The Cohen’s D Test used as part of the differential pricing analysis (‘DPA’) conducted by the U.S. Department of Commerce (‘USDOC’) to unmask targeted dumping was analyzed in our previous article (see here). The present article discusses the recent judgement of the U.S. Court for Federal Circuit in Marmen Inc. v. United States[1] wherein the court ‘invalidated’ the application of Cohen’s D test and the implications of the ruling for exporters to the U.S. market. In the context, article traces the origins of this methodology in the WTO’s Anti-Dumping Agreement, its evolution over the years in USA’s context, and the road ahead after the Marmen Inc. judgement.
WTO’s provisions – The origins of DPA
One of the cornerstones of an anti-dumping investigation is the computation of a dumping margin which is the rate at which an exporter is dumping its goods to the importing country. Article 2.4.2 of the WTO’s Anti-Dumping Agreement (‘ADA’) equips an investigating authority with a toolkit of methodologies for the calculation of dumping margins. While the methods normally expected to be employed are ‘weighted average-to-weighted average (A-A)’ or ‘transaction-to-transaction (T-T)’, the Article carves out a specific exception allowing for a calculation on the ‘weighted average-to-transaction (A-T)’ basis. The preconditions being –
* Pattern Clause: The Authority finds a pattern of export prices which differ significantly among different purchasers, regions or time periods, and
* Explanation Clause: The Authority provides an explanation as to why the other two methods cannot account for these differences.
The WTO provisions provide for this third method (A-T) to counter the selective dumping focused on specific markets, customers or points in time. Investigating Authorities globally employ various methods to identify targeted dumping, and one of the most well-known statistical techniques employed by the USDOC for this purpose is the ‘Cohen’s D Test’.
Evolution of targeted dumping analysis – The past
The USDOC derives its authority to conduct DPA from the Tariff Act, 1930 (‘Act’) which is like Article 2.4.2 of the ADA. The previous US Federal Circuit judgements (Apex Frozen Foods Private Ltd. v. United States) have upheld the broad principles applied by the USDOC’s in its DPA to the dumping margin calculations as legally sound and based on reasonable and sound interpretation.
Previously in the US, Petitioners were required to allege targeted dumping by exporters for the USDOC to undertake the DPA, absent which the USDOC followed the usual A-A methodology. From December 2008, the regulatory provisions governing DPA were amended in favor of a finer approach to extensively analyze target dumping through the ‘Nails Test’[2] to identify significant differences in pattern of export prices among purchasers, regions or periods in time. The test comprised two main steps –
* Standard Deviation Test: to evaluate differences (product-specific) in standard deviation of weighted-average prices of alleged target group v/s all transactions.
* Gap / Significance Test: to evaluate the gap in weighted-average prices between the sales to target group and next highest non-target group v/s average gap in weighted-average prices between the non-targeted groups.
If the two-step analysis confirmed the allegation of targeted dumping along with requisite sufficiency, the USDOC evaluated differences between A-A method and A-T method. If the differences in results were meaningful, the dumping margin was calculated using the A-T method for all export sales.
The beginning of Cohen’s D
Despite the Nails Test being in line with Section 777A(d)(1)(B) of the Act, the USDOC had been keen on continually refining its methodologies by applying alternative methods which are more characteristic of the requirements under section 777A(d)(1)(B) of the Act. Under the new approach, the USDOC, instead of seeking an allegation from the Petitioner, itself started conducting the analysis in each anti-dumping investigation. The Xanthan Gum from China and Xanthan Gum from Austria are two illustrative investigations wherein the USDOC initially applied the new approach.
The first stage (price pattern determination) of DPA involves using the ‘Cohen’s D Test’ and the ‘Ratio Test’. The Cohen’s D test is a popular statistical measure of the extent of the difference in the means between a test group and a comparison group. It evaluates the extent of these differences which are then classified as - small, medium or large, with ‘large’ variance in export prices providing the strongest indication of targeted dumping. Subsequently, in ‘Ratio Test’, the Department assesses how much value of the export sales exhibit significant differences as measured by the Cohen’s D, and this is grouped in following categories –
* 66% or more, where A-T applied to all sales
* More than 33%, and less than 66%, where a mix of A-T and A-A is applied
*33% or less, where A-A is applied to all sales.
In the second stage (Meaningful Difference Test), the Department evaluates the differences in dumping margins using the A-A method and other alternative methods (A-T or mixed). Where the result of applying DPA changes the dumping margins by 25% or more, or where the difference makes a margin jump beyond ‘de minimis’ threshold, it is considered meaningful.
The above DPA method has been extensively employed by the USDOC for over a decade, and the approach has received broad legal reinforcement from several judgements of the Federal Circuit Courts (e.g., Apex Frozen Foods Private Ltd. v. United States).
The ‘Paradox’ of Zeroing: Difference between A-A and A-T methods
The USDOC resorts to zeroing the negative margins when it employs A-T method, whereas zeroing is not done in the case of the A-A method. This zeroing results in higher dumping margins because the export transactions of an exporter during the investigated period which have high export prices (and therefore, negatives dumping margins) are ignored or, in other words, the negative margins in those transactions are set to ‘zero’. The zeroing approach of DPA has been challenged at the WTO numerous times with most cases directed at the USDOC’s practice. The WTO Appellate Body as well as Panels have mostly held that zeroing is inconsistent with Article 2.4.2 of the ADA, although in some cases the Panel has taken a contrary view.
