Commerce's 2% Test Could Anchor a Targeted Dumping Rule
Commerce may use Marmen's 2% remand test in a targeted-dumping proposal that could also codify Commerce's abandonment of mixed A-to-A and A-to-T calculations. Companies with AD exposure should test sales data and prepare positions before any NPRM appears.
Primary lensTrade policy
Sub-topicPolicy monitoring
Evidence base11 records used
Use casePolicy monitoring
Commerce may bring a 2% price test and a one-third ratio structure into its next targeted-dumping proposal. On the Marmen remand, a price band of 2% and a pass ratio just above one-third opened the way to full average-to-transaction treatment. The calculated margin moved from 1.22% under the standard method to 2.93% under the alternative, crossing the 2% de minimis threshold for an antidumping investigation.
The U.S. Court of International Trade sustained Commerce's explanation and application on that record on June 15. The now says Commerce is considering whether to codify or revise its use of the method in investigations and certain reviews. Its target for a notice of proposed rulemaking, or NPRM, is July 2026. That is a planning marker with no legal deadline. No proposed text or comment period existed as of July 15.
Companies with active or expected antidumping exposure have a reason to run one focused sensitivity test now. The priority group includes respondents with margins near the applicable de minimis line, companies that depend on offsets from non-dumped sales, and businesses whose customer, destination, date, or product coding changes across systems. The output should be a documented sales-data test and a short list of provisions the company could support, oppose, or replace if Commerce publishes an NPRM.
Marmen Gives Commerce a Three-Part Rulemaking Prototype
In an investigation, 19 USC 1677f-1(d)(1)(B) permits the average-to-transaction method only after two findings. Commerce must identify a pattern of export prices that differ significantly among purchasers, regions, or periods of time. It must also explain why average-to-average or transaction-to-transaction comparisons cannot account for those differences. The current regulation defines the methods. It does not codify the complete differential pricing sequence Commerce has built through agency practice.
Commerce formerly used Cohen's d to identify significant differences among price groups. Its ratio test measured how much of the respondent's U.S. sales passed. A meaningful-difference test then compared the margins produced by average-to-average, known as A-to-A, and average-to-transaction, known as A-to-T.
The Federal Circuit's April 2025 Marmen opinion left the statutory A-to-T authority intact but rejected Commerce's use of Cohen's d for data sets like Marmen's. The court identified normal distribution, equal variability, and equal and sufficient numerosity as assumptions associated with Cohen's d. Commerce could develop another sound way to gauge the price differences.
Commerce did so on remand. For comparable merchandise, it compared the weighted-average net price for a purchaser, region, or quarter with the weighted-average net price for all other members of that group. A price outside a band of 2% above or below the comparison average passed. Commerce then calculated the share of U.S. sales value that passed and asked whether the margin changed meaningfully when it switched methods.
The June 2026 Marmen decision sustained that approach on the second-remand record. The trade court had already sustained Commerce's post-Marmen price difference test in the May 2026 Toyo Kohan decision, which arose from an administrative review. The agency now has trial-court decisions in an investigation and a review. Neither decision is a regulation, a binding Federal Circuit precedent on the new test, or proof that Commerce will propose the same method nationally.
Who Should Spend Money Before Commerce Publishes
An executive should fund a sensitivity run when the company has an active investigation, expects a petition, or faces a review in which Commerce may use a parallel differential pricing analysis. The case for doing it is stronger when a recent margin sits near de minimis or when A-to-A offsets materially reduce the result. A distant possibility of antidumping exposure is not enough on its own.
The legal work product is a position paper, not a forecast. Counsel should identify what the company needs a proposal to say about significant price differences, the ratio threshold, group size, the mixed method, offsets, and transition. In investigations, those positions should address the two statutory findings. In reviews, counsel should separately address 19 CFR 351.414(c)(1) and Commerce's use of analogous differential pricing practice. Treating the authorities as identical would invite a weak comment record.
The trade or data team needs a reproducible bridge from ordinary systems to the fields Commerce used in Marmen. Customer master records should map to consolidated customer codes. Shipment destinations should map to regions. Contract and invoice events should support dates of sale. Product data should support control numbers. Each mapping needs an owner and a source.
