Canadian Gov't Tells Federal Circuit US Defense of 'd' Test in AD Cases 'Complete Nonsense'

No explanation could make the Commerce Department's use of the Cohen's d test, used to root out "masked" dumping, reasonable, the Canadian government and eight Canadian exporters argued in a proposed amicus brief at the U.S. Court of Appeals for the Federal Circuit. Seeking to tack their arguments onto the case in which the appellate court originally questioned the agency's use of the test, the amici said that Commerce is not using the d test "in any coherent sense" (Stupp Corp. v. U.S., Fed. Cir. # 23-1663).

The agency plugs in the wrong inputs but insists the outputs "mean the same thing as when the right inputs are used," the brief said. It called Commerce's principal defense of the use of the test -- that various statistical assumptions need not apply because it used the whole population of data instead of a sample -- "complete nonsense."

Making their claims following exporter SeAH Steel Corp.'s opening brief (see 2307260057), the proposed amici said there is no merit to the claims that the assumptions underlying the test do not apply. Echoing SeAH, the Canadian government argued the agency's defense of the test is a "red herring" and said the test is "demonstrably sensitive to the assumptions of normality, equal variance, and equal size."

Commerce said the assumptions come into play when statistician Jacob Cohen, the test's founder, is considering the extent to which two sets of data do not overlap. The agency argues that it "has a different use for Cohen’s d," the Canadian government said. "This argument makes no sense."

The agency acknowledged that to "quantify the amount of non-overlap," the areas under each bell curve must be known, which requires the statistical assumptions cited by Cohen. Commerce then says the non-overlap measurements in analysis of sampled data do not restrict the real-world observed differences used by Cohen to define his thresholds. This is belied by Cohen's own reference to "populations" instead of "sampled data," the brief said, adding that Commerce's positions "have the merit of neither coherence nor consistency."

While the Court of International Trade sustained the use of the Cohen's d test in the antidumping duty investigation on welded line pipe from South Korea contested here, the Canadian government said Commerce "cannot launder the tainted results" of the "unprincipled" test in the other steps of the differential pricing mechanism. The agency said the ratio and meaningful-difference tests conducted as part of this analysis absolve the need to satisfy the statistical assumptions typically required to run the test.

The Canadian government said this claim does not work, adding that the three tests conducted as part of the differential pricing mechanism "do not operate as an amalgamated whole," but instead serve a distinct function. If the Cohen's d test does not identify price differences between a particular region or time period or any other factor, "then the subsequent tests do not perform any statutory function," the brief said.