The Court of International Trade has sustained Commerce’s replacement for the discredited Cohen’s d test, leaving Canadian softwood margins essentially unchanged and validating a methodology that now reaches every antidumping case in the American system
NEW YORK / OTTAWA, July 30, 2026 – The United States Court of International Trade has upheld the Commerce Department’s new statistical method for detecting targeted dumping, rejecting a challenge brought by the Government of Canada, four provincial governments, and virtually every significant Canadian softwood lumber producer, and leaving the disputed dumping margins from the 2021 administrative review almost exactly where they were before the litigation began.
In Slip Opinion 26-82, dated July 27, 2026 and reported the following day, Judge Jennifer Choe-Groves sustained Commerce’s Remand Redetermination in Consolidated Court Number 23-00187. The decision closes a chapter of a dispute that has run through the Federal Circuit and back, and it does so on terms that will disappoint Canadian mills in the narrow case while carrying implications for antidumping respondents from every country in every industry.
The outcome, in numbers: Canfor Corporation’s weighted-average dumping margin held at 5.25 per cent. West Fraser Mills Ltd. moved from 6.96 per cent to 7.06 per cent, a ten basis point increase. The rate for non-selected companies moved from 6.20 per cent to 6.26 per cent. Commerce discarded the statistical test the Federal Circuit had condemned, built a new one, applied it, and arrived at essentially the same place.
How the Case Reached This Point
The proceeding began with an ordinary administrative review. In March 2022, Commerce initiated a review of the antidumping duty order on softwood lumber from Canada covering calendar year 2021. In the Final Determination published at 88 Federal Register 50,106 on August 1, 2023, Commerce assigned weighted-average dumping margins of 5.25 per cent to Canfor and its affiliates, 6.96 per cent to West Fraser and its affiliated mills including Blue Ridge Lumber, Manning Forest Products, and Sundre Forest Products, and 6.20 per cent to non-selected companies.
To reach those numbers, Commerce used its differential pricing analysis, and at the heart of that analysis sat the Cohen’s d test.
The statutory background matters for understanding the fight. Under 19 United States Code section 1677f-1(d)(1)(A), Commerce normally determines dumping by comparing weighted-average normal values to weighted-average export prices, the average-to-average or A-to-A method. Subsection (d)(1)(B) creates an exception: Commerce may compare average normal values to individual transaction prices, the average-to-transaction or A-to-T method, if there is a pattern of export prices that differ significantly among purchasers, regions, or periods of time, and Commerce explains why those differences cannot be accounted for using the standard method.
Congress created that exception, according to the Statement of Administrative Action accompanying the Uruguay Round Agreements Act, because averaging could conceal what the statute’s drafters called targeted dumping: low prices to some customers offset by high prices to others, netting out to an innocuous average. The A-to-T method exists to unmask that pattern. Because A-to-T calculations typically produce higher margins than A-to-A, the choice of method is often worth more to a respondent than any individual factual dispute in a review.
Commerce operationalized the exception through a three-step differential pricing analysis. Step one asked whether prices differed significantly, and for years the tool used at that step was Cohen’s d, a statistical effect-size measure borrowed from the social sciences. Step two, the ratio test, measured how much of the respondent’s sales volume exhibited the pattern. Step three, the meaningful difference test, asked whether switching methods actually changed the margin materially.
In Marmen Inc. v. United States, reported at 134 F.4th 1334 and decided in 2025, the Court of Appeals for the Federal Circuit vacated and remanded, holding that Commerce could not rely on the Cohen’s d test for data sets of the kind at issue. The problem, in essence, was that Cohen’s d carries statistical assumptions about normal distribution, comparable variance, and sample size that do not hold for the full-population, non-random sales data Commerce works with in antidumping proceedings. The Federal Circuit did not forbid Commerce from using ideas underlying Cohen’s analysis of group differences. It required that whatever Commerce used be justified as sound for the data actually in front of it.
The Court of International Trade remanded this case for compliance in June 2025. Commerce responded by retiring Cohen’s d entirely and constructing a replacement.
