod. Corp. v. United States, 16 CIT 1101, 1104 (1992).
Because Commerce did not issue a draft of the Fourth Remand Redetermination on which SeAH could comment, SeAH could not have raised its arguments at the administrative level and therefore did not fail to exhaust its administrative remedies. Thus, SeAH’s arguments are not barred by the doctrine of administrative exhaustion.
II. Placement of Academic Material on the Record Judicial review is generally limited to the administrative record before the agency at the time it rendered its decision. See Camp v. Pitts, 411 U.S. 138, 142 (1973). “The purpose of limiting review to the record actually before the agency is to guard against courts using new evidence to convert the arbitrary and capricious standard into effectively de novo review.” Axiom Res. Mgmt., Inc. v. United States, 564 F.3d 1374, 1380 (Fed. Cir. 2009) (internal quotations and citation omitted).
10
In its prior opinion, the Court directed SeAH to place on the administrative record copies of academic literature cited by Commerce in the Final IDM. NEXTEEL V, 47 CIT at __, 633 F. Supp. 3d at 1203. Commerce confirmed that SeAH placed the academic literature on the administrative record in compliance with the Court’s remand. Fourth Remand Redetermination at 2, 5; see also Letter from Jeffrey M. Winton, Winton & Chapman PLLC, re: Remand Redetermination in the Second Administrative Review of Oil Country Tubular Goods from Korea – Resubmission of Publications Pursuant to Department’s May 26 Letter (May 31, 2023) (“SeAH’s May 31, 2023 Submission”), PRR 3; Letter from Jeffrey M. Winton, Winton & Chapman PLLC, re: Remand Redetermination in the Second Administrative Review of Oil Country Tubular Goods from Korea – Resubmission of Publications Pursuant to Department’s June 8 Letter (June 12, 2023) (“SeAH’s June 12, 2023 Submission”), PRR 10 (collectively, “Academic Literature”).
III. Differential Pricing Analysis A. Legal Framework
When calculating a weighted dumping margin, Commerce may use three methodologies depending on the circumstances: the “average-to-average” (A-to-A) method, the “transaction-to-transaction” (T-to-T) method, or the “average-to- transaction” (A-to-T) method. See 19 U.S.C. § 1677f-1(d)(1); 19 CFR § 351.414(b)(1)–(3); Union Steel v. United States, 713 F.3d 1101, 1103 (Fed. Cir.
11
2013). Commerce ordinarily uses an A-to-A comparison of normal values to export prices for comparable merchandise in an investigation when calculating a dumping margin. See 19 U.S.C. § 1677f-1(d)(1)(A); 19 C.F.R. § 351.414(b)(1). The A-to-A methodology sometimes fails to detect targeted or masked dumping because a respondent’s “sales of low-priced ‘dumped’ merchandise would be averaged with (and offset by) sales of higher-priced ‘masking’ merchandise, giving the impression that no dumping was taking place.” Apex Frozen Foods Priv. Ltd. v. United States (“Apex”), 862 F.3d 1337, 1341 (Fed. Cir. 2017). “Targeted dumping” occurs when an exporter sells at a dumped price to particular customers or regions, while selling at higher prices to other customers or regions, and as a result, uses higher-priced products to mask dumped products when Commerce averages the sales using the A-to-A method. See id.
Congress has not provided a method for Commerce to use for determining whether a pattern of significantly different prices exists, but the Statement of Administrative Action (“SAA”) 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 843 (1994), reprinted in 1994 U.S.C.C.A.N. 4040, 4178.
12
Commerce can depart from using the A-to-A methodology and instead compare the weighted average of normal values to the export prices of individual transactions for comparable merchandise using the A-to-T methodology when (1) Commerce finds a pattern of export prices for comparable merchandise that differ significantly among purchasers, regions, or periods of time, and (2) Commerce explains why such differences cannot be taken into account using the A-to-A methodology. See 19 U.S.C. § 1677f-1(d)(1)(B). Commerce has adopted the same basis for applying its A-to-T methodology in administrative reviews. See JBF RAK LLC v. United States, 790 F.3d 1358, 1364 (Fed. Cir. 2015).
