Cohen’s D test – An insight into the statistical tool used by the USDOC for determining targeted dumping

02 七月 2024

by Shyam Matanhelia


In international trade law, ‘targeted dumping’ refers to a specific trade practice where exporters sell their goods at an export price which differs significantly among different purchasers, regions or time periods. This unfair trade practice has the potential to harm specific regions of the importing country’s market. For this reason, Article 2.4.2 of WTO’s Anti-Dumping Agreement permits the investigating authority to employ specific methods for determination of dumping margin in case of targeted dumping.

The first sentence of Article 2.4.2 identifies two methodologies for calculating the dumping margin that are ordinarily to be used: (1) the comparison of weighted average normal value with weighted average export price (‘W-W methodology’); and (2) the comparison of transaction-wise normal value with transaction-wise export price (‘T-T methodology’). The second sentence of Article 2.4.2 identifies a methodology for calculating the dumping margin if the investigating authority finds a pattern of export prices which differ significantly among different purchasers, regions or time periods, viz. in the case of targeted dumping: the comparison of weighted average normal value to prices of individual export transactions, called the ‘W-T methodology’. 

However, before employing the W-T methodology for determining the dumping margin in case of targeted dumping, the investigating authority is required to establish that targeted dumping is indeed taking place. This is usually done by employing a statistical method for identifying the difference in the pattern of export prices. 

In the United States, the United States Department of Commerce (‘USDOC’) employs the ‘Cohen’s D test’. Named after the American statistician, Jacob Cohen, the Cohen’s D test is an important statistical tool that is used for determining the standardised difference between the statistical means of two groups of data. The USDOC uses this statistical tool for determining the standard deviation of the means of export prices between different purchasers, regions or time periods to identify targeted dumping by exporters to the USA.

This article examines how the Cohen’s D test is utilized by the USDOC for analyzing the existence of targeted dumping.

What is Cohen’s D test?

‘Cohen’s D’ is computed by dividing the difference between means (weighted average) of the two data sets, i.e., test group and comparison group, by the pooled standard deviation. It is a statistical metric that quantifies the effect size, indicating the magnitude of difference between two groups i.e., test group and comparison group.

Below is the formula of calculating the ‘effect size’ of Cohen’s D test:

d = (M1 - M2) / SD pooled


M1 and M2 are the means of test group and comparison group, and

SD pooled is the pooled standard deviation of the test group and comparison group

In the context of targeted dumping, Cohen’s D test helps the USDOC in determining if substantial price differences exist between different purchasers (based on customer codes), regions (based on grouped destination zip codes) and time periods (based on the quarter within the period of review) that may indicate potential targeted dumping. The test group typically consists of export prices of the article under investigation sold by an exporter to a particular purchaser, region or time period in the US market, which are under investigation for potential targeted dumping. On the other hand, the comparison group comprises of exports prices under all other export transactions of that exporter to the US market excluding the transactions under the test group of purchaser, region or time period. By comparing the test group with the comparison group, Cohen's D test assesses the magnitude of difference in export prices between the two groups. This statistical analysis is crucial in determining whether significant price differentials exist that could warrant anti-dumping duties to be imposed on targeted basis.

Application of Cohen’s D test

  1. Calculating the Cohen’s D for export sale transactions

The first step in applying Cohen’s D test is to identify the existence of deviation in the export prices. The USDOC does this by employing Cohen's D test in its differential pricing analysis. This involves comparing prices between test groups and comparison groups to identify the patterns that suggest deliberate pricing strategies. The extent of the difference between the test group and the comparison group can be quantified by one of three thresholds defined by Cohen’s D test:

  1. small (0.2)
  2. medium (0.5)
  3. large (0.8)

Steps for computing Cohen’s D

  • Grouping of export sales database: The export sales data to the USA is divided into different groups based on purchasers, regions and time periods.
  • Calculating mean and standard deviation: For each group, the USDOC calculates the mean and standard deviation.
  • Computing effect size (d): The effect size (d) is calculated for each comparison between groups
  • Computing significance: The USDOC assesses whether the effect size (d) is greater than or equal to 0.8.

The USDOC considers the difference or deviation in export prices to be significant if the calculated effect size (d) is equal to or exceeds the large (i.e., 0.8) threshold. These transactions “pass” the Cohen’s D test. A transaction may pass the Cohen’s D if it passes the test either with respect to purchaser, or region or time period. If a transaction fails to pass the Cohen’s D test on all three fronts, it is deemed to not pass the test.

It may be noted that Cohen’s D can be computed only when the test group and comparison group satisfy the following conditions:

  • The test group and comparison group for a particular purchaser, region, or time period each have at least two observations; and
  • The sales quantity for the comparison group accounts for at least 5 percent of the total sales quantity of both the groups taken together.
  1. Calculating the ratio of transactions passing or failing the Cohen’s D

The USDOC then calculates the percentage of sales values of transactions passing the Cohen’s D test over the total sales value of all export transactions to the USA and divides them on following basis:

  • 66 percent or more sales value transactions passing the Cohen’s D: In this situation, the deviations in export prices are large enough for USDOC to calculate the dumping margin by employing the ‘W-T methodology’.
  • More than 33 percent but less than 66 per cent sales value transactions passing the Cohen’s D: In this situation, the deviations in export prices are large but not enough to apply W-T methodology on all transactions. The USDOC calculates the dumping margin in this case by employing the W-T methodology only to those transactions which “pass” the Cohen’s D test and employs W-W methodology for transactions that do not pass the Cohen’s D test.
  • 33 percent or less sales value transactions passing the Cohen’s D: In this situation, the deviations in export prices are not large and hence the USDOC calculates the dumping margin by employing the W-W methodology.


The United States has been one of the earliest and heaviest users of trade remedy measures. The USDOC’s use of a statistical method like the Cohen’s D test for determining targeted dumping reflects the United States’ advanced history and experience in anti-dumping investigations.

The Cohen's D test is a crucial tool for the USDOC in identifying targeted dumping. The USDOC relies on this method to detect dumping practices that would otherwise not be detectable under ordinary methods. Understanding the complexities of Cohen's D test and how it is used may help exporters to the United States to formulate appropriate pricing strategies.Top of Form

Bottom of Form


[The author is an Associate in WTO and International Trade Division at Lakshmikumaran & Sridharan Attorneys, New Delhi]

Browse articles