Commerce Defends Use of Statistical Test to Detect Masked Dumping in Remand at CIT

The Commerce Department again defended the use of the Cohen's d test as part of its differential pricing analysis to detect "masked" dumping in remand results filed on May 26 at the Court of International Trade. Responding to the court's order instructing the agency to address questions on the use of the test raised by the U.S. Court of Appeals for the Federal Circuit, Commerce said that the appellate court's chief concern -- that the test as used by Commerce did not satisfy certain statistical criteria -- is not applicable in the present case (Marmen Inc. v. United States, CIT #20-00169).

The case concerns the antidumping duty investigation into utility scale wind towers from Canada. In the trade court's opinion, Judge Jennifer Choe-Groves upheld four elements of the investigation, but sent back Commerce's decision to reject respondent Marmen's additional cost reconciliation information and use of the average-to-transaction methodology to account for masked dumping (see 2110220069).

To address the first remanded issue, Commerce said it took another look at Marmen's additional reconciliation information that it originally rejected, which contained an additional reconciling item over converting purchases of sections from Marmen from U.S. dollars to Canadian dollars. However, the agency still turned down the option to use the information since it adjusts for amounts already accounted for in costs reported to Commerce. The agency said it still adjusted for amounts already attributed to Marmen's cost of production (COP).

"We have obtained additional cost reconciliation information and have evaluated all of the submitted information and find that Marmen’s additional cost reconciliation item was already reflected in its audited financial statements," the remand results said. "Based on the evidence on the record, it appears that Marmen’s proposed additional reconciling item would duplicate an adjustment amount that was already reflected in its revised audited financial statements, which is the basis for a respondent’s reporting of costs under section 773(f)(1)(A) of the Act. For these reasons, it is not necessary or accurate to adjust Marmen’s reported [cost of manufacturing (COM)] for the item. Therefore, we have not adjusted Marmen’s COM and COP for this redetermination."

As for the differential pricing analysis, or DPA, Commerce continued to use its average-to-transaction comparison for Marmen, releasing a stouter defense of the Cohen's d test. In July 2021, the Federal Circuit called the use of the d test into question in Stupp Corp. v. U.S., finding that Commerce did not satisfy certain statistical assumptions, such as normal distribution and roughly equal variances, before running the test (see 2107150032). Choe-Groves remanded Commerce's use of the test in the present case to address the questions raised in Stupp.

On remand, Commerce said that these statistical assumptions are not relevant to its use of the Cohen's d test since it does not use sampled data, which would require these statistical assumptions, but uses the entire data set instead. The agency made the same defense in the Stupp case, which currently sits before the trade court (see 2204050058).

Commerce said that the statistical criteria cited in academic literature such as normal distribution are related to the statistical significance of sampled data and establish the reliability of an estimated parameter based on the sample data. But no such estimation of the parameters exists for the Cohen's d test when applied to the DPA since the calculation of the parameters is based on the "complete universe of sales prices to the test and comparison groups," Commerce said.

"Unlike with a sample of data where the estimated parameters will change with each sample selected from a population, each time these parameters would be calculated as part of Commerce’s Cohen’s d test, the exact same results would be found because the calculated parameters are the parameters of the entire population and not an estimate of the parameters based on a sample," the remand said.

"Accordingly, the means, standard deviations, and Cohen’s d coefficients calculated are not estimates with confidence levels or sampling errors as would be associated with sampled data, but, rather, are the actual values which describe a company’s pricing behavior. Consequently, the statistical significance of the results of the Cohen’s d test is not relevant in Commerce’s application of the differential pricing analysis, which measures practical significance."