A regulatory owner should monitor Regulation Identifier Number, or RIN, 0625-AB27 and the Federal Register. There is no filing calendar yet. The trigger for comments is an actual notice that supplies proposed text and a deadline.
Ending the Mixed Method May Matter More Than 2%
The 2% band is only the first gate. In Marmen, 38.18% of the value of U.S. sales passed. Commerce treated more than 33% as enough to establish a pattern and reach the meaningful-difference inquiry. The calculation then produced 1.22% under A-to-A and 2.93% under A-to-T. Crossing the 2% investigation threshold satisfied one prong of Commerce's meaningful-difference test.
The other prong applies when both margins are above de minimis and the relative change between them is at least 25%. A future proposal should make both paths visible. Focusing only on the initial price band would miss the decision that connects the screen to the final comparison method.
The ratio structure changed too. Under the practice Commerce described in its 2014 request for comments, a result at 33% or less did not support an alternative method. More than 33% but less than 66% could produce a mixed calculation. Commerce used A-to-T for the passing sales and A-to-A for the rest. A result of 66% or more could support A-to-T for all sales.
On the Marmen remand, Commerce removed that middle category. It told the trade court that it had discontinued the mixed method as an administrative practice for all of its administrative proceedings. Under the practice Commerce described, a pass ratio just above 33% can lead to full A-to-T once the meaningful-difference condition is met. The trade court sustained that choice in Marmen. Neither the agency statement nor the decision turns it into regulatory text.
That combination may be more consequential than 2% alone. The price band identifies groups, but full A-to-T depends on the later ratio and method choice. Once Commerce removed the middle category, a result just over one-third could reach every sale after the meaningful-difference step. A proposal that codifies only the first number would leave the larger method choice outside the rule.
A Regulation Would Move the Fight Upstream
Commerce's differential pricing method grew through notices and explanations in individual proceedings. The 2014 comment request described the practice but did not put the whole sequence into 19 CFR 351.414. Companies often challenged the method after a record had formed and Commerce had calculated a margin.
That 2014 episode is the closest institutional comparison, but the direction is different. Commerce then described a method it was already using and asked for views while continuing to develop the approach case by case. The current Agenda contemplates regulatory text after the Federal Circuit rejected a central test and the agency replaced it on remand. The earlier notice helps explain the old 33% and 66% structure. It does not supply the justification for a new 2% band or for eliminating the middle category. Commerce would have to build that record in any NPRM.
Regulatory text would move the foundational dispute into notice and comment. A proposal could define the comparison groups, price measure, numerical band, minimum group size, and treatment of sparse or skewed data. It could determine whether one threshold applies across products or whether Commerce retains room for industry-specific significance. It could also address the ratio line and explain when full A-to-T replaces A-to-A.
Once a final rule fixes those choices, case records will still matter. A respondent can dispute customer coding, region assignment, date-of-sale fields, product matching, or Commerce's application of the rule. The broader challenge would point back to the regulatory text and the record Commerce assembled when it adopted that text.
A useful comment therefore needs more than an objection to 2%. It should test the proposal against actual transaction structures, show how results behave when groups are small or prices cluster tightly, and offer an administrable alternative. Domestic producers can use the same record to show when averages mask price patterns that affect a margin. The comment period may be the best chance to contest the architecture before a company has a case-specific rate at stake.
A Fixed 2% Rule Would Need an Industry Escape Valve
The proposal would also have to confront the tension between a fixed number and a case-specific inquiry. The Marmen court accepted Commerce's explanation that 2% was reasonable on that remand. The Statement of Administrative Action discussed in the opinion says small differences can matter for one industry or product but not another. A national default may offer consistency, while a flexible test may better reflect different pricing markets. Either choice needs a rule that tells parties what evidence can move Commerce away from the default.
An escape valve would matter only if parties can use it before the margin calculation is settled. Commerce would need to identify who bears the burden, what market evidence can justify another threshold, and where the decision appears on the record. The rule should also say whether Commerce makes that decision before or after it sees the margin effect. Too little guidance would turn flexibility into another case-specific dispute. Too much rigidity would ignore the industry differences that Congress expected Commerce to consider.