To determine whether a pattern of significant price differences exists among purchasers, regions, or periods of time, Commerce uses its two-stage differential pricing analysis: first, Commerce applies the “Cohen’s d test,” which measures the degree of price disparity between two groups of sales by calculating the number of standard deviations by which the weighted-average net prices of U.S. sales for a particular purchaser, region, or time period (the test group) differ from the weighted-average net prices of all other U.S. sales of comparable merchandise (the comparison group) to come up with the d coefficient; and second, Commerce then applies the “ratio test,” which considers the ratio of the sales in the targeted groups that have passed the Cohen’s d test to the exporter’s total U.S. sales to measure the
13
extent of significant price differences. See Apex, 862 F.3d at 1341 n.2Ѹ Differential Pricing Analysis[:] Request for Comments, 79 Fed. Reg. 26,720, 26,722–23 (Dep’t of Commerce May 9, 2014). The coefficient calculated from the first step may be situated within fixed thresholds, being small, medium, or large, and the targeted test groups pass the Cohen’s d test if they yield coefficients that are equal to or exceed the large threshold, or the 0.8 threshold. Apex, 862 F.3d at 1341 n.2ѸDifferential Pricing Analysis[:] Request for Comments, 79 Fed. Reg. at 26,722.
Based on how much of a percentage the passing sales account for the exporter’s total U.S. sales, Commerce applies a different methodology to the data. If the passing sales account for 66% or more of the value of total sales, then the pattern of significant price differences warrants application of the A-to-T method to all sales. SeeDifferential Pricing Analysis[:] Request for Comments, 79 Fed. Reg. at 26,722–23. If the passing sales make up more than 33% and less than 66% of the value of all sales, Commerce takes a hybrid approach, applying the A-to-T method to the sales that passed the Cohen’s d test and applying the A-to-A method to all other sales. See id. at 26,723. If the passing sales make up 33% or less of total sales, then Commerce will apply the A-to-A method. See id. If both the Cohen’s d test and ratio test demonstrate that the A-to-T methodology should be considered, and Commerce has not selected the A-to-A methodology, it applies its
14
“meaningful difference” test, with which Commerce evaluates whether the difference between the weighted-average dumping margins calculated by the A-to- A method is meaningfully different than the weighted-average dumping margins calculated by the A-to-T method. See id.
B. CAFC’s Concerns in Stupp III In Stupp III, the CAFC recognized that Mid Continent Steel & Wire, Inc. v.
United States (“Mid Continent”), 940 F.3d 662 (Fed. Cir. 2019), had resolved the issue of whether Commerce’s adoption of Professor Cohen’s 0.8 threshold to determine whether price differences were significant was reasonable. Stupp III, 5 F.4th at 1356–57. In Mid Continent, the CAFC held that Commerce was within the wide discretion left to it under 19 U.S.C. § 1677f-1(d)(1)(B) in adopting the 0.8 threshold because “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.” Mid Continent, 940 F.3d at 673. The CAFC did not reach the question of whether the 0.8 threshold could be applied when the data did not satisfy the statistical assumptions of the Cohen’s d test. Stupp III, 5 F.4th at 1356–57. The CAFC expressed concerns about the reasonableness of Commerce’s use of the Cohen’s d test in less-than-fair-value adjudications because such application to data that do not satisfy the assumptions
15
on which the test is based “may undermine the usefulness of the interpretive cutoffs.” Id. at 1357.
The CAFC identified three potential scenarios when the Cohen’s d test could raise such concerns: when the distribution of a respondent’s sales data is not normal (lack of normality), when the test data have few data points (lack of sufficient data), and when there is not normal variance in a respondent’s sales (lack of roughly equal variances). Id. at 1357–59. Additionally, the CAFC presented two hypotheticals to illustrate its concerns: (1) in its first hypothetical, an analysis of a group of only eight export sales across four groups, such that each test group would consist of only two sales, raising the concern that the prices’ lack of normality and sufficient data will produce inaccurate results on the Cohen’s d test; and (2) in its second hypothetical, an analysis of five test groups that contain sales prices that hover around the same value of about $100 each, differing from each other by up to two cents, raising the concern that prices with small variances will produce inaccurate results on the Cohen’s d test. Id. at 1358–59. In the first situation, the CAFC noted that an analysis of groups of such small numbers would potentially lack normality and produce an upward bias in effect size, which, in turn, will produce more passing results under the Cohen’s d test and create an exaggeration of dumping margins. Id. at 1359. In the second situation, the CAFC pointed out that the lack of variance in the data would artificially inflate the
16
dumping margins, and an objective examiner looking at these similar sales prices “would be unlikely to conclude that they embody a ‘pattern’ of prices which ‘differ significantly.’” Id. (citing 19 U.S.C. § 1677f-1(d)(1)(B)(i)). The CAFC remanded Commerce’s determination with instructions for “Commerce to clarify its argument that having the entire universe of data rather than a sample makes it permissible to disregard the otherwise-applicable limitations on the use of the Cohen’s d test.” Id. at 1360.