The Price Difference Test
What Commerce built is considerably simpler than what it replaced, and simplicity is the source of both its administrative appeal and the Canadian objection.
Under the new price difference test, Commerce examines whether the weighted-average net price to a given purchaser, region, or time period falls within two per cent of the weighted-average net price to all other purchasers, regions, or time periods. If it falls outside that band in either direction, the prices to that purchaser, region, or period are found to differ significantly and those sales pass the test.
Purchasers are defined by reported consolidated customer codes. Regions are defined by destination ZIP code, grouped using United States Census Bureau definitions. Time periods are calendar quarters within the period of review. Comparable merchandise is defined by product control number.
The ratio test then calculates the share of total United States sales value that passed the price difference test. If 33 per cent or less passes, A-to-T is not considered. If more than 33 per cent passes, Commerce finds a pattern of prices existed and proceeds to step three.
The meaningful difference test asks whether using A-to-T yields a meaningfully different margin than A-to-A. A difference is meaningful if there is a 25 per cent relative change between the two calculations where both rates exceed de minimis, or if the two calculations fall on opposite sides of the de minimis threshold.
Applied to the 2021 review, the results were lopsided. Commerce found that 99.40 per cent of the value of Canfor’s United States sales passed the price difference test, and 99.82 per cent of West Fraser’s. Both figures obviously cleared the 33 per cent ratio threshold. At step three, Commerce determined that the margins crossed the de minimis threshold depending on method, so A-to-A could not account for the differences, and A-to-T applied.
Commerce also abandoned the mixed methodology it had previously used, under which it applied A-to-A to some sales and A-to-T to others.
The Canadian Argument, and Why It Lost
The Plaintiff Canadian Parties, a coalition comprising the Government of Canada, the Governments of Alberta, Ontario, and Quebec, the British Columbia Lumber Trade Council, the Conseil de l’industrie forestière du Québec, the Ontario Forest Industries Association, and producers including Canfor, West Fraser, Interfor, Fontaine, Carrier, Olympic, Resolute FP Canada, Tolko, Gilbert Smith, Chaleur, J.D. Irving, Delco, Devon Lumber, H.J. Crabbe and Sons, Langevin, Marwood, North American Forest Products, and Twin Rivers Paper, advanced several lines of attack.
Their threshold argument was about the standard of review. Canada contended that after the Supreme Court’s decision in Loper Bright Enterprises v. Raimondo, 603 United States 369 (2024), which ended Chevron deference, the question was not whether Commerce’s reading of the statute was reasonable but whether it was the best reading. That distinction, if accepted, would have shifted the whole analysis.
The court rejected it. Citing Stupp Corp. v. United States, 5 F.4th 1341 (Federal Circuit 2021), Judge Choe-Groves held that “the relevant standard for reviewing Commerce’s selection of statistical tests and numerical cutoffs is reasonableness, not substantial evidence,” and noted that the Federal Circuit itself had applied a reasonableness standard in Marmen III after Loper Bright was decided. The court reviewed the price difference test for reasonableness.
On the merits, Canada argued that a flat two per cent threshold ignores the context in which prices exist and defies Congress’s instruction, expressed in the Statement of Administrative Action, that Commerce proceed case by case “because small differences may be significant for one industry or one type of product, but not for another.”
Commerce’s answer was that the test is context-sensitive in the only way that matters: the comparison is always against the respondent’s own prices to all other purchasers, regions, and time periods, so the data underlying the test changes with every case. Commerce also pointed to its use of a two per cent threshold elsewhere, including the arm’s-length test under 19 Code of Federal Regulations section 351.403(c) and the statutory de minimis threshold for dumping margins.
Canada raised a pointed historical objection. In Certain Steel Nails from the United Arab Emirates, 73 Federal Register 33,985 (2008), Commerce had rejected a two per cent test of this general character, explaining that such a bright-line threshold “does not account for price variations specific to the market in question.” Commerce distinguished that precedent on the ground that the earlier Nails Test derived from the P/2 test and looked only at whether prices to alleged targets were at least two per cent lower than others, whereas the price difference test asks whether prices to any purchaser, region, or period are at least two per cent higher or lower. “The only common aspect of the P/2 test and the price difference test is the two percent threshold,” Commerce wrote in the Remand Redetermination.
