AD Petitioner Joins Suit on 'd' Test at CAFC as Amicus
Antidumping duty petitioner the Committee Overseeing Action for Lumber International Trade Investigations or Negotiations on Aug. 22 moved to file an amicus brief at the U.S. Court of Appeals for the Federal Circuit in a case on the Commerce Department's use of the Cohen's d test to detect "masked" dumping. The committee filed the brief in response to arguments from amici led by the Canadian government, which invoked various academic literature on the use of the test (Mid Continent Steel & Wire v. U.S., Fed. Cir. # 24-1556).
In the case, the Court of International Trade sustained Commerce's use of a simple average of standard deviations in the Cohen's d test denominator in an antidumping duty investigation on steel nails from Taiwan (see 2402120036). The trade court said it couldn't find any fault with Commerce's logic.
The Canadian government, along with a group of eight Canadian lumber exporters, said the Cohen's d denominator represents the variance of the groups of data, which are tested for differences using the d test (see 2406130037). The Canadian parties said if the d coefficient is calculated without a denominator that consistently and accurately standardizes mean differences in the numerator, it doesn't provide useful information about the difference between the groups.
In response, the committee discussed the use of a simple average of standard deviations, along with Commerce's other options, which are a weighted average of the test and control groups under limited conditions or the standard deviation of the groups "comingled together." The academic literature, which addresses how to assign statistical significance to certain variables in sampled data, "is not relevant to Commerce's determination of whether there is a pattern of significant price differences in a full dataset of all U.S. sales by an antidumping respondent along three specific axes," the brief said.
Comingling the data "ignores the distribution of prices within each test group" and is "unreasonable for Commerce's purposes," the brief said. A weighted average is similarly unfit because it "smooths out the impact of unusual price distributions within particular test groups," which are the very distributions that can show patterns of price differences among test groups, the committee argued.
The petitioner then echoed a common line used in defense of the d test: what works for sample data simply isn't applicable to the use of the entire population of data.