C. Commerce’s Remand Redeterminations In the Third Remand Redetermination, Commerce explained that its results from its utilization of the Cohen’s d test do not require statistical inferences because Commerce looks at the full population of sale prices and calculates the actual parameters of data, and does not rely on sampled data that are estimates of the actual values, which would require the observation of such statistical criteria. Third Remand Redetermination at 20–21, 58–60; see also NEXTEEL VI, 633 F. Supp. 3d at 1200–01. The Court concluded that this explanation did not sufficiently resolve the CAFC’s concerns or address the CAFC’s observation that Professor Cohen had derived his interpretive cutoffs under certain assumptions and required Commerce to consider the Academic Literature to discuss whether the application of the Cohen’s d coefficient and 0.8 threshold were reasonable. NEXTEEL VI, 633 F. Supp. 3d at 1200–01.
17
In the Fourth Remand Redetermination, Commerce provided a more complete explanation, stating that it reviewed the Academic Literature but found “no support in the Academic Literature for the claim that [Professor] Cohen’s 0.8 threshold was derived based on the statistical criteria, or that the use of [Professor] Cohen’s threshold should be limited to situations where the sampled data exhibit a normal distribution or similarly equal variances.” Fourth Remand Redetermination at 6. Commerce contended that the assumptions of normality, equal variances, and sufficient data are only relevant as a matter of statistical significance, that these assumptions do not apply when analyzing a whole population, and that the Academic Literature does not support the contention that Professor Cohen’s thresholds are derived from statistical assumptions, including two hypothetical frameworks of its own to illustrate its position. Id. at 12–14.
SeAH challenges the Fourth Remand Redetermination and Commerce’s application of the Cohen’s d test in its differential pricing analysis by arguing that: (1) the statistical assumptions of normality, equal variances, and sufficient data must be present regardless of data consisting of a sample or an entire population, and the data do not satisfy these assumptions; (2) Professor Cohen’s thresholds were based on normally-distributed data and not on the analyses of entire populations, but SeAH’s pricing data did not follow a normal distribution or have equal variances or a sufficient number of data points in the groups being compared;
18
and (3) Professor Cohen’s thresholds are not universally applicable, so Commerce’s use of the 0.8 threshold to calculate effect sizes for price differences is not an acceptable use of such threshold. SeAH’s Br. at 5–16.
Defendant argues that SeAH’s arguments should not prevail because Commerce provided a reasonable explanation for its methodology and the Academic Literature does not support the claim that Professor Cohen’s 0.8 threshold was based on the statistical assumptions for examining the entire population or limited to application for situations when only sampled data exhibit a normal distribution or similarly equal variances. Def.’s Br. at 4–5.
D. Differential Pricing Analysis The Court must determine whether Commerce’s explanation of its use of the 0.8 threshold resolves the CAFC’s concerns with Commerce’s methodology applied without the observation of certain statistical assumptions, including the normality, sufficient size, and roughly equal variances of the considered populations on which the Cohen’s d test was based. Stupp III, 5 F.4th at 1357–58; see also NEXTEEL IV, 28 F.4th at 1238–39 (citing the same concerns). The Court’s task is to determine whether Commerce’s methodology is reasonable, rather than to interpret the meaning of the Academic Literature and the correct application of the Cohen’s d test. See Motor Vehicle Mfrs. Ass’n of U.S. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983); Ceramica Regiomontana, S.A.
19
v. United States, 10 CIT 399, 404–05, 636 F. Supp. 961, 966 (1986), aff’d, 810 F.2d 1137 (Fed. Cir. 1987) (“As long as the agency’s methodology and procedures are reasonable means of effectuating the statutory purpose, and there is substantial evidence in the record supporting the agency’s conclusions, the court will not impose its own views as to the sufficiency of the agency’s investigation or question the agency’s methodology.”).
Commerce explained in the Fourth Remand Redetermination that Professor Cohen’s 0.8 threshold is not dependent on the statistical criteria identified by the CAFC in Stupp III, and there is no role for the statistical criteria to examine whether the test results are reliable and representative of the results when the calculations are based on the full populations of data because the prices used in the Cohen’s d test include all prices of comparable merchandise for the test and comparison groups. Fourth Remand Redetermination at 12. Commerce provided its own illustrative framework of hypotheticals to demonstrate the relationship between Professor Cohen’s thresholds and the statistical criteria. Id.
