In Marmen Inc. v. United States, No. 23-1877 (Fed. Cir. Apr. 22, 2025), the Federal Circuit vacated and remanded the Department of Commerce’s antidumping duty (AD) determination for utility-scale wind towers from Canada. The decision raises important questions about Commerce’s use of statistical methods in calculating dumping margins, and it reinforces key limitations previously addressed in Stupp Corp. v. United States, 5 F.4th 1341 (Fed. Cir. 2021).
Background
In response to a 2019 AD petition by the Wind Tower Trade Coalition, Commerce assigned Marmen a 4.94% dumping margin on Canadian wind tower imports. Marmen challenged three aspects of Commerce’s analysis:
- Commerce’s smoothing of steel plate input costs across product types (CONNUMs),
- Rejection of a supplemental correction for currency conversion discrepancies, and
- Use of the average-to-transaction (A-to-T) method instead of the default average-to-average (A-to-A) method, based on the results of the Cohen’s d test in its differential pricing analysis.
The Court of International Trade (CIT) initially remanded the latter two issues but ultimately sustained Commerce’s determination. On appeal, the Federal Circuit affirmed in part and reversed in part.
1. Steel Plate Cost Smoothing: Affirmed
Commerce had weight-averaged steel plate costs for all but one CONNUM (due to a high-thickness surcharge). Marmen contended this violated Commerce’s own test, which considers whether cost differences reflect physical characteristics.
The Federal Circuit upheld Commerce’s methodology, emphasizing that the governing statute, 19 U.S.C. § 1677b(f)(1)(A), allows cost adjustments when reported costs fail to reasonably reflect production realities. It reiterated its reasoning from Dongkuk S&C Co. v. United States, No. 23-1419 (Fed. Cir. Apr. 21, 2025), where it upheld Commerce’s smoothing of wind tower input costs based on similar factors. Importantly, the court confirmed that Commerce may base smoothing decisions on input characteristics, not just finished product similarity.
2. Currency Conversion Correction: Rejection Overturned
Marmen sought to correct an error in its cost reconciliation stemming from a failure to convert USD purchases into CAD for half of the period of investigation. Commerce had rejected the correction as duplicative or unreliable.
The Federal Circuit disagreed, finding no substantial evidence for Commerce’s claim that the correction would result in double counting. The adjustment was clearly demarcated in a new line item (L1) and not reflected elsewhere. Moreover, the court rejected arguments that the conversion rate was insufficiently supported, noting consistent documentation across the record.
3. Cohen’s d Test Use: Rejected as Unreasonable
The most consequential part of the decision addresses Commerce’s use of Cohen’s d test to justify switching from the A-to-A to the A-to-T methodology. Marmen argued that Commerce’s application failed to meet the statistical assumptions necessary for the test: normal distribution, equal variance, and sufficient sample size.
Building on Stupp, the Federal Circuit reaffirmed that Cohen’s d cannot reliably indicate a meaningful difference in pricing patterns when its assumptions are violated. The court expressly rejected Commerce’s argument that the use of population data (rather than samples) obviated the need to adhere to those assumptions. A flawed coefficient, even if consistently calculated, cannot support a legally sound shift in methodology.
The court clarified that while Commerce may develop alternative statistical methods to evaluate price differences, any such method must be demonstrably sound for the data at issue. Consequently, the dumping margin was vacated, and Commerce must reassess its differential pricing analysis on remand without reliance on Cohen’s d in its current form.
Key Takeaways
- Cost Averaging Permitted: Commerce may smooth input costs across CONNUMs when such costs fail to reflect true production costs—even when finished products are not identical.
- Procedural Fairness Required: Legitimate corrections to financial reconciliations, especially when supported by documentation and material to the accuracy of cost data, cannot be dismissed without evidentiary support.
- Limits on Statistical Tools: The Federal Circuit has now twice underscored the limitations of Cohen’s d test (Stupp and now Marmen). Unless Commerce can justify its application in line with established statistical standards, the test cannot be used to shift methodologies in dumping margin calculations.
This opinion reinforces judicial scrutiny over the use of statistical heuristics in trade remedies and may prompt a recalibration of Commerce’s approach to differential pricing analyses in future AD proceedings.