The Agenda Combines Three Policy Tracks Across Two Notice Records
The planned package is wider than targeted dumping. Its title refers to unaffiliated reseller assessment rates and the possible modification or removal of expedited countervailing duty, or CVD, reviews. Its abstract adds the possible codification or revision of A-to-T.
Those issues began in separate public records. Commerce's June 2025 advance notice under RIN 0625-AB27 covered reseller rates and expedited CVD reviews. It did not ask about A-to-T or Cohen's d. The reseller question concerns entries exported by an unexamined unaffiliated reseller that did not obtain its own review, including the current-policy scenario in which the reviewed producer did not know the merchandise was destined for the United States.
Commerce opened the statistical question in a separate May 2025 notice under another docket. The notice requested alternatives to Cohen's d after Marmen and closed before the reseller docket. A company that followed one record may not have addressed the others.
The three policy tracks determine which rate applies or how Commerce calculates it, but they entered the current Agenda through two notice records. The combined Agenda entry raises a record-integration risk, not a present defect in notice. Any NPRM would need enough proposed detail on each component to permit meaningful comment. It should also identify which earlier comments Commerce is using for which choice.
The agenda announces intended work. Only an NPRM can show whether Commerce will propose one integrated rule, separable amendments, the Marmen test, or another approach.
The Marmen Data Fields Need a Sensitivity Run
On the Marmen remand, Commerce treated reported consolidated customer codes as purchasers, grouped destination ZIP codes under Census regions, and used quarters for time periods. Product control numbers and the characteristics used for price comparisons defined comparable merchandise. Net U.S. prices supplied the measure for the 2% band.
Small coding choices can alter the groups. Two codes for one commercial buyer can split a pattern. One code shared by distinct buyers can combine it. A billing destination may produce a different region from the delivery point. A corrected date can move a sale across a quarter. Detailed product characteristics can create thin comparison cells.
A sensitivity run should first reproduce the Marmen definitions. Variants can then change one assumption at a time without being mistaken for Commerce policy. Useful comparisons include billing location against delivery location, monthly periods against quarters, and a uniform band against a measure that accounts for product volatility. The record should show which changes affect the pass ratio or method outcome.
This test will not predict a future margin. It will show where the company's data definitions and commercial structure create a policy position worth presenting. The same exercise can expose weak source fields before an investigation questionnaire makes them part of an official record.
The sensitivity run should preserve failed tests as well as successful ones. If changing a location field or period definition does not affect the pass ratio, that stability is useful evidence. If a small mapping change flips the method outcome, the team should trace the commercial reason and the source record before drawing a policy conclusion. The point is to separate a genuine price pattern from an artifact of how the sales file was assembled.
Investigations and Reviews Need Separate Answers
The Agenda says Commerce is considering A-to-T changes for investigations and certain reviews. That phrasing carries a legal distinction the proposal will need to respect.
The two conditions in 19 USC 1677f-1(d)(1)(B) expressly govern investigations. In reviews, the regulation says Commerce normally uses A-to-A unless the Secretary determines that another method is appropriate in a particular case. Commerce has applied a parallel differential pricing analysis in reviews as agency practice.
A single test might operate in both settings, but Commerce should identify the authority and explanation for each. It should say whether the same 2% band, ratio threshold, and meaningful-difference test would apply. It should also explain whether ending the mixed method is part of the proposal or remains an administrative practice.
The distinction affects transition too. A new investigation begins with data requests built for a developing record. An administrative review may involve a familiar reporting system and an established deposit or assessment history. If Commerce changes the comparison method in both settings, parties need to know whether existing questionnaires collect every field the new test requires and whether a changed method applies to periods already underway.
Offsets require a separate answer. A-to-T describes a comparison method. Denial of offsets, commonly called zeroing, is another methodological choice. The Agenda does not say Commerce will codify zeroing. Any proposed text should reveal whether offsets are addressed directly or left to agency practice and later case records.
That division is the useful line to hold before any proposal appears. A company can build one reliable sales record now, but its eventual investigation and review arguments should not be collapsed into a single legal theory.
From reading to review
Run the numbers on your lane.
The duty calculator runs the current stack for any HTS code and origin. A free account opens full tool output, AD/CVD detail, Chapter 98 processing, and available exports.