Commerce’s first hypothetical involved the differential pricing analysis of the prices of BigBill’s bicycles in Virginia and Maryland in 2020 using Professor Cohen’s 0.8 threshold, with two sets of analyses to show the difference between the results obtained by using a small sample size and the results obtained by using a sample size that represents the full population of data. Id. at 12–13. The first
20
analysis involved five random sale prices from each state, resulting in the d coefficient value of 0.9, and the second analysis involved twenty random sale prices from each state, resulting in the d coefficient value of 0.75. Id. Regarding the slight difference in results obtained by the two sets of data, Commerce argued that “the calculated results using the sample reliably represent the results as if the calculations had been based on the full populations of sale prices to each state.” Id. at 13.
Commerce’s second hypothetical involved the determination of the price difference between the sale prices of exotic sport cars in Vermont and New Hampshire. Id. at 13–14. Commerce posited that this example demonstrates that the concern of a small sample size, or concern over the lack of sufficient data, is not relevant when using the Cohen’s d test because “although the number of observations is small, the results reflect the actual values of the full population of sale prices.” Id. at 14.
In addition, Commerce asserted that the Academic Literature does not contain any express mention of criteria or assumptions necessary when examining data of an entire population and argued that there is no support for the claim that Professor Cohen’s 0.8 threshold was derived from statistical criteria. See id. at 6– 10. Commerce discussed Professor Cohen’s real-life “operational definitions” to illustrate small, medium, or large effects to support this proposition. See id. at 17–
21
21 (citing SeAH’s June 12, 2023 Submission at Att. 2 (“Cohen”) at 21–23, 26). In its discussion of the second “operational definition” of “percent nonoverlap,” which uses two bell curves to illustrate the difference in the means, Commerce contended that the use of Professor Cohen’s threshold should not be limited to situations in which sampled data exhibit a normal distribution or similarly equal variances because such limitations would “not apply to [Professor] Cohen’s thresholds themselves, but only to the calculations which permit this example of interpreting different effect sizes [such as the percent of overlap].” Id. at 19–20. In its discussion of the third “operational definition” of each threshold, which provided real-life examples in which small, medium, and large effects have been found, Commerce contended that these illustrative examples do not link Professor Cohen’s thresholds with the statistical criteria. Id. at 20. Commerce reiterated the results of its illustrative framework and stated that “although the statistical criteria may be used to determine whether the result of an analysis is representative of the full populations of data, it is not part of [Professor] Cohen’s proposed thresholds to qualify an effect as small, medium, or large.” Id. at 21.
In Stupp III, the CAFC raised concerns about the assumptions of population size and normalcy, questioning whether small sample sizes without normal distributions could produce an upward bias in the calculated effect size, and ultimately “exaggerate” dumping margins. Stupp III, 5 F.4th at 1358–59.
22
In the Fourth Remand Redetermination, Commerce reasonably explained that the Cohen’s d test does not apply a sampling methodology, but instead relies on the entire populations of sales observations and the “use of the statistical criteria to determine the statistical significance of the calculated results is not relevant for the Cohen’s d test or the differential pricing analysis as a whole.” Fourth Remand Redetermination at 22. Commerce noted that the purpose behind the three statistical criteria of normality of distribution, equal variances, and the size of the sample is to make samples more reflective of the population, which in turn increases the confidence level that the results are reflective of the whole population. Id. at 10. Commerce explained that:
If researchers examine the whole population, these assumptions become unnecessary, as there is no need to make a whole population more reflective of the population. To repeat, any adjustment to the population data serves only to distort the actual population’s parameters, making it no longer reflective of the whole population and reducing the 100 percent confidence level. Nowhere in the cited literature is there any mention of criteria or assumptions necessary when examining the entire population. In contrast, [Professor] Cohen’s thresholds do not depend on the subjective composition of a particular sampled population. The only references to assumptions are related to drawing a sample and to efforts to improve the probability that the samples are as representative of the whole population as possible.
Id. at 1011. Commerce determined that its application of the Cohen’s d test and Professor Cohen’s thresholds to the entire population of relevant price observations
23
does not require the application of the three statistical criteria identified by the CAFC in Stupp III. Id. at 11.