In particular, the Panel in US-Softwood Lumber (Canada) partially deviated from the long-held view of the WTO Panels and AB (US – Washing Machines and US – Anti-Dumping Methodologies (China)) on zeroing. The Panel re-affirmed the prior decisions that in case of differences in pattern of export prices, A-T methodology must be applied. However, it deviated in case of ‘non-pattern transactions’ and held that all export transactions must be considered to properly assess the pricing behavior of foreign exporter. In effect, the Panel aligned with the USDOC’s practice and held that mixed methodology involving A-T be applied for ‘patterned-transactions’ and A-A for ‘non-patterned-transactions’. The Panel reasoned that if zeroing is prohibited in all cases, it may lead to a mathematical equivalence between A-T and A-A methods causing failure in unmasking targeted dumping.
Marmen Inc., judgement – Cohen’s D meets its fate
The initial observations calling into question the appropriateness of USDOC’s use of Cohen’s D test in certain situations was made by the Federal Circuit Court in Stupp Corp. v. United States.
Some of the key underlying assumptions of Cohen’s D test are - statistical assumptions of normal distribution, equal variability, and sufficiently numerous data. In the Stupp Corp. v. United States, the Court relied on extensive statistical literature to point out serious flaws in results based on Cohen’s D test when sufficiency of size, normality of distribution, homogeneity-of-variances were absent. The Court held that absence of any of the underlying assumptions in the export price data may undermine the usefulness of interpretative cut-offs prescribed under the test. The Cohen’s D test was found to produce an ‘upward bias’ i.e., tendency to overestimate and produce more ‘passing results’ which eventually result in higher dumping margins. The Court eventually remanded the case back to the USDOC to re-evaluate whether prescribed limits on Cohen’s D test were satisfied in the case or whether those limits are required to be observed in anti-dumping investigations.
Marmen Inc., judgement begins where the Stupp Corp. v. United States ended i.e., whether it is unreasonable for the USDOC to apply Cohen’s D as part of its DPA on data that doesn’t satisfy the underlying statistical assumptions.
The Department’s arguments – first, underlying assumptions are irrelevant as the Department uses full population and not sample sales prices. Second, Cohen’s D test is relied not to calculate the weighted-average dumping margin but only to determine if significant patterned variation in export prices exists. The first argument was rejected citing statistician Jacob Cohen’s insistence on the necessity of underlying assumptions remaining true. The Court held that the test determines ‘effect size’ and the same is independent of whether a sample is taken or an entire population. The essential characteristics remain critical - normal distribution, equal variability, and sufficiently numerous data.
Therefore, the Court held that absent underlying assumptions, the Cohen’s D test cannot be utilized to conduct ‘pattern analysis’.
The road ahead: Case in point - Certain new Pneumatic Off-the-Road Tires from India
As on date, the decision in Marmen Inc. is not final and conclusive, as there remains a possibility of rehearing and/or appeal, and the Court’s mandate has not yet been issued. Under the Federal Rules of Appellate Procedure, the opinion of the appeals court is not final until it issues its mandate. However, the USDOC has already started to adopt alternative measures in its DPA. For example, in the Preliminary Decision Memo issued in the latest Admin Review (AR) for OTR – India,[3] the Department has discontinued the use of the Cohen’s D test as well as of the ‘mixed method’ under its DPA.
Instead, the USDOC relied upon ‘Price Difference Test (PDT)’ and ‘(modified) Ratio Test (MRT)’. In the PDT, the USDOC examined whether the weighted-average net export price from a foreign producer/exporter to a given purchaser, region or time-period is within +/- 2% of the weighted average net export price to all other purchasers, regions or time periods in the US. If the difference were beyond +/-2 % threshold, the export prices were identified as differing significantly and hence, passing the test. Next, in the case of the MRT, if more than 33% of the total export transactions in terms of sales value pass PDT, then the Department concludes that a differential pattern in export prices exists for applying the A-T method (with zeroing). Thereafter, the Department performs the ‘meaningful difference test’ i.e., whether using the A-T method yields a meaningful difference in the weighted-average dumping margin as compared to that resulting from the use of the A-A method.
Conclusion
The observable consequence of the latest ruling and consequent changes of methods by USDOC clearly leads to a possibility of higher dumping margins for the exporters going ahead. This is because the threshold of +-2% price differential seems to be very low compared to 0.8 (large deviation) under the Cohen’s D. Further, USDOC no longer follows the mixed approach, and the dumping margins are computed basis either extremes A-A or A-T. This increases the possibility that more transactions may be subject to ‘zeroing’ and thereby inflating in dumping margins. The exporting businesses to the US must plan their exports in a manner that they are not caught dumping heavily to the US under the revised methodology. This requires a thorough analysis of the export prices variations between purchasers, regions and time periods and strategizing the ways to minimize the dumping margins.
‘Targeted dumping’ has remained one of the most contentious issues of international trade investigations and is expected to remain so in the foreseeable future. The evolving nature of trade, global geo-political dynamics, and the USDOC’s adamancy in resorting to ‘zeroing’ through novel statistical techniques is likely to keep the chase going. It is exceedingly important for exporters to keep track of the USDOC’s moves proactively and ensure proper up-to-date methods are employed in their simulation margin runs.
[The authors are Executive Partner and Principal Associate, respectively, in International Trade and WTO practice at Lakshmikumaran & Sridharan Attorneys, New Delhi]
[1] Marmen Inc. v. United States, 2025 U.S. App. LEXIS, 9506 (Fed. Cir. 22 April 2025)
[2] The specific details and intricacies of this test are beyond the scope of this article.
[3] Certain New Pneumatic Off-the-Road Tires from India (A-533-869), 89 FR 23973, 5 April 2024.