Canada also mounted a market-realities argument that will resonate with anyone who has watched lumber futures. Prices in the American softwood market are volatile and move for reasons wholly outside any exporter’s control. Treating quarter-to-quarter price variation as evidence of significant price differences among purchasers and regions, Canada argued, is not supported by substantial evidence, and the fact that nearly 90 per cent of Commerce’s applications of the price difference and ratio tests produce a finding of pattern suggests the test is detecting market volatility rather than targeting behaviour. Market-specific evidence of volatility, Canada argued, “dispels the notion that targeted dumping is occurring at all.”
Commerce replied, and the court accepted, that under JBF RAK LLC v. United States, 790 F.3d 1358 (Federal Circuit 2015), section 1677f-1(d)(1)(B) does not require Commerce to determine why a pattern of differing prices exists, and does not impose an intent requirement. Commerce’s position is that step one identifies a pattern and step three measures whether the pattern conceals dumping the standard method cannot capture. The reason for the pattern is not an element.
On the abandonment of the mixed method, the court found the statutory text dispositive. Section 1677f-1(d) refers to A-to-A and A-to-T and makes no reference to simultaneous application of both. Neither does the Statement of Administrative Action. That silence supported Commerce’s decision to discontinue the practice.
The court noted that the Federal Circuit had already upheld the ratio test and its 33 per cent and 66 per cent cutoffs in Stupp, and had upheld the meaningful difference test in Apex Frozen Foods Private Ltd. v. United States, 862 F.3d 1337 (Federal Circuit 2017). It also observed that the new price difference test had already been sustained in Toyo Kohan Co. v. United States, Slip Opinion 26-54, decided May 22, 2026, and in Marmen IV, decided June 15, 2026.
“Commerce’s Remand Determination was reasonable,” Judge Choe-Groves concluded, sustaining the differential pricing analysis in full.
The Wider Significance
Canadian mills lost roughly ten basis points on one respondent in one review year. That is not the story.
The story is that Commerce has now had its replacement for Cohen’s d validated three times at the Court of International Trade in ten weeks, across unrelated industries. The price difference test is not a lumber methodology. It is the first step of the differential pricing analysis in every antidumping proceeding Commerce conducts. When the Federal Circuit struck down Cohen’s d in Marmen III, respondents across dozens of orders saw an opening. Toyo Kohan, Marmen IV, and now this softwood decision have largely closed it.
Two features of the new test make it, from a respondent’s perspective, worse than what it replaced. First, a flat two per cent band is a low bar in any market with meaningful price dispersion, and the 99-plus per cent pass rates recorded here for Canfor and West Fraser illustrate the point starkly. Second, because the test is mechanical rather than statistical, the sophisticated attacks that succeeded against Cohen’s d, arguments about distributional assumptions and variance and sample size, have no obvious analogue. There is not much statistics left to challenge.
The Loper Bright question is the live one. Canada’s argument that post-Chevron review requires the best reading of “differ significantly” rather than merely a reasonable one is not frivolous, and the tension between Loper Bright and the Stupp line of reasonableness cases in the technical-methodology space will need appellate resolution eventually. Given the number of parties and the amounts involved, a Federal Circuit appeal in this or a parallel case is a realistic prospect.
Stacked on Everything Else
This ruling lands on an industry already carrying an unusual accumulation of American trade measures.
Combined antidumping and countervailing duties plus tariffs hold the effective rate on most Canadian softwood lumber producers at 34.83 per cent, according to figures reported by Wood Central. In October 2025 Washington added a 10 per cent Section 232 duty on softwood timber and lumber, which applies even to shipments meeting CUSMA rules of origin and stacks on top of the trade remedy duties. That Section 232 measure sits entirely outside the CUSMA framework and requires its own settlement; Commerce’s final determination on the associated duties is not expected until late August.