The CAFC specifically asked Commerce to explain why it can use the 0.8 threshold identified by Professor Cohen as a measure of significant price difference when Commerce evaluates data that fail to meet statistical assumptions of normality, size, and variance. Stupp III, 5 F.4th at 1360. The U.S. Court of International Trade previously sustained Commerce’s remand redetermination pursuant to the CAFC’s concerns regarding the use of the 0.8 threshold in Stupp Corp. v. United States (“Stupp IV”), 47 CIT __, __, 619 F. Supp. 3d 1314, 1328 (2023); see Final Results of Redetermination Pursuant to Court Remand [Pursuant to Order (Oct. 8, 2021)], Consol. Court No. 15-00334, ECF No. 208-1. The court concluded that “Commerce’s decision to adopt [Professor] Cohen’s 0.8 (‘large’) threshold as a measure of significance because it is widely accepted in the statistical literature does not undermine the reasonableness of that choice, if it is based on Commerce’s expertise and Commerce demonstrates the reasonableness of that choice with reference to the impact it has on the differential pricing analysis.” Stupp IV, 47 CIT at __, 619 F. Supp. 3d at 1327. Further, the court concluded that “Commerce’s reference to [Professor] Cohen’s work does not circumscribe its discretion to choose the same values in a new context, because that choice is itself reasonable.” Id. Because Congress delegated to Commerce the authority to
24
determine where a price difference is significant, 19 U.S.C. § 1677f-1(d)(1)(B)(i), and made clear that the definition of a “significant price difference” would depend on the product at issue, see SAA at 843, “Commerce’s choice of a measurement that is a function of standard deviation as a uniform approach to identify differences as significant is reasonable, even if the absolute difference in means is small.” Stupp IV, 47 CIT at __, 619 F. Supp. 3d at 1326.
Similarly, in the Fourth Remand Redetermination, Commerce explained that “although [the 0.8 threshold is] arbitrary, the proposed conventions [of Professor Cohen’s thresholds] will be found to be reasonable by reasonable people.” Fourth Remand Redetermination at 18 (citing Cohen at 13); see also SeAH’s June 12, 2023 Submission at Att. 1 (“Ellis”) at 41. Commerce noted that for an analysis based on the difference of the means, Professor Cohen proposed numerical thresholds to define a small, medium, and large effect, i.e., 0.2, 0.5, and 0.8 respectively. Fourth Remand Redetermination at 18. Commerce stated that Professor Cohen expected these numerical thresholds to be reasonable and argued that these thresholds have been widely accepted as recognized in Mid Continent. Id.
Commerce discussed scenarios in which the numerical thresholds present different real-life examples in which small, medium, and large effects have been found. Id. at 20. Commerce identified Professor Cohen’s examples involving the
25
differences in the IQs of various groups of people or the differences in the heights of various ages of teenage girls. Id. at 20 & n.58; see also Cohen at 27. Commerce noted that when the IQ data were collected, they were not collected from everyone in the group, but from a selected sample from the group. Fourth Remand Redetermination at 20. Commerce stated that:
The results of the analysis would have been calculated based on the sampled data from each group, and also, through statistical inferences, the representativeness of those results for the entire populations would have been determined. If the statistical analysis of the sample demonstrated that the sample-based results are representative of the population, then the sample-based results would be applied to the entire populations of Ph.D. holders and college freshmen. This use of statistical inference, however, is necessary to ensure that the sample is representative, but it was not part of [Professor] Cohen’s proposed small, medium, and large thresholds, which are numerical values that have been widely accepted in the academic community.
Id.
SeAH argues that Commerce has failed to provide an explanation that demonstrates that its use of the Cohen’s d test is reasonable. SeAH’s Br. at 16. The Court disagrees. Commerce explained that its analysis in the Cohen’s d test is to determine whether prices differ significantly between the sales to a specific purchaser, region, or time period (i.e., the test group), and all other comparable sales (i.e., the comparison group), and these sales prices include all of the sales prices that are used to calculate each respondent’s weighted-average dumping
26
margin and represent the full population of sales prices to each test and comparison group. Fourth Remand Redetermination at 2122.
As noted above, the question before the Court is whether Commerce’s use of the Cohen’s d test is reasonable, when applied as a component of its differential pricing analysis. See Ceramica Regiomontana, S.A., 10 CIT at 404–05, 636 F. Supp. at 966. The Court holds that Commerce has adequately explained how its methodology is reasonable.
CONCLUSION
For the foregoing reasons, Commerce’s Fourth Remand Redetermination is supported by substantial evidence, in accordance with law, and therefore is sustained. Judgment will enter accordingly.
/s/ Jennifer Choe-Groves
Jennifer Choe-Groves, Judge
Dated: December 18, 2023 New York, New York