The trade remedy litigation itself has been busy. On July 22, 2026, the Court of International Trade sustained countervailing duty orders affecting Canadian softwood producers after Commerce’s fourth attempt at the rate calculation, resolving that strand of the dispute. Six days later came the antidumping decision in Slip Opinion 26-82. Canadian producers lost both.
The commercial consequences are visible. Canadian producers supply roughly 75 per cent of the softwood timber and lumber the United States imports, which makes them the most exposed suppliers to any American measure on wood. Softwood imports into the United States fell 28 per cent in the year to January as stacked duties priced Canadian mills out of the market, pushing British Columbia operations into curtailment and closure.
Ottawa and British Columbia have written a deadline into their own response, committing to softwood support measures by December 1. Premier David Eby has pressed the federal government to treat the dispute as a national priority while mills across his province shed shifts.
One point of relief: softwood and lumber were not among the categories hit by the July 20 Section 338 proclamations imposing an additional 50 per cent duty on approximately $20 billion in Canadian goods effective August 19. Wood products more broadly were included, so the distinction between covered and uncovered lines matters intensely for firms shipping mixed product ranges.
LeBlanc named lumber alongside steel, aluminum, and autos as the sectoral tariffs Canada wants rolled back at the fifth CUSMA joint review on July 1. Nothing has moved since.
What This Means for Importers and Exporters
For American importers of Canadian softwood, the practical effects are immediate and specific.
Cash deposit rates on the affected respondents are essentially confirmed at the revised levels, subject to appeal. Importers of Canfor product see no change at 5.25 per cent. West Fraser product moves to 7.06 per cent and non-selected suppliers to 6.26 per cent. These are small movements, but they are movements in the wrong direction for anyone who financed the litigation hoping for relief, and they apply to entries during the 2021 period of review that have been suspended pending resolution. Final assessment and liquidation instructions will follow, and importers should reconcile their accrued liability against actual deposits to identify any shortfall.
Anyone modelling landed cost on Canadian softwood should be building the full stack: the antidumping rate applicable to the specific producer, the countervailing duty rate, and the 10 per cent Section 232 duty. Producer-specific rates differ materially, and supplier selection is now a duty-rate decision as much as a price and quality decision.
For respondents in other antidumping proceedings, the message is broader and more important than the lumber numbers. The price difference test is the operative step-one methodology going forward and it has now been sustained in three separate cases. Respondents should assume that in any market with normal price dispersion across customers, regions, or quarters, the pattern finding will be made and the analysis will turn on the meaningful difference test at step three. Litigation strategy and pricing strategy both need to account for that. Firms whose United States pricing varies by more than two per cent across customer tiers, geographic regions, or quarters should understand that this variation, however commercially ordinary, now reliably triggers the targeted dumping analysis.
For Canadian forest products companies, the strategic calculus has not changed but the odds have worsened. Litigation before American courts has produced four remands on the countervailing duty side and a rebuilt methodology on the antidumping side, and the cumulative effect on rates has been marginal. The durable solution to the softwood dispute has always been a negotiated one, and negotiation is currently subordinate to a broader bilateral confrontation with an August 19 deadline of its own. The federal and British Columbia commitment to measures by December 1 is now the more relevant date for most operators.
Outlook
Appeals are likely. The Loper Bright standard-of-review question is genuinely unresolved in this corner of administrative law, and the coalition of governments and producers assembled in this case has both the resources and the incentive to take it up. The Federal Circuit has already shown, in Marmen III, that it will scrutinize Commerce’s statistical choices seriously.
But respondents should be clear-eyed about what a win would look like. Marmen III eliminated Cohen’s d and Commerce replaced it within a year with something that produced nearly identical results. A successful challenge to the price difference test would likely produce a fourth methodology rather than a different outcome. Commerce has demonstrated considerable durability in reaching the destinations it intends to reach.
For now, the effective rate on most Canadian softwood entering the United States stands at 34.83 per cent, Commerce’s final Section 232 determination is due late August, Ottawa and Victoria have promised measures by December 1, and the methodology that generates the antidumping component of that burden has